Sustaining the Peace After Civil War Pub819

8/7/2019 Sustaining the Peace After Civil War Pub819

1/117

SUSTAINING THE PEACE AFTER CIVIL WAR

T. David Mason

December 2007

This publication is a work of the U.S. Government as dened

in Title 17, United States Code, Section 101. As such, it is in thepublic domain, and under the provisions of Title 17, United StatesCode, Section 105, it may not be copyrighted.

Visit our website for other free publication downloadshttp://www.StrategicStudiesInstitute.army.mil/

To rate this publication click here.
http://www.strategicstudiesinstitute.army.mil/http://www.strategicstudiesinstitute.army.mil/pubs/display.cfm?pubID=819http://www.strategicstudiesinstitute.army.mil/pubs/display.cfm?pubID=819http://www.strategicstudiesinstitute.army.mil/


2/117

ii

*****

The views expressed in this report are those of the authorand do not necessarily reect the ofcial policy or position of theDepartment of the Army, the Department of Defense, or the U.S.Government. This report is cleared for public release; distributionis unlimited.

*****

This manuscript was funded by the U.S. Army War CollegeExternal Research Associates Program. Information on thisprogram is available on our website, www.StrategicStudiesInstitute.army.mil, at the Publishing button.

*****

Comments pertaining to this report are invited and should beforwarded to: Director, Strategic Studies Institute, U.S. Army WarCollege, 122 Forbes Ave, Carlisle, PA 17013-5244.

*****

All Strategic Studies Institute (SSI) publications are availableon the SSI homepage for electronic dissemination. Hard copiesof this report also may be ordered from our homepage. SSIshomepage address is: www.StrategicStudiesInstitute.army.mil.

*****

The Strategic Studies Institute publishes a monthly e-mailnewsletter to update the national security community on theresearch of our analysts, recent and forthcoming publications, andupcoming conferences sponsored by the Institute. Each newsletteralso provides a strategic commentary by one of our researchanalysts. If you are interested in receiving this newsletter, pleasesubscribe on our homepage at www.StrategicStudiesInstitute.army.

mil/newsletter/.

ISBN 1-58487-331-0


3/117

iii

FOREWORD

Since the end of World War II, there have been fourtimes as many civil wars as interstate wars. For a smallsubset of nations, civil war is a chronic condition: abouthalf of the civil war nations have had at least two andas many as six conicts. The author of this monograph,Dr. David Mason, seeks to spell out what social scienceresearch can tell us about how civil wars end and whatpredicts whether (and when) they will recur. Aftersummarizing research on what factors dene the riskset of nations that are susceptible to civil war onset,he presents an analytical framework that has beenused, rst, to explain and predict how civil wars endwhether in a government victory, a rebel victory, or anegotiated settlementand, second, whether the peace

will last following the termination of the conict (or,alternatively, the nation will experience a relapse intocivil war). Research suggests that the outcome of theprevious civil warwhether it ended in a governmentvictory, a rebel victory or a negotiated settlementaswell as the duration and deadliness of the conict,affect the durability of the peace after civil war.

The international community can reduce theprospects for a resumption of armed conict by 1)introducing peacekeeping forces, 2) investing ineconomic development and reconstruction, and 3)establishing democratic political institutions tailoredto the conguration of ethnic and religious cleavagesin the society. The author closes by applying thesepropositions in an analysis of the civil war in Iraq: What

can be done to bring the Iraq conict to an earlier, lessdestructive, and more stable conclusion?


4/117

iv

The Strategic Studies Institute is pleased to publishthis work as part of our External Research Associates

Program.

DOUGLAS C. LOVELACE, JR.Director

Strategic Studies Institute


5/117

v

BIOGRAPHICAL SKETCH OF THE AUTHOR

T. DAVID MASON is is the Johnie Christian FamilyPeace Professor at the University of North Texas andEditor in Chief of International Studies Quarterly. He hasheld faculty positions at Mississippi State University(1981-92) and the University of Memphis (1992-2002).He is the author of Caught in the Crossre: Revolution,Repression, and the Rational Peasant (Rowman &Littleeld 2004) and co-editor (with James Meernik)of Conict Prevention and Peace-building in Post-WarSocieties: Sustaining the Peace (Routledge, 2006) as wellover 40 book chapters and journal articles in suchjournals asAmerican Political Science Review, Journal ofPolitics, Political Research Quarterly, International StudiesQuarterly, Journal of Conict Resolution, Journal of Peace

Research, Terrorism and Political Violence, Social ScienceQuarterly, Public Choice, and Comparative PoliticalStudies. Dr. Mason holds a Ph.D. in Political Sciencefrom the University of Georgia.


6/117


7/117

vii

SUMMARY

Without exception, every widely used data seton civil wars indicates that once a civil war ends in anation, that nation is at risk of experiencing another oneat a later date. I will present a conceptual frameworkthat allows us to identify the factors that make thepost-civil war peace more likely to break down into aresumption of civil war.

Alternatively, this framework will allow us topoint to those factors that make the post-civil warpeace more durable. Many of these factors are policy-manipulable variables: there are policy tools at thedisposal of the international community that caninoculate a post-civil war nation against the prospectsof a relapse into renewed civil war. The analytical

framework that informs the analysis suggests thatthe outcome of the previous civil warwhether itended in a government victory, a rebel victory, or anegotiated settlementas well as the duration anddeadliness of the civil war affect the durability of thepeace after civil war. In addition, characteristics of thepost-civil war environmentthe extent of democracy,the level of economic development, and the degree ofethnic fractionalizationalso affect the durability ofthe peace.

Finally, there is a set of policy interventions at thedisposal of the international community that can bedeployed to enhance the prospects of sustaining thepeace. These include the introduction of peacekeepingforces, modest levels of investment in economic

development and reconstruction, and supporting theestablishment of a set of democratic political institutionsthat are tailored appropriately for the particular


8/117

viii

conguration of ethnic and religious cleavages inthe society. One critical nding from several recent

studies is that the longer the peace lasts, the less likelyit is to break down into renewed conict, regardlessof the characteristics of the society, its economy, or itspolitical system. Therefore, the critical task is to bringthe conict to an end and take the steps necessary tosustain it past the rst few years, after which the peacebecomes increasingly self-sustaining.

This analysis will not only review the evidence onwhat factors account for the duration of the peace (or,conversely, the prospects for renewed war), it will alsooffer theoretically grounded explanations of why wewould expect each factor to have the effect that it doeshave on the durability of peace following civil war.These propositions will be illustrated with examplesfrom specic cases. The analysis will conclude with a

discussion of policy implications: what can be doneto bring civil wars to an earlier and less destructiveconclusion and prevent them from recurring, and howcost effective these policy interventions are comparedto the cost of continued or renewed conict.


9/117


10/117

1


It is widely recognized that over the last half century,civil warrevolution, secessionist conict, and ethno-religious conicthas replaced interstate war as themost frequent and deadly form of armed conict in theinternational system. The Correlates of War (COW)Project, the long-standing armed conict data archiveproject, reports that there were only 23 interstate wars

between 1945 and 1997, resulting in 3.3 million battledeaths. By contrast, there were more than four times asmany civil wars (108), resulting in almost four times asmany casualties (11.4 million).1 While COW includesonly major armed conicts, the Armed Conict Dataset(ACD) compiled by the International Peace ResearchInstitute of Oslo (PRIO) and Uppsala University codes

major, minor and intermediate conicts.2

Of the 231incidents identied in ACD as occurring between1946 and 2005, 167 were internal conicts, 21 wereextrastate conicts (mostly anticolonial wars), andonly 43 were interstate wars.3

To date, the end of the Cold War has not broughtmuch relief from the epidemic of civil wars. Harbom,Hgbladh and Wallensteen report that since 1989 there

have been 121 conicts in 81 locations. Only seven ofthose conicts were interstate wars; the rest were civilwars.4 What the end of the Cold War did bring was thediffusion of civil war to Yugoslavia and the republicsof the former Soviet Union. Following the dissolutionof the Soviet Union into its constituent republics, civilwars erupted in the former Soviet republics of Georgia,

Azerbaijan, Moldova, Tajikistan, and Russia itself.At the same time, the relatively peaceful secession ofSlovenia from Yugoslavia was followed by secessionist


11/117

2

revolts in Croatia and Bosnia. Eventually, Yugoslaviaalso dissolved into its constituent republics, with armed

conict continuing in Bosnia, Croatia, and Serbia-Kosovo. During the Cold War, these two nations andEurope generally had been more or less immune toarmed rebellion on the scale of civil war. It is clear fromthese observations that, whether we are consideringthe Cold War era or its aftermath, armed conict since1945 has been largely a matter of civil war.

What is less often recognized about this same periodis that once a nation experienced one civil war, it washighly likely to experience another one. The 108 civilwars in theCOW data set occurred in only 54 nations.Only 26 of those nations experienced one and only onecivil war, 10 had two civil wars, 12 had three, four hadfour, and two experienced ve civil wars. The 124 civilwars listed in the Doyle and Sambanis data set occurred

in just 69 nations. Only 36 of those nations had one andonly one civil war, while 18 had two separate conicts,nine nations had three, ve nations had four, and onenation had ve.5 In an updated data set, Sambanisreports 151 civil wars occurring in 75 nations, withonly 36 of those nations experiencing one and only onecivil war, 20 nations had two, nine nations had three,four nations had four, ve nations had ve, and onenation (Indonesia) had seven civil wars.6 This leads usto a second conclusion about patterns of armed conictsince 1945: for a certain subset of nations, civil war hasbecome a chronic condition.

That observation raises the question of why it isso difcult to sustain the peace after a civil war. Moreprecisely, what factors inuence whether the peace

established once a civil war ends will endure or,alternatively, the nation will experience a relapse intorenewed civil war? These questions guide the analysis


12/117

3

that follows. We begin with the proposition that thedurability of the peace after a civil war is conditioned,

rst, by how the civil war ended: in a rebel victory, agovernment victory, or a negotiated settlement. Thisimplies that to understand the durability of the peacefollowing civil waror, alternatively, the likelihood ofpeace failure and a resumption of civil warwe mustrst understand what factors determine whether thecivil war ends in a rebel victory, a government victory,or a negotiated settlement.

A body of social science research has identieda set of national attributes that determine a nationssusceptibility to the initial outbreak of civil war.Presumably, these same factors should be implicatedin the failure of peace (i.e., the relapse into renewedcivil conict) following the termination of a civil war.However, characteristics of the previous civil war

itselfincluding its destructiveness, its duration,and the stakes of the conict (e.g., secession versusrevolution, ethnic versus ideological)inuence howthe civil war will end. Independent of the nationalattributes that rendered the nation susceptible to civilwar in the rst place, characteristics of the civil waritself inuence the cost-benet calculations of theprotagonists over the joint decision to continue ghtingor stop. Combining the national attributes that denethe risk set of nations that are susceptible to civil warwith the conict characteristics that predict how thecivil war will end, we can identify a set of factors thatcondition the post-civil war environment in ways thatmake a relapse into civil war more or less likely or,alternatively, affect the capacity of the post-civil war

regime to sustain the peace.The question of how civil wars end points us to a

third, more encouraging trend in the patterns of conict


13/117


14/117

5

more new wars started than ended in the rst 5 yearsof the 1990s. This trend fueled public perceptions

that the post-Cold War era would be fraught withdanger. However, thereafter, the trend reversed: agreater number of wars ended than began duringthe latter half of the 1990s. The trend has continuedinto the new millennium: between 2000 and 2005 thenumber of conicts that ended exceeded the numberof new conicts that began in each year, resulting in anaverage net decline of 1.5 conicts per year.11 The neteffect is that by 2003 there were 40 percent fewer state-based conicts underway than in 1992. Moreover, thenumber of high intensity conicts (1000+ battle deaths)declined by 80 percent between 1990 and 2000.12

The increase in the number of civil war terminationsover the last 15 years has been largely a function of anincrease in the frequency with which civil wars have

been brought to an end by negotiated settlements.Since the end of the Cold War more wars have beenbrought to a conclusion by negotiated settlement (42)than by military victory (23). By contrast, during theCold War, the number of civil wars ending in militaryvictory (by the government or the rebels) was twice aslarge as the number that were concluded by negotiatedsettlements. Hartzell reports that three-fourths of allconicts that ended after 1990 did so by means of anegotiated settlement, whereas a majority of thosethat ended between 1950 and 1990 did so by means ofa military victory by the government or the rebels.13Harbom et al., report that one-third of the 121 conictsthat were active since the end of the Cold War (1989)have been brought to a conclusion by a formal peace

agreement between rebels and government, a rate thatis twice that for the previous 4 decades.14 The trendaccelerated in the new millennium: between 2000 and


15/117

6

2005, 17 conicts ended in a negotiated settlement, whileonly four ended in military victory by the government

or the rebels. In short, since 1990 negotiated settlementhas surpassed military victory as the modal outcomein civil wars. Hartzell also points out that negotiatedsettlements reduce the human costs of civil war byending them sooner. Military victories produced anaverage of 170,706 battle deaths whereas negotiatedsettlements produced only about half that number ofdeaths (87,487) and negotiated truces produced lessthan one-quarter of the death toll (35,182).15 Theseobservations lead to a third conclusion concerningpatterns of conict over the last half century: since theend of the Cold War, more civil wars have been brought toan end by negotiated settlement than by military victory onthe part of the government or the rebels.

The debate over how civil wars endand what the

international community can do to bring them to anearlier and less destructive conclusionhas centeredaround two competing propositions. On the one hand,several studies note that the decline in the numberof ongoing civil wars is largely a function of existingconicts being brought to an end by third partymediation of negotiated settlements to protractedconicts. The implication of this school of thought isthat the best way to reduce the number of conictsgoing on in the world is to build on this trend ofinternational mediation to bring civil wars to an earlierand less destructive conclusion.

On the other hand, another group of scholarsargues that, while brokering settlements to ongoingconicts may bring them to a conclusion for now, peace

agreements all too often preserve the protagonistsorganizational capacity intact and thereby preservethe conditions for a resumption of conict at a later


16/117

7

date. Luttwaks give war a chance thesis contendsthat international mediation of civil wars does little

more than provide breathing space for warring partiesto prepare for the next round of ghting.16 As such,it simply makes recurrence of civil war more likely.Instead, Luttwak contends that it is preferable to givewar a chance: let the warring parties ght it out toa decisive military victory by one side or the otherbecause the decisive defeat of one side makes it lesslikely that civil war will resume in that nation for sometime. In other words, letting them ght it out until oneside achieves decisive victory produces a more durablepeace than brokering a peace agreement between thewarring parties.

Explaining how civil wars end and what factorspredict their recurrence is critical to any effort todevise policy remedies to reduce the frequency and

destructiveness of armed conict. The general patternsof conict make this apparent. First, there is a set ofnational attributes that distinguish those nations thatare at risk of civil war from those that are not. Second,nations that experience one civil war are highly likely toexperience a relapse into armed conict after the initialconict has ended. Therefore, any policy prescriptionsdesigned to reduce the amount of armed conict inthe international community should rst target thosenations that have experienced one civil war with policyinterventions designed to minimize the risk of civil warrecurrence. In order to design such interventions, werst must determine what factors affect the likelihoodof a nation that has had one civil war relapsing intorenewed conict at a later date. Research suggests that

the probability of civil war recurrence is inuenced by(1) the attributes of the nation that put it at risk of civilwar onset in the rst place, (2) the manner in which the


17/117

8

previous civil war endedwhether in a governmentvictory, a rebel victory, or a negotiated settlement, (3)

the attributes of the now-ended civil war that conditionthe post-conict environment in ways that make therecurrence of civil war more or less likely, and (4)attributes of the post-conict environment itself.

Drawing on recent empirical research on civilwars and the larger body of theoretical works onwhat factors make nations susceptible to civil war, Iwill present an analytical framework to assess thesecompeting remedies for bringing civil wars to aconclusion and preventing them from recurring. Ithen use this framework to analyze recent ndings onwhat factors predict how civil wars end and how longthey last. This same framework provides us with someinsights into what factors inuence whether the peacewill endure following the termination of a civil war

or, alternatively, the peace will fail with a relapse intorenewed conict. These insights point to some policyprescriptions for sustaining the peace in the aftermathof civil war. I will conclude with a post-script on whatthis body of research suggests about how to end thewar in Iraq.

DEFINING THE RISK SET: WHICH NATIONSARE SUSCEPTIBLE TO CIVIL WAR?

Research on civil war onset has identied a set ofnational attributes that render a nation more or lesssusceptible to the outbreak of civil war. In effect theydene the risk set of nations susceptible to the outbreakof civil war by specifying what national attributes

distinguish those nations from the large majority ofnations that are generally immune to civil war. Amongthe attributes that dene this risk set are (1) the level of


18/117

9

economic development, (2) the type of political regime(democracy, autocracy, or weak authoritarian), and (3)

the degree of ethnic and religious fractionalization. It isreasonable to expect those same factors to be implicatedin the recurrence of civil war or the failure of the peaceafter a civil war has ended.

Economic Development: Poverty Breeds Confict.

The most consistent and robust nding acrossempirical studies of civil war onset is that economicunderdevelopment (measured as gross domesticproduct (GDP) per capita, infant mortality rate, or lifeexpectancy) is a signicant predictor of civil war onset.Among all nations, those that are the most impoverishedare at the greatest risk of experiencing civil war in agiven year. Conversely, relatively prosperous nations

are largely immune to civil war. Fearon and Laitin,Sambanis, Collier and Hoefer, and others have foundthis relationship to be robust regardless of which civilwar data set one uses or what statistical estimationtechnique or model specication one employs.17 Fearonand Laitin report that $1,000 less in per capita incomeis associated with a 41 percent greater annual odds ofcivil war onset. According to them, the poorest 10percent of nations have an 18 percent chance of civilwar breaking out in a given year, while the wealthiest10 percent of nations have only a 1 percent chance ofexperiencing civil war onset in a given year.18

This nding provides empirical support forgrievance-based theories of civil war: where morepeople suffer from deeper levels of poverty, grievances

are likely to be more widespread and more deeplyfelt, and it is in such environments that civil wars aremost likely to occur.19 However, Collier and Hoeferinterpret this effect as a function of the opportunity


19/117

10

costs of participating in armed rebellion.20 The lowerthe average income in a nation, the lower the recruiting

costs will be for rebel organizations. Where incomeand education levels are low (especially among youngmales), the payoffs from joining a rebel movementexceed what one can expect to earn by devoting onestime to conventional legal economic activity. Thisrelationship is exacerbated by rapid population growththat often characterizes low-income nations. Rapidpopulation growth creates youth bulges whichoverwhelm the supply of legal jobs and provide anever-expanding pool of potential recruits for aspiringrebel movements.21

While the statistical relationship between measuresof poverty and the probability of civil war onset isrobust, there is nothing very surprising about thisnding. There is nothing counterintuitive about the

notion that civil war is more likely to occur in themost impoverished nations of the world. Moreover, itis still the case that, even among poor nations, mostnations in most years do not experience an outbreakof civil war; civil war is still a rare event, in space andtime, even among the most impoverished nations ofthe world. Fearon and Laitins study identies 127new civil war onsets in all nations for all years from1945 through 1997.22 Out of a total of 6,610 nation-years in which a new civil war could have started, inonly 127 of those nation-years did a civil war actuallystart. Therefore, the more challenging task is to specify,among poor nations, what factors distinguish thosethat do experience civil wars from those that do not,and in those that do, what factors determine the timing

of civil war onset.


20/117

11

Regime Type: Democracy vs. Autocracy vs.

Anocracy.

Drawing on the seminal work of Theda Skocpol,state-centric theories of civil war narrow the civilwar risk set by proposing that, among impoverishednations, those governed by certain regime types aremore susceptible to civil war than those governed byother regime types.23 The task then becomes how tospecify the regime types or regime characteristics thatmake a nation (especially an impoverished nation)more or less likely to experience a civil war onset in agiven year.

The consensus is that weak states are more prone toviolent opposition, including civil war. There is lessagreement on what attributes dene a state as weak.

Barry Buzan argues that, weak states either do nothave or have failed to create a domestic political andsocial consensus of sufcient strength to eliminatethe large-scale use of force as a major and continuingelement in the domestic political life of the nation.24The state is seen by one or more signicant socialgroups as representing the interests of a particularethnic group (as is the case with many multiethnicstates) or a particular social sector (such as the agrarianelite in Latin America) or an economic or military elite(as was the case in Nicaragua of the Somoza era orthe Philippines of the Marcos era). Those who are notmembers of the group favored by the state withholdtheir support from the state, either tacitly by neglectingto comply with state laws and regulations and evading

taxes, or actively by organizing opposition movementsto challenge the incumbent regime. Because the stateperceives those alienated social groups as a threat, it


21/117

12

responds by increasing its coercive capacity in orderto defend itself against anticipated challenges to its

authority. The threat of state repression further alienatesmarginalized groups and gives them incentives toorganize for armed rebellion. This cycle escalates intowhat Brian Job has termed an insecurity dilemma.25

Regimes that manifest this weak state syndromehave been labeled neo-patrimonial regimes,26 sultanisticregimes,27 or protection racket states.28 The commonfeature of these regimes is that they typically areheaded by a personalist dictator presiding over a stateapparatus that is staffed not on the basis of competenceand experience but on the basis of personal loyalty tothe dictator.

Goodwin identies ve practices common to weakstates that render them susceptible to armed revolt.This list captures most of the attributes that others have

listed as characteristic of the weak state syndrome.First, state sponsorship of unpopular social and economicarrangementsmakes the state the target for the grievancesthat the extremes of poverty and economic inequalitygenerate.29 These arrangements can be based on classdifferences or ethnic differences. Stanleys protectionracket state is typical of the former: in a nation suchas El Salvador, where export agriculture was thedominant sector of the economy, the military protectedthe interests of a small landed elite from redistributivepressures emerging from the large landless and land-poor peasant population. The military systematicallyrepressed dissent and dissident organizations amongthe peasants, thereby preserving the landed elites incontrol over landed wealth. In return, the military was

allowed to control the institutional machinery of thestate and use it to extract rents from society for thepurposes of enriching the ofcer corps.30 Where ethnic


22/117

13

differences are the basis of the unpopular social andeconomic arrangements, a dominant ethnic group uses

its control over the institutional machinery of the state tofurther subordinate other ethnic groups, economically,politically, and socially, through discriminatory lawsand practices. The dominance of the Hutu majorityin Rwanda under Juvenal Habyarimina or theSinhalese majority in Sri Lanka is exemplary of thisarrangement.31

Second, where a weak state excludes newly mobilizedgroups from access to state power or economic opportunity,it may leave those groups with few alternatives otherthan direct challenges to the states authority.32 Theregime types listed earlier are, as a rule, intolerant ofany sort of grassroots political mobilization. Whencollective dissent does emerge, such states typicallyreact with repression. This leaves even moderate

reformers with few options other than withdrawingfrom politics and suffering in silence or resorting toviolent tactics of their own. Otherwise, those leadersrisk being marginalized among their own constituentsfor being ineffectual. Even the choice of withdrawingfrom politics is not viable because, as known leaders ofan opposition organization, they have to assume thatthey remain on the states list of targets for repressiveviolence. Hence, they have powerful incentivesi.e., the threat of being victims of state-sanctionedrepressionto remain active in opposition politicsbut to shift to violent tactics of their own.33 Repressiontends to radicalize dissent.

Third, when confronted with political opposition,weak states typically respond with indiscriminate but

not overwhelming repressive violence, which tends toradicalize the opposition.34 Mason and Krane arguethat the escalation to indiscriminate violence is highly


23/117

14

likely among weak states, in part because they lackthe institutional capacity or redistributable resources

to pursue more accommodative reform strategies.Moreover, given the origins and composition of suchregimes, they also generally lack the political will topursue reform and accommodation as opposed torepression. Repression is the one policy response forwhich weak states are well-equipped. Therefore, whenconfronted with opposition challenges, they almostreexively employ the resources with which they arebest endowed: the repressive machinery of the state.35

Usually the state begins by targeting oppositionleaders. This compels those leaders who manageto escape the repressive arm of the state to gounderground and shift to violent tactics of theirown. Lacking sufcient numbers to mount insurgentattacks, the small surviving cadre of opposition

leaders often resorts to forms of terrorist violenceintended to provoke the state into expanding itsrepressive targeting, thereby driving more people tothe side of the opposition. If the states initial effortsto decapitate the opposition do not silence it, the weakstate typically responds by expanding repression toinclude rank and le participants in and supportersof opposition organizations and social movements.They target members of labor unions, political parties,peasant associations, and other social organizationsthat have some degree of autonomy from the state,some established constituency, and a record of publicopposition to the state, its leaders, and its policies.When repression becomes more widely targeted,nonelite supporters of opposition movements are then

compelled to go underground as well. This providesthe previously radicalized dissident leadership withthe human resources to escalate terrorist violence


24/117

15

to guerrilla insurgency. Faced with the escalation ofopposition violence, weak states typically respond

by further expanding the targeting of their repressionto include the civilian support base of the insurgentopposition.36

At this point, distinguishing the guerrilla irregularand his/her supporters from the uninvolved civilianpresents the states security forces with the classiccounterinsurgency dilemma.37 Troops in the eld, whoseimmediate goal is to survive the mission, are likelyto target anyone remotely suspected of supportingthe insurgents rather than risk allowing a suspectedinsurgent to escape detection and later kill them. AsLeites and Wolf put it, without adequate intelligenceto allow them to target rebel supporters and only rebelsupporters, government security forces may notfeel too guilty about fullling their professional duty

of spending ammunition.38

From the point of viewof civilians, the indiscriminate application of staterepression means that their chances of being victimizedare largely unrelated to whether or not they actuallysupport the insurgents, actively or tacitly, overtly orcovertly. Under those circumstances, it may becomerational for them to join the insurgents if for no otherreason than to secure protection from indiscriminatecounterinsurgent violence by the states securityforces.39

In this sense, repression by itself can and often doesfail to suppress opposition. Instead, it can instigate theescalation from nonviolent protest to violent oppositionand, eventually, civil war. It may bring about a temporarylull in opposition activity in the early stages, largely

by disrupting the ability of conventional (nonviolent)opposition organizations to mobilize their supporters.However, once a campaign of repression begins, it is


25/117

16

difcult to keep it from becoming indiscriminate. Overthe longer term, as repression escalates, it is likely to

become indiscriminate, which compels oppositionorganizations to shift to violent tactics of their ownand, eventually, to escalate the level of violence fromterrorist acts to low level insurgency to civil war.

Fourth, Goodwin points to weak policing practicesand infrastructural power that enable insurgent groupsto establish security zones within the territorial jurisdiction of the state.40 From secure base areas,insurgents can mount and sustain armed challenges tothe state. There are two components of this dimensionof state weakness. First, if the states policing poweris geographically uneven, then rebels can establishsecure bases of operation in those regions where thestates police presence is weakest. Fearon and Laitinfound evidence that geographic features of a nation

that make it easier for insurgents to establish securebase camps increase that nations susceptibility tocivil war.41 The second component is a function ofthe states relationship with the population. Wherelarge segments of the population are alienated fromthe state, the states power becomes more strictly afunction of its troop strength. It cannot count on thepopulation to provide it with intelligence on rebeloperations. Indeed, all that insurgents need in orderto survive is a population that tolerates their existence,which amounts to a form of tacit support. Leites andWolf observe that, the only act that [the rebel] needsdesperately from a large proportion of the populaceis nondenunciation (that is, eschewing the act ofinforming against R[ebels]) and noncombat against

[them].42 Joel Migdal adds, in the early stages ofrevolution, revolutionaries stake their lives on the hopethat peasants will not expose them to authorities.43


26/117

17

Neo-patrimonial regimes are especially prone to weakpolicing capability because their security forces, like

other state institutions, are staffed according to theirloyalty to the leader, not their competence. As long asthe security forces remain loyal, the leader is usuallytolerant of a certain level of corruption, incompetenceand venality on their part. This simply exacerbates thestates weakness by alienating the civilian populationas a source of intelligence on the rebels and drivingthem to the side of the rebels.

Finally, Goodwin argues that the corrupt andarbitrary rule of neopatrimonial dictators tends toalienate, weaken, and divide elite groups and externalsupporters who otherwise might share the leadersinterest in repressing opposition challenges.44 For thisreason, neopatrimonial regimes are not only susceptibleto revolutionary challenges but also vulnerable to defeat

by them. When an opposition challenge escalates tothe point of posing a threat to the survival of the state,whatever elite coalition has supported the regime canquickly dissolve if elements of that coalition becomedissatised with the dictators distribution of thespoils of rule among his coalition of supporters. Signsof divisions within the elite coalition are often readilyapparent, and insurgents can exploit them by escalatingthe level of violence. A military establishment thathas been deprofessionalized by the corruption that istolerated by the neopatrimonial leader as the price forthe militarys loyalty can quickly dissolve in the faceof an effective rebel challenge, especially when theysee the leaders civilian coalition defecting and hisability to deliver the spoils of patronage eroding. The

sudden collapse of the Somoza regime in Nicaragua,the Mobutu regime in Zaire, and the Barre regime inSomalia illustrate the vulnerability of neopatrimonial


27/117

18

regimes not only to the emergence of armed challengesbut to defeat by them.

The empirical evidence on the susceptibility ofweak states to civil war is generally supportive, thoughhampered by measurement issues. Fearon and Laitinargue that nancially, organizationally, and politicallyweak central governments render insurgency morefeasible and attractive due to weak local policing andcorrupt counterinsurgency practices.45 However, theirstatistical models include no direct measure of theseaspects of the weak state syndrome. They add that weakstates have a propensity for brutal and indiscriminateretaliation that helps drive noncombatants into rebelforces, an argument that echoes Mason and Kranestheory about the impact of escalating repression onthe distribution of popular support between the stateand the opposition. However, Fearon and Laitins

models contain no direct measure of this weak statecharacteristic either. Indeed, their primary measure ofstate weakness is income per capita, which most theoriesof civil war onset treat as a measure of grievances46 oropportunity costs,47 not state strength.

The more common test of the relationship betweenstate strength and civil war is the domestic version ofthe democratic peace proposition: that democracies areless susceptible to civil war than are nondemocracies.Numerous studies have tested this proposition,employing the 21-point (+10 to -10) POLITY IVdemocracy-autocracy scale. States with scores of 7or more on this scale are treated as full democracies,while those with scores of -7 or below are treated asfully autocratic regimes. Both fully democratic and

fully autocratic states are treated as strong states, atleast in the sense of their capacity to avoid civil war. Itis the middle range of weak authoritarian regimes


28/117

19

(-6 to 0), semi-democracies (0 to +6), or (generally)anocracies (-6 to +6) that are alleged to be the most

prone to civil war.At one end of the scale, democracies are less likely to

experience civil war because civil war is not necessaryfor the opposition to have its concerns accommodated(or at least considered) by the state.48 Under democracy,opposition groups are free to organize for peacefulcollective action, to form their own political partiesand run candidates for ofce, and otherwise to engagein a variety of forms of peaceful collective action toseek redress of grievances or to secure the enactmentof their preferences into policy. And they are free to doso without fear of state repression. Elections confrontpolitical leaders with incentives to accommodatepopular demands in order to expand their vote share.Those same electoral incentives discourage state leaders

from employing repression against a loyal opposition,lest those leaders suffer the repercussions at the polls.

At the other end of the scale, fully autocratic regimesare also unlikely to experience civil war because theypossess the overwhelming coercive capacity to repressopposition movements preemptively. In autocracies,rebellion is irrational because the coercive capacity ofthe state is so overwhelming that dissident movementsare crushed before they can mobilize any base of popularsupport. Citizens are intimidated into withholdingsupport for or participating in such movements forfear of the severe repressive consequences.49

It is that middle range of weakly authoritarianregimes or semi-democracies that are most prone tocivil war because they lack the institutional capacity

to accommodate peaceful opposition movements orthe coercive capacity to repress them preemptively.The ndings on the democracy/autocracy-civil war


29/117

20

relationship are mixed, but generally, there is supportfor this inverted-U relationship: fully democratic

regimes and highly autocratic regimes are less likely toexperience civil war, while weak authoritarian regimesand semi-democracies are most susceptible to civilwar.50

A critical addition to this hypothesis is the ndingby Hegre et al., that new democraciesi.e., regimes thathave recently undergone the transition to democracyare especially susceptible to civil war. Indeed, changein a nations level of democracyregardless ofwhether it is becoming more democratic or moreautocraticappears to be especially destabilizing.51New democracies may have the formal institutions toaccommodate dissident interests in a peaceful manner,but it takes time for a civic culture to emerge wherebythe population views democratic processes as the only

game in town. Until a stable party system evolves,elections create space for anti-democratic demagoguesto run for ofce and win. Unchecked by an effectiveand institutionalized loyal opposition, such leaderscan use the power of elective ofce to attack rivalleaders and their parties and gradually but inexorablytransform a edgling democracy into what FareedZakaria has termed an illiberal democracy thatsuccumbs to the perverse principle of one man, onevote, one time.52 Such regimes are susceptible to civilwar, despite the democratic facade that elections conferupon them. In Zimbabwe, once Robert Mugabe wonthat nations rst presidential election, he attempted toenact legislation to make Zimbabwe a one party state.When that failed (due to constitutional constraints

established by the Lancaster House Agreement thatended the civil war), he accused his chief rival, JoshuaNkomo, of plotting an insurrection and unleashed a


30/117

21

campaign of repression against Nkomo, his party, andhis ethnic Ndebele support base. Zimbabwe has since

degenerated into a virtual dictatorship that maintainsonly the thinnest veneers of democratic appearances.The empirical research discussed so far would suggestthat Mugabes rule has put Zimbabwe rmly in therisk set of nations susceptible to civil war.

Ethnic Divisions.

Among impoverished nations, those in whichthe population is fragmented along ethnic lines areespecially susceptible to civil war. Indeed, ethnicfragmentation contributes to state weakness as well. Inethnically divided societies, the state itself can becomethe spoils over which ethnic groups compete. The stateoften does not command the support and loyalty of one

or more ethnic groups. This is especially true wherethe state becomes dominated by one ethnic group tothe exclusion of others. Excluded ethnic groups cometo view the state as predatory, unresponsive to theirinterests at best and threatening to their ethnic identityat worst. Under these circumstances, the state comesto see itself as threatened by the excluded groups. Asa result, a domestic security dilemma can emerge,whereby the state and excluded ethnic groups arm inorder to defend themselves against the other, and eachinterprets the others actions as a threat that warrantsfurther arming.53

Shared ethnic identity serves as a powerful basis formobilizing supporters for collective action. Dissidentleaders can frame grievances in ethnic terms. Shared

ethnic identity also facilitates recruitment by insurgentorganizations. Dissident leaders can target theirrecruitment more efciently to the extent that ethnic


31/117

22

cleavages dene the grievances that motivate rebellion.Shared ethnic identity also facilitates the identication

and sanctioning of free riders in that defectors from arebel movement can be identied by ethnic markersand sanctioned for not supporting the movement.

The ndings on the relationship betweenethnic fragmentation and the onset of civil war aresurprisingly mixed. Most studies employ a versionof the Ethnolinguistic Fractionalization Index (ELF)which uses the number and relative size of each ethnicgroup in a nation to calculate an index that estimatesthe probability that two randomly chosen individualswould be from different ethnic groups.54 Theoreticalarguments for the impact of ethnic fractionalizationusually propose an inverted-U relationship betweenELF and the likelihood of conict: conict is least likelyin ethnically homogeneous societies and in those that

are fragmented among a relatively large number ofsmall ethnic groups, while ethnically based conictis most likely in societies that are divided between asmall number of relatively large ethnic groups.

Where society is composed of a large number ofrelatively small ethnic groups, no single group hassufcient numbers to threaten the establishment ofethnic hegemony over the other groups. Ethnic securitydilemmas that would motivate groups to mobilizeand arm defensivelyand thereby motivate a similarresponse on the part of other ethnic groupsare lesslikely to arise because no single ethnic group is largeenough to pose a threat of ethnic dominance. Collierand Hoefer add that in highly fragmented societies,coordination problems between ethnic groups reduce

the likelihood that multiple ethnic groups will be ableto form a coalition of sufcient magnitude to mountand sustain a major rebellion.55 Each group has little


32/117

23

incentive to devote much of its collective resources topolitical activities beyond its own communal borders.56

The state is more able to accommodate the demandsof one group without threatening the interests of theothers.

By contrast, where there are fewer groups andone or more is sufciently large in number (relativelyand absolutely) to aspire to ethnic hegemony, ethnicsecurity dilemmas are more likely to arise, makingconict more likely.57 If one group mobilizes to assertits control over the machinery of the state, othergroups are likely to react defensively by mobilizingand perhaps arming themselves to prevent that or todefend their group against subordination by the groupaspiring to dominance.

Elbadawi and Sambanis did nd support for aninverted-U relationship between the degree of ethnic

fractionalization and the probability of civil war.58

Elbadawi and Reynol-Querol found that ethnicallypolarized societies (i.e., those divided between twoethnic groups) have a greater risk of experiencingcivil war.59 Similarly, Collier and Hoefer did nd arelationship between civil war onset and a conditionof ethnic dominance, dened as a nation in whichthe largest ethnic group constitutes between 45 and 90percent of the population.60 Ellingsen also found thatsocieties that were divided among a relatively smallnumber of relatively large groups were more likely toexperience civil war.61 Her key measure was the relativesize of the second largest ethnic group. Cederman andGirardin found that governments controlled by ethnicminorities are more likely to experience civil war, and

the smaller the ratio of the dominant ethnic groupssize to a challenger groups size, the more likelycivil conict is to arise between those two groups.62


33/117

24

However, neither Fearon and Laitin nor Collier andHoefer found much support for a direct relationship

between the degree of ethnic fractionalization and theprobability of civil war.63

These ndings suggest that ethnic civil war ismore a function of the ability of groups to mobilize forviolent collective action than of the depth of the ethnicgrievances that motivate rebellion.64 The extent towhich an ethnic group is concentrated geographicallystrongly affects its ability to mobilize.65Ethnic minoritiesthat are concentrated in their own territorial enclaveare less subject to monitoring and repression by rivalethnic groups than are groups that are interspersedamong other ethnic groups (including a dominantethnic group). Geographic concentration also makesit easier for the group to establish secure base campsfrom which to organize and sustain an armed rebellion.

Geographic concentration also facilitates the detectionand sanctioning of free riders among the members ofthe group.66

Resource Wars: Do Oil, Drugs, and Gems FuelConfict?

A recent addition to the civil war research programhas been the resource curse hypothesis: nations that areheavily dependent on mineral exports as a source ofnational income are especially susceptible to civil war.There are two streams of research that come out ofthis program. The rst is that oil-exporting nations areprone to state weakness and, therefore, civil war. Oilwealth increases the value of controlling the state and,

as such, creates incentives for rebel groups to emergeand challenge the incumbent government for controlof the state.67 Similarly, oil wealth creates incentives


34/117

25

for regional ethnic groups to launch secessionist warsintended to wrest control of oil-rich regions from the

existing regime.68

The second theme in this literature is the greedhypothesis, championed by Paul Collier and hiscolleagues at the World Bank. They propose that civilwar is driven not so much by grievance as by greed. Thatis, civil war is more likely where rebel organizationshave access to lootable commodities, such as illegaldrugs or gemstones. What makes these commoditiesvaluable for rebels is that they can be produced onlyin limited geographic regions. Only some nations havedeposits of gemstones, and those deposits are locatedin very specic regions of those countries. Illegal drugssuch as opium and coca can only be grown in certainclimates, altitudes, and soil types. Where rebels cancapture the territory where such commodities are

produced or control the supply routes from productionsites to markets, they can extract rents from this sectorof the economy that they can use to nance theirrebellion.

In Peru, Shining Path guerrillas provided cocagrowers with protection from drug eradication effortsof the government. They also provided protection for anumber of clandestine landing strips in remote regionsof the Andean highlands where drug cartels could y inplanes to transport coca leaf or coca paste to laboratoriesoutside the country. Protection fees from coca growersand landing fees from drug cartels produced a revenuestream that enabled the rebel organization to equipand pay guerrilla soldiers.69 Revolutionary ArmedForces of Colombia (FARC) guerrillas have developed

a similar symbiotic relationship with coca growersthere, as has the Taliban (and, before them, regionalwarlords) with opium poppy growers in Afghanistan.


35/117

26

Rebels in Congo/Zaire, Sierra Leone, and Angola havesustained their operations with revenues from alluvial

diamonds.Empirical support for the greed hypothesis is

somewhat mixed. A stronger case can be made forlootable resources contributing to the duration of civilwar rather than to the onset of civil war.70 Fearon ndsthat the availability of lootable resources is positivelyassociated with the duration of civil wars, especiallysecessionist conicts in peripheral regions of a nationwhere the resources are located.71

A related element of the resource curse thesis isthat oil exporting nations are particularly susceptibleto civil war, especially of the secessionist variety. Thelogic underlying this relationship is that the rents thatcan be derived from oil exports create incentives forrebel groups to contest over control of the state or

for regional groups to seek secession in the hopes ofgaining monopoly control over oil-rich regions in acountry.72 Ross adds that the same incentives can induceexternal intervention in civil wars, as was the case withLiberian President Charles Taylors intervention intothe conict in Sierra Leone.73 Fearon and Laitin arguethat oil wealth contributes to state weakness, which inturn makes civil war more likely.74 States that derivesignicant rents from oil revenues have the capacityto buy popular quiescence by providing extensivesocial welfare benets, without investing in developinga growth economy that would be sustainableeven without oil revenues. Humphreys adds thatdependence on mineral exports can retard the growthof domestic commerce, thus making the nation more

vulnerable to external shocks, such as rapid declinesin the price of oil.75 Leaders of oil-rich nations canalso use the rents from oil exports to nance extensivethe coercive machinery necessary to repress political


36/117

27

opposition, a practice that (as discussed earlier) canhave the effect of transforming nonviolent dissent into

revolutionary opposition (especially where the payoffsfrom rebel victorycontrol over oil revenuesare sosubstantial).

Evidence on the resource curse explanation ofcivil war onset is mixed, with results highly dependenton how one measures natural resource dependence andhow one species the dependent variable, civil war.Collier and Hoefer nd that a states dependence onnatural resource exports increases the likelihood thatthe nation will experience a civil war.76 They do ndthat this effect is nonlinear: the probability of civil warincreases up to a ratio of natural resource exports to GDPof 32 percent and declines beyond that point. However,Fearon and Laitin found no signicant relationship(linear or otherwise) between primary commodity

exports and civil war onset, though they did nd thatcountries that derive at least a third of their exportrevenues from oil were twice as likely to experiencecivil war as similar nations that did not export oil.77Elbadawi and Sambanis found some support for thisrelationship but also found that such ndings werehighly sensitive to how the model was specied andwhich civil war data set one employed.78They concludedthat the relationship is fragile at best and certainlynot robust across data sets or model specications.In a later paper, Collier and Hoefer found that anations dependence on primary commodity exportsis more strongly related to the onset of secessionistconicts than revolutionary civil wars,79 but Reynol-Querol presents evidence that revolutionary conicts

(rather than secessionist conicts) are catalyzed by adependence on primary commodity exports.80 Partof the problem with this debate is that these studies


37/117

28

lump together oil, and other minerals and evenagricultural exports under the category of primary

commodity exports, whereas the theories presentedearlier focus separately on the effect of oil exports, onthe one hand, and lootable commodities (such as illegaldrugs or alluvial gemstones), on the other, on civil waronset.

The discussion up to now has surveyed theempirical ndings and theoretical arguments onwhat national attributes dene a risk set of nationssusceptible to civil war. Impoverished nations withweak states dene the broad parameters of this riskset. Among impoverished nations, those governedby neopatrimonial regimes appear to be especiallyvulnerable to civil war. Democracy does appear toimmunize nations against civil war to some degree, butthat effect emerges only after democratic institutions

have been in place long enough to earn some degreeof popular legitimacy and establish some degree ofinstitutional stability. There is some evidence for aresource curse affecting the susceptibility of nations tocivil war, but this effect is probably more catalytic thancausal: oil-exporting nations that manifest the otherrisk factors may be somewhat more likely to experiencecivil war, but only if they manifest those other criticalrisk attributes such as state-weakness and widespreadpoverty. Ethnic divisions exacerbate most of theserisk factors: weak states presiding over impoverishedpopulations that are also ethnically divided are morelikely to experience civil war than similar nations thatare not divided among a small number of relativelylarge ethnic groups. Moreover, the pacifying effects of

democracy are less likely to emergeand democracy isless likely to survivein ethnically divided societies.

With this survey of what makes nations susceptibleto civil wars, we now turn to the question of how civil


38/117

29

wars end and what factors inuence the durability ofthe peace following the termination of civil war.

WIN, LOSE, OR DRAW: HOW CIVIL WARS END

While an extensive body of research has deneda set of national attributes that dene the risk set ofnations susceptible to civil war, there is considerablyless research on how civil wars end or what factorspredict the durability of the peace after civil war. Thisauthor has completed several studies on these subjects,and some ndings appear to hold up across data setsand model specications.

A useful way to think about how civil wars cometo an end is to consider the decision calculus by whichrebels and governments decide whether to stop ghtingor continue to prosecute the war. The model I present is

built on the assumption of two rational actors involvedin a civil war. The rationality assumptions and theassumption of two decisionmakers are, admittedly, anover-simplication of the reality of civil war. However,models such as these are evaluated on the basis ofwhether they enable us to derive some predictionsabout what conditions affect how civil wars end andwhether those predictions are supported by evidencefrom the real world. This decisionmaking model hasbeen used to identify what conditions make a civil warmore likely to end in a negotiated settlement ratherthan a military victory by either side, and the modelcorrectly predicted 87 percent of the outcomes.81 It hasbeen used to predict whether a civil war will end ina government victory, a rebel victory, or a negotiated

settlement, and it correctly predicted 86 percent of theoutcomes in that study.82 Most recently, it was usedto predict whether a nation that had experienced one


39/117

30

civil war would experience a relapse into renewed civilwar, and that model correctly predicted recurrence/

nonrecurrence 85 percent of the time.83

It has also beenused to model the duration of civil wars,84 and theduration of the peace after a civil war.85 Similar logichas also been used to explain how interstate wars end,86and how foreign intervention affects the duration ofcivil wars.87 Thus, whatever one might think aboutthe realism of rationality assumptions or the extent towhich the decision calculus oversimplies the realityof civil war, the model does allow us to develop somepredictions about what factors affect civil war outcome,and those predictions are supported by the empiricalevidence.

At any given point in a civil war, the government (G)and the rebels (R) each must choose between quittingor continuing to ght. This implies four possible

outcomes from their joint decisions at any given time,t

i: (1) if R continues ghting and G quits, R wins and

the government is overthrown; (2) if G continues toght and R quits, G wins and the revolt is defeated;(3) if both G and R choose to quit at the same time,the civil war ends in a truce or a negotiated settlement;(4) if neither decides to quit, the civil war continues.Following Stam, the four outcomes can be representedas an iterated two-person game (see Figure 1), withcontinued ghting as the dominant strategy for bothsides.88 If one or both parties prefer to continue ghting,it must be that they expect either to win at some point inthe future or at least achieve more favorable settlementterms than what they estimate they can secure in thepresent. In either case, as long as one or both parties

expect that their net benets from victory (or a futuresettlement) will exceed the benets they can get from


40/117

31

a settlement now (or from defeat), they have a strongincentive to continue ghting.

GOVERNMENT

Fight Quit

REBELS Fight Civil War Continues Rebels Win

Quit Government Wins Negotiated Settlement

Derived from Allan C. Stam III, Win, Lose, or Draw: DomesticPolitics and the Crucible of War, Ann Arbor, University ofMichigan Press, 1996, p. 35.

Figure 1. Civil War Outcomes as a Function of Rebeland Government Choices.

The decision calculus by which both actors choose

between continuing to ght and stopping is a functionof the expected payoffs from victory versus defeatversus a negotiated settlement.89 The expected payoffsfrom continuing to ght are a function of: (1) the actorssubjective estimate of the total payoffs from victory, (2)the actors estimate of the probability of victory, (3) theactors estimate of the rate at which s/he will have toabsorb the costs of conict if s/he continues to ght in

hopes of eventually achieving victory, (4) the actorsestimate of the amount of additional time needed toachieve victory (and, therefore, the amount of timethat actor will have to absorb the costs of conict inorder to achieve victory). Mason and Fett represent theexpected utility of continuing to ght as follows:

EUC=P

V(U

V)+(1-P

V)(U

D) - C

ti(1)

where EUcis the expected utility of continuing to ght,

Uv

is the actors estimate of the payoff from eventual

ti=0

tv


41/117

32

victory, Pv

is the actors estimate of the probabilityof achieving victory, U

dis the actors estimate of the

cost from defeat, (1-Pv) is the estimated probability ofdefeat, C

tiis the actors estimate of the rate at which the

costs of conict will accrue from the present (t0) to that

time in the future when the actor estimates victory canbe achieved (t

v). Generally, an actor will continue to

ght as long as its expected payoff from victory (PvU

v)

exceeds the costs it expects to absorb in order to achieve

victory ( Cti). However, even if one protagonist

believes its chances of victory are better than even, thatactor may still prefer to seek a negotiated settlement ifits estimate of the cumulative costs required to achievevictory come to approach or exceed its expected payofffrom victory. Under these circumstances, victory, even

though more likely than defeat, would be pyrrhic.For a negotiated settlement to be preferred to

continued ghting, the expected utility of a negotiatedsettlement, EU

s, must be greater than the expected

utility of continuing the conict, EUc. The expected

utility of a negotiated settlement can be represented asfollows:

EUS=P

s(U

S) + ( C

ti) - C

ti(2)

where Usrepresents that actors estimate of the payoffs

from the terms of the settlement and the cost termsare the same as in Equation (1). The payoffs froma settlement (U

s) are presumed to be less than the

payoffs from victory (Uv). However, by agreeing to asettlement now rather than continuing to ght in searchof victory, the actor saves the additional costs of conictthat would have to be absorbed in order to achieve

s

ti=0

tv

ti=0

tv

t

i=0

t


42/117

33

victory ( Cti). Instead, that actor has to absorb onlythose additional costs that accrue between the presentand that time in the more immediate future (t

s) when

the settlement goes into effect and the ghting stops

( Cti; we assume that t

s< t

v).

The logic of this decision calculus implies that

any factor that (1) decreases an actors estimate of theprobability of victory (P

v), (2) reduces that actorss

estimate of the payoff from victory (Uv), (3) increases

the rate at which that actor absorbs the costs ofcontinued conict (C

ti), or (4) extends that actors

estimate of the time required to achieve victory (tv)

should make negotiating a settlement more attractive

than continuing to ght. From this decision calculus,we can derive some propositions concerning thecharacteristics of a civil war that affect the outcome ofthe conict by inuencing one or both partys incentivesto continue ghting, capitulate, or enter negotiationsfor a settlement.

Duration Matters.

The decision calculus implies, rst, that the longer acivil war last, the more likely it is to end in a settlement(as opposed to a military victory by either side). Indeed,one fairly consistent nding on civil war outcomes isthat the longer a civil war lasts, the less likely it is to end ina decisive military victory by eitherthe government or the

rebels. If the rebels win (the least likely outcome), theytypically do so within the rst few years of the conict.Mason, Weingarten, and Fett found that 12 of 16 rebel

s

ti=0

tv

ti=0

t


43/117

34

victories in their data set of 57 civil war terminations(1945-92) occurred within the rst 5 years of the conict.

Similarly, if governments succeed in putting down arebellion decisively, they also usually do so within therst 5 years of the conict. They also found that all butthree of 28 government victories occurred within therst 5 years of the conict.90 If neither side prevails early,the conict settles into a mutually hurting stalemate inwhich neither side has the capacity to defeat the other,but each side has sufcient strength to prevent theirown defeat.91 At that point, the only way out of theconict is through a negotiated settlement. Otherwise,the conict simply drags on interminably. Mason andFett (1996) found that negotiated settlement was by farthe most likely outcome to civil wars lasting more than5 years.92 Fearon found that one-quarter of the civil warsthat occurred between 1945 and 1997 lasted 2 years

or less, and another quarter lasted at least 12 years;13 lasted 20 years or more.93 Consistent with Mason,Weingarten, and Fetts study, he found that those thatended quickly terminated in a decisive victory by oneside or the other while those of long duration endedin a negotiated settlement or simply dissipated afterreaching a protracted stalemate. Fearon concludescivil wars last a long time when neither side can disarm theother, causing a military stalemate. They are relatively quickwhen conditions favor a decisive victory (emphasis in theoriginal).94

These ndings imply that, contrary to EdwardLuttwaks give war a chance thesis, civil wars willnot burn themselves out like brush res, nor will theconditions of a more lasting peace emerge naturally from

the course of the war if the international communitysimply stands aside and allows the protagonists toght it out to a decisive victory by one side or the


44/117

35

other. The decisive military victory that Luttwakclaims will produce a more lasting peace occurs early

in the conict or it usually does not occur at all. Civilwars that do not end in early victory simply dragon, disrupting the nations economy, destroying itsinfrastructure, and bleeding its population. Protractedcivil wars may wax or wane in intensity, but theyrarely burn themselves out. Contrary to Luttwaksrecommendation, if the international community doeschoose to stand aside and give war a chance, whatwill result is not a more durable peace but a protractedbloodletting that is not likely to end on its own and,even if it does, will leave the nation so decimated thatit immediately becomes a prime candidate for a relapseinto renewed civil war. Once protracted conicts havesettled into a mutually hurting stalemate, they areripe for resolution (in Zartmans words). However,

as I will discuss later, breaking the stalemate usuallyrequires the involvement of a third party to serve asmediator. Left to their own devices, protagonists in acivil war are rarely able to get to a settlement on theirown, for reasons that Barbara Walter has spelled outand which I will discuss later.95

The duration of the conict affects the outcomein several ways. First, the progression of a civil waris an information revealing process, in the sense thatthe experience of ongoing conict forces both thegovernment and the rebels to revise their estimates oftheir chances of victory and the costs they will have toabsorb to achieve victory. The longer the conict lasts,the more likely both sides are to discount their estimateof their chances of achieving victory (P

v). Likewise, the

experience of a protracted conict compels them toadjust their estimate of the amount of time required toachieve victory (t

v) and, therefore, the cumulative costs


45/117

36

required to achieve victory ( Cti).For conicts ending in government victory, there

is evidence that the size of the governments army asa proportion of the nations population does increasethe odds of government victory and shortens thetime to government victory.96 DeRouen and Sobekalso found that increases in the relative size of thegovernments army shorten the war; however, theydid not nd that it affected which side won, only thatthe conict ended sooner.97 These ndings on the effectof the governments military follow from the cost ofconict factor in the decision calculus presentedearlier: where governments have a relatively largearmy, they can inict heavy costs on the rebels earlyand thereby prevail. Rebels start out with a decided

military disadvantage: they have to build a militaryforce from scratch in the shadow of a government thatalready has an established military capability. Thus,we would expect rebels to be especially vulnerableto early defeat. Given this initial disadvantage, ifrebels overestimate their chances of victory whenthey initiate the conict, they are subject to an earlyand decisive defeat. The example of Ch Guevara inBolivia illustrates this vulnerability. Guevara foundlittle interest among Bolivian peasants in his call forarmed uprising against landlords, the Bolivian stateor their foreign benefactors, in large part because landwas relatively abundant in Bolivia.98 Unable to recruita guerrilla army or build a civilian support base of anysize, Guevara was soon tracked down and his small

armed band annihilated by the Bolivian governmentbefore it could build a base of support sufcient toavoid early defeat.

ti=0

tv


46/117

37

How, then, do we explain early victories by reb-els? Previous research has suggested that the type of

regime that is most susceptible to civil warand thetype most likely to be overthrown by an armed rebel-lionis the neopatrimonial dictatorship described ear-lier. This regime type is marked by the dominance ofa single personalist dictator presiding over a govern-ment and a military staffed on the basis of their loyaltyto the dictator rather than their competence, training,or battleeld capabilities. Such regimes tend to be cor-rupt to the point of being parasitic and administra-tively incompetent. When challenging neopatrimonialregimes, rebels often prevail early, despite their initialdisadvantage, because the government is so corrupt,incompetent, repressive, and parasitic that large seg-ments of the population are willing to abandon the re-gime at the rst sign that the rebels can win. Moreover,

the states own military is often deprofessionalized bythe ethos of patronage and corruption that character-izes recruitment and promotion. Not only are they notvery competent on the battleeld, their loyalty to thestate is contingent upon the continued patronage ofthe dictator. When faced with a battleeld challenge,these militaries often collapse, with units choosing todesert or defect to what they see as a rebel bandwagonrather than risk their lives to defend a leader whoseloyalty to them is suspect at best. Thus, when facedwith a rebel challenge, a neopatrimonial regime oftenimplodes as a result of its own corruption rather thanas a function of the rebels military capacity or tacticalbrilliance. Laurent Kabila had led a rebel movementin Zaire for 30 years (indeed, Ch Guevara went to

work with him in 1965 only to depart later that year,frustrated by Kabilas unwillingness to prosecute theinsurgency more aggressively). When Kabilas forcesnally overthrew the Mobutu regime in 1997, their


47/117

38

success was clearly more a function of the implosion ofMobutus regime than of any change in Kabilas strat-

egy, tactics, or level of popular support. The collapse ofthe Somoza regime in Nicaragua and the Lon Nol re-gime in Cambodia present additional examples of thiseffect. Not surprisingly, all of these regimes collapsedsoon after external sponsors withdrew their support.

Another effect that is somewhat surprising is that themore deadly the conict is (measured in casualties asa proportion of the population), the longer the conictlasts. The decision calculus presented earlier impliesthat the deadliness of the conict should shorten itsduration as one or both sides calculate that the higherthe rate at which they absorb costs, the shorter thetime until the accumulated costs of conict begin toapproach the expected payoffs of victory. However,Brandt et al. (2005), found that higher casualty rates are

associated with longer wars.99

They interpret this as asunk cost effect: the more deadly the conict is, themore likely both sides are to continue ghting, perhapsin hopes of avenging or justifying the losses they havesuffered up to that point.

Military Intervention Prolongs Civil Wars.

Contrary to the notion that major powers canimpose a peace by intervening militarily in civil wars,the consistent nding across empirical studies is thatmilitary intervention by outside powers in supportof one side or the otherusually prolongs the conict.100While counterintuitive at rst glancewhy would anoutside power commit troops and treasure to a foreign

military venture if it did not believe that action wouldenable its favored side to score a decisive victory?when one considers the question of Whats in it forthe intervener? this effect makes more sense.


48/117

39

Intervention in another nations civil war involvesa substantial risk to the intervener, often with the

promise of little direct payoff to the interveningnation.101 On the downside, intervention does imposedirect costs on the intervener, in terms of troops andtreasure expended in prosecuting the intervention.Moreover, interventions also carry opportunity costsfor the intervening nation. Military forces committedto the intervention are military forces not availablefor other national security needs. Funds expended onnancing the intervention are funds not available forother national priorities. Finally, interventions carrypolitical risks for the decisionmakers who initiate them.Audience costs to leaders can take a number of forms,from the risks that elected leaders will face at the pollsto the risk that authoritarian leaders face in the formof opposition from within their own authoritarian

coalition.102

Given the costs and risks, nations are more likelyto intervene when the potential costs to that nation(including the political costs to the nations leader) ofnot intervening come to approach or exceed those ofintervening. Under what circumstances would thiscondition arise? When that nations favored side in thecivil war (whether the government or the rebels) is onthe verge of defeat, it then becomes more feasible forthe external power to intervene in order to prevent thatdefeat. If, for instance, an external power depends onanother nation for some vital natural resource such asoil, and the government of that nation is in imminentdanger of being overthrown by a rebel movement,then the risks of intervening can quickly be more

than offset by the now-near certain costs that willfollow from the overthrow of the incumbent regime,saddling the external power with the much greater


49/117

40

(and more certain) costs that accrue from loss of accessto that vital natural resource. The Cuban intervention

in Angola took place not for the purpose of enablingthe government of Angola to deal a decisive blow toUNITA rebels and end that civil war but to preventthe governments overthrow by those rebels. Similarly,the Soviet intervention in Afghanistan and the U.S.intervention in Vietnam were motivated by the desireto prevent the imminent overthrow of a favoredgovernment. Interventions of this type prolong the warby preventing the imminent defeat of the intervenersfavored side in the conict. Rarely do external powersintervene when their favored side is on the verge ofvictory. Why would a leader assume the risks and thecosts of intervention when his/her preferred outcomeis already imminent?

Direct military intervention in the form of sending

armed forces into the middle of another nations civilwar is, of course, rare. More common are indirectforms of intervention, such as supplying one side orthe other with funds and military equipment. Suchmeasures also tend to prolong civil wars in that theyrepresent a subsidy to that sides capacity to sustaincombat operations. External support is a criticaldeterminant of the duration of civil wars because theprotagonists in a civil war, unlike their counterpartsin an interstate war, draw on the same population andthe same economy to sustain their operations. In theabsence of external subsidies (in the form of foreignmilitary and economic assistance to one or both sides),civil wars might come to an earlier conclusion simplyas a function of the protagonists exhausting the human

and material resources available to sustain armedconict. In terms of the decision calculus presentedearlier, subsidies to a civil war protagonist increase the


50/117

41

amount of cost that actor can absorb in the quest forvictory and extend the amount of time that actor can

sustain combat in the quest for victory. The evidencesuggests, however, that these subsidies serve to wardoff defeat rather than enhance the prospects of victoryor shorten the time to victory.

The importance of these subsidies can be seen in howquickly a number of civil wars came to an end after theCold War waned and the two superpowers no longerhad any compelling reason to continue subsidizingtheir favored side in these conicts. The civil war in ElSalvador and the Contra War in Nicaragua both endedin negotiated settlements, in part because the UnitedStates and the Soviet Union no longer had compelling(and competing) strategic interests in subsidizing theirfavored sides in these conicts. Likewise, civil warsin Mozambique, Namibia, and Angola all came to an

end soon after external support for one or both sidesended. Moreover, once Cold War rivalries disappearedfrom UN Security Council (UNSC) deliberations,the ability of that body to achieve the consensusnecessary to authorize UN mediation of these conictswas enhanced considerably, with the result being aremarkable increase in the frequency (and the success)of UN mediation of ongoing civil wars. Many of thesemediation efforts would not have been possible duringthe Cold War because either the Soviet Union or theUnited States (or both) had (competing) interests at stakein these civil wars and, therefore, would have vetoedany UNSC resolution that would have jeopardized theability of their favored side to prevail in the conict.


Once a civil war ends, that nation is confrontedwith the reality that it is at grave risk of experiencinga relapse into renewed conict. As noted earlier,


51/117

42

nations that experience one civil war are highlylikely to experience another. Indeed, a nation that has

experienced one civil war is more likely to experienceanother one than a nation that has never had a civilwar is to experience its rst, even among those that arein the risk set of nations especially susceptible to civilwar onset. To use a medical analogy, a nation that hashad one civil war is like a person who has had a heartattack. That person is more likely to have another heartattack than are others who share the same risk factorsbut have so far not had their rst heart attack.

What do we know about the factors that predict therelapse into renewed civil war? More precisely, whatfactors predict the duration of the peace after a civilwar and, conversely, what factors predict peace failure?Two general conditions affect the durability of the peaceafter a civil war. First, for the peace to fail, a new rebel

group (or a reconstituted old one) must develop theorganizational capacity to mount an armed challengeto the post-civil war regime. The emergence of sucha challenger represents what Charles Tilly has termeda condition of dual sovereignty, dened as a conditionmarked by the appearance of contenders or coalitionsof contenders, advancing exclusive alternative claimsto the control over the government . . .; commitmentto those claims by a signicant segment of the subjectpopulation . . .; the incapacity or unwillingness of thegovernment or its agents to suppress the challengercoalition . . .103

For Tilly, dual sovereignty makes civil war possible.Therefore, the extent to which the condition of dualsovereignty persists or reemerges in the post-conict

environment affects the likelihood that the peace willfail with a relapse into civil war. Thus, factors thataffect the extent to which dual sovereignty persists


52/117

43

or reemerges in the post-conict environment shouldaffect the durability of the peace after civil war.

While dual sovereignty makes civil war possible,whether or not renewed civil war does erupt (and,if so, when) is a function of whether or not dissidentgroups have the incentive to revolt rather than sustainthe peace. This element of agency can be modeled asa function of the potential rebels estimate of the costsand benets of resuming conict versus sustaining thepeace. This decision calculus is similar to that speciedearlier in Equation 1.

Presumably, dissidents would prefer a resumptionof conict only if they believe they can eventually winor at least extract more favorable settlement terms inthe future by resuming the ght now. The decisioncalculus presented earlier (Equation 1) can also be usedto represent an actors expected payoffs from resuming

conict versus sustaining the peace.104

The payoff fromresuming conict is depicted as follows:

EUC=P

V(U

V)+(1-P

V)(U

D)- C

ti(3)

where EUCis the expected utility of resuming the conict,

Uv is the actors estimate of the payoff from eventualvictory, Pv

is the actors estimate of the probability ofachieving victory, U

dis the actors estimate of the cost

from defeat, (1-Pv) is the probability of defeat, C

tiis the

actors estimate of the rate at which the costs of conictaccrue from the present (t

0) to that time in the future

when the actor expects to achieve victory (tv). For a

resumption of civil war to be preferred, the expected

utility of resuming the war, EUc, must be greater thanthe expected utility of sustaining the peace, EU

p. The

payoffs from sustaining the peace are:

ti=0

tv


53/117

44

EUP=UP+ Cti (4)

where EUp

is the expected utility from sustaining thepeace and U

sis the payoff from the post-civil war sta-

tus quo. The payoff from the status quo is augmentedby avoiding the costs that would have to be absorbed

in order to achieve victory ( Cti

).This model suggests that any attribute of the post-

conict environment that (a) decreases the actorsestimate of the probability of victory (P

v), (b) decreases

the actors estimate of the payoffs from victory (Uv), (c)

increases their estimate of the rate at which the costs ofconict would have to be absorbed to achieve victory(C

ti), (d) increases the protagonists estimate of the time

required to achieve victory (tv), or (e) increases their

estimate of the payoffs from sustaining the peace (Up)

should increase the duration of the peace following acivil war by reducing the incentives for that actor toinitiate a new rebellion. One critical difference betweenthe initial onset and the recurrence of civil war is that

the experience of the previous war enables potentialprotagonists in the post-civil war environment toestimate more realistically the likely duration, costs,and probability of victory of a new war, informationthey did not have prior to the onset of the originalwar.

In summary, we expect the peace following acivil war to be less durable if (1) the condition of dualsovereignty persists in the post-war environment, and(2) for at least one politically mobilized group, theexpected utility of resuming armed conict is greaterthan the expected utility of sustaining the peace.

ti=0

tv

ti=0

tv


54/117

45

How a Civil War Ends Affects Whether Another

Will Occur.

If the persi

Documents

Sustaining the Peace After Civil War Pub819