blattman and miguel 2010_civil war

7/28/2019 Blattman and Miguel 2010_civil War

http://slidepdf.com/reader/full/blattman-and-miguel-2010civil-war 1/56

Journal of Economic Literature 2010, 48:1, 3– 57 http:www.aeaweb.org/articles.php?doi=10.1257/jel.48.1.3

3

1. Civil War and the Study of Economics

Internal civil conflict has been common-place during the past half-century, a

fact that, unti l recently, escaped the noticeof most economists. Civil wars, or those

internal conflicts that count more than1,000 battle deaths in a single year, haveafflicted a third of all nations. Countingcivil conflicts, or those that count at leasttwenty-five battle deaths per annum,increases the incidence to more than

Civil WarC B E M*

Most nations have experienced an internal armed conict since 1960. Yet whilecivil war is central to many nations’ development, it has stood at the periphery of economics research and teaching. The past decade has witnessed a long overdueexplosion of research into war’s causes and consequences. We summarize progress,

identify weaknesses, and chart a path forward. Why war? Existing theory is provocative but incomplete, omitting advances in behavioral economics and makinglittle progress in key areas, like why armed groups form and cohere, or how more

than two armed sides compete. Empirical work nds that low per capita incomesand slow economic growth are both robustly linked to civil war. Yet there is lit-

tle consensus on the most effective policies to aver t conicts or promote postwar recovery. Cross-country analysis of war will benet from more attention to causal

identication and stronger links to theory. We argue that micro-level analysis andcase studies are also crucial to decipher war’s causes, conduct, and consequences.We bring a growth theoretic approach to the study of conict consequences to high-light areas for research, most of all the study of war’s impact on institutions. We

conclude with a plea for new and better data. ( JEL D72, D74, O17)

* Blattman: Yale University. Miguel: University of California, Berkeley and NBER. We thank Ana Arjona,Karen Ballentine, Bob Bates, Tim Besley, David Card,Ernesto Dal Bó, Jesse Driscoll, Bill Easterly, Jim Fearon,Karen Ferree, Mary Kay Gugerty, Anke Hoefer, Patri-cia Justino, Stathis Kalyvas, David Leonard, Jason Lyall,Andrew Mack, Daniel Maliniak, Gerard Padro-i-Miquel,Torsten Persson, Dan Posner, Robert Powell, Vijaya Ram-achandran, Debraj Ray, Marta Reynal-Querol, GérardRoland, Shanker Satyanath, Jacob Shapiro, Ryan Sheely,

Stergios Skaperdas, Abbey Steele, Julia Strauss, Dennis deTray, Philip Verwimp, Barbara Walter, Jeremy Weinstein,our anonymous referees and the editor, Roger Gordon, forcomments and discussion. We are deeply grateful to ourcoauthors on related research: Jeannie Annan, SamuelBazzi, Bernd Beber, John Bellows, Khristopher Carlson,John Dykema, Rachel Glennerster, Dyan Mazurana,Gerard Roland, Sebastian Saiegh, Shanker Satyanath, andErnest Sergenti. Camille Pannu, Abbey Steele, and Mela-nie Wasserman provided superb research assistance.



Journal of Economic Literature, Vol. XLVIII (March 2010)4

half.1 This internal warfare is not just extremely common, it is also persistent. Figure 1 displaysthe cumulative proportion of all nations expe-riencing wars and conicts since 1960. Twenty percent of nations have experienced at leastten years of civil war during the period.

The proportion of countries embroiledin civil conict at a single point in timeincreased steadily through the last half of thetwentieth century, peaking in the 1990s atover 20 percent (see gure 2). In sub-Saha-ran Africa, the world’s poorest region, nearly a third of countries had active civil wars orconicts during the mid-1990s.2 The preva-lence of war prompted scholars to ask a sim-

ple question: why there is so much civil warin the world? A decade later, observers beganto ask where some of the civil wars had gone;there were “just” thirty-two active conictsin 2006, the result of a steady decline in con-ict from a peak of fty-one in 1992.

The outbreak of internal wars is commonly attributed to poverty. Indeed, the correlationbetween low per capita incomes and higherpropensities for internal war is one of themost robust empirical relationships in the lit-erature. Figure 3 illustrates the relationshipbetween per capita income (percentiles) andcivil war using a nonparametric Fan regres-sion; countries towards the bottom of the world income distribution—many in Africa—have several times more wars than those in

1 These denitions come from the well-known UCDP/ PRIO dataset developed by Nils Petter Gleditsch etal. (2002) and extended in Lotta Harbom and Peter

Wallensteen (2007). UCDP/PRIO denes conict as “acontested incompatibility that concerns government and/ or territory where the use of armed force between twoparties, of which at least one is the government of a state,results in at least 25 battle-related deaths.” As notedbelow, the denition and coding of civil war is contested,but our main points are robust to alternative approaches.

2 The proportion of individuals directly affected by war violence is lower than suggested by this gure, asmany armed conicts are conned to subregions. Yeteven individuals living in largely peaceful regions can beadversely affected by insecurity, the public policy changes,and economic consequences of a civil war.

the top quartile, while middle income coun-tries still face considerable conict risk.

Yet claims of a direct causal line from pov-erty to conict should be greeted with caution.One reason is that this line can be drawn inreverse. Conicts devastate life, health, andliving standards. A chilling example is theDemocratic Republic of Congo, where surveyssuggest millions may have died as a result of the recent civil war, primarily due to hungerand disease (Benjamin Coghlan et al. 2007).Although the accuracy of mortality gures insuch war zones is open to question, estimatedmortality gures for Rwanda, Angola, andSudan are likewise shocking. Massive loss of life

inevitably affects the economy. Warfare alsodestroys physical infrastructure and humancapital, as well as possibly altering some socialand political institutions. Moreover, internal wars are contagious; refugee ows, disease,lawlessness, and the illicit trades in drugs,arms, and minerals have generated “spillover”effects into the countries neighboring conictzones. Some have argued that the destructiveconsequences of internal warfare may be sogreat as to be a factor in the growing incomegap between the world’s richest and poorestnations (Paul Collier et al. 2003).

A seeming paradox, however, is that warfare is also sometimes credited for thetechnological and institutional developmentthat underpins Western economic prosperity.Both internal and external wars are com-monplace in European history. Severalscholars have claimed that inter-state warsand wars of territorial conquest served a

critical role in enabling the development of strong and capable government institutionsin Europe (e.g., Daron Acemoglu and JamesA. Robinson 2006; Niall Ferguson 2002;Charles Tilly 1975; Tilly 1992). The evidenceon institution-building and internal warfareis limited, but cases of stronger states emerg-ing out of contemporary civil wars also existin East Africa and Southeast Asia (Dan Slater2005; Jeremy M. Weinstein 2005a).



5Blattman and Miguel: Civil War

3 The survey included thirty-eight undergraduate andtwenty-ve graduate syllabi. We searched for online syllabifor undergraduate institutions ranked in the top fty of theU.S. News and World Report college rankings (2007a), and

for Ph.D. economics programs ranked in the top twenty-ve of either the National Research Council (1995), PiperFogg (2007), or U.S. News and World Report economicsPhD (2007b) ranking. Details are available upon request.

It seems clear to us that civil war ought to becentral in the study of international economicdevelopment. Yet leading development econo-mists have too often overlooked it; for instance,two respected and widely taught undergradu-ate development economics textbooks (DebrajRay 1998; Michael Todaro 1999) do not con-tain the words “war,” “conict,” or “violence”

in their subject index. Moreover, a 2007survey (by the authors) of sixty-three devel-opment economics course syllabi in leadingU.S. universities reveals that only 13 percentof undergraduate courses and 24 percent of

graduate courses mention any of these topicsat all.3 Over the past decade, however, many economists and other social scientists have worked to better understand the causes andthe economic legacies of internal warfare,often in collaboration with political scientistsand other scholars. This article’s main goal is tosummarize this progress and help chart a pro-

ductive path forward. As bets an emergingeld, this article focuses as much on what wecannot say today as what we know.

Unfortunately, our survey deals too briey with important topics such as civil war endings

0.6

0.4

0.2

0

–

–

–

–

0 10 20 30 40

Number of years war and war or conict

P r o p o r t i o n

o f c o u n t r i e s

– – – – –

Proportion of countries with civil war

Proportion of countries with civil war or conict

Figure 1: The Distribution of Civil War or Conict Years across Countries, 1960–2006

Sources: Data based on UCDP/PRIO armed conict database. Civil wars are those internal conicts that countmore than 1,000 battle deaths in a single year. Civil war or conict includes cases with at least twenty-ve battledeaths in a single year.




and duration, postwar reconstruction, andthe emergence of peaceable institutions, inpart because these subliteratures are stilllargely in ux. We must also neglect relatedforms of violence—interstate war, terror-ism, coups, communal violence, politicalrepression, and crime—to keep this articlea reasonable length. This is a pity because

the distinction between civil wars and otherforms of political instability has largely been assumed rather than demonstrated.4

Our principal conclusions challengeresearchers to focus on new questions,econometric methods, and data. First,beginning with the origins of conict, weargue that existing theory is incomplete.

0.25

0.2

0.15

0.1

0.05

–

–

–

–

–

1960 1970 1980 1990 2000

Year

P r o p o r t i o n

o f c o u n t r i e s

– – – – –

Proportion of countries with civil war

Proportion of countries with civil war or conict

Figure 2: Proportion of Countries with an Active Civil War or Civil Conict, 1960–2006

Sources: Data based on UCDP/PRIO armed conict database (Gleditsch et al. 2002). Civil wars are thoseinternal conicts that count more than 1,000 battle deaths in a single year. Civil war or conict includes cases with at least twenty-ve battle deaths in a single year.

4 Related literatures investigate the logic andorganization of terrorism, including: self-selectionand screening of terrorist recruits (Ethan Bueno deMesquita 2005); why radical religious clubs special-ize in suicide attacks (Eli Berman and David D. Laitin2008); how terrorist organizations use bureaucracy toalign the asymmetric preferences for violence amongleaders and operatives (Jacob N. Shapiro 2008); theeconomic logic of hostage-taking and governmentresponse (Todd Sandler and Walter Enders 2004); thesplintering and ideology of terrorist groups (Bueno de

Mesquita 2008); the logic of suicide missions (DiegoGambetta 2005); and why terrorists employ roadsidebombs (Matthew A. Hanson 2007). The line betweenrebel and terrorist groups is blurry, and many of thelessons we draw may apply to terrorism. Further theo-retical work laying out the analytical distinctiveness of civil wars versus terrorism and other forms of politi-cal violence would be useful. Anjali Thomas Bohlkenand Ernest Sergenti (2008) document the close linkbetween local economic conditions and the outbreakof communal (inter-religious) riots across Indian states.




Central theoretical problems remainunresolved, including the sources of armedgroup cohesion amid pervasive collectiveaction problems. Moreover, we have yetto develop persuasive arguments for non-traditional mechanisms—myopic or selshleaders, for example, or the role of ideology

and identity in reducing free-riding withinarmed groups. As a consequence, too littleempirical work is motivated by (and explicitly derived from) formal models.

Second, the leading existing theoriesremain untested. Simple contest mod-els—ones that link conict to geographicconditions that favor insurgency, or ones where poverty triggers political violence—have been tested often. Yet one of the most

dominant rational explanations for civil war,conict as the result of commitment prob-lems that prevent socially desirable agree-ments between ghting sides, has barely been examined.

Third, theories seldom specify the empiri-cal predictions that can test between com-

peting accounts. What, for instance, is thealternative to purely rational theories of war-fare? “Irrational” warfare? In fact, there areseveral plausible alternatives: rational actors who do not internalize the social costs of war;maximizing actors with systematic defectsin decision making or expectations forma-tion; strategic interactions between multipleactors within coalitions; idiosyncratic war(described below); and so forth. Theory will

0.4

0.3

0.2

0.1

0

–

–

–

–

–

0 0.2 0.4 0.6 0.8 1

GDP per capita

I n c i d e n c e o f c i v i l w a r

– – – – – –

95% upper band

Fan regression

95% lower band

Figure 3: Incidence of Civil War by Country Income per Capita, 1960–2006

Sources: Figure displays the results of a Fan regression of the incidence of civil war on GDP per capitapercentiles (bandwidth = 0.3, bootstrapped standard errors). Population and GDP data are drawn fromthe World Development Indicators (World Bank 2008). Civil war incidence is drawn from the UCDP/PRIO

armed conict database (Gleditsch et al. 2002; Harbom and Wallensteen 2007).




lead to better empirical testing if and whenit better species the empirical predictionsthat distinguish between models.

Fourth, further cross-country regres-sions will only be useful if they distinguishbetween competing explanations using morecredible econometric methods for establish-ing causality. Up to now this literature hasbeen enormously provocative but has facedequally important limitations: convincingcausal identication of key relationships israre; robustness to alternative specicationsor assumptions is seldom explored; country- years are often assumed to be independentunits in time and space; measurement error

is rarely addressed; an absence of evidenceabout particular effects has often been inter-preted as evidence of absence; and theoriesof individual or armed group behavior aretested at the country level despite obviousaggregation difculties. It would be easy toconclude that the cross-country literature hasbeen exhausted, but that would go too far. We highlight new macro-level research thataddresses some of these challenges head on.

Fifth, we believe the most promisingavenue for new empirical research is on thesubnational scale, analyzing conict causes,conduct, and consequences at the level of armed groups, communities, and individu-als. We refer mainly to the blossoming num-ber of microeconomic statistical studies of armed conict and combatants, as well asto the integration of quantitative evidence with case and historical analysis. The empiri-cal microeconomic work sometimes employs

more credible research designs, yet so farthe results are scattered and many ndingsmay be context-dependent. More studies areexactly what is needed. In our view, the mostinteresting directions for research includethe internal organization of armed groups,rebel governance of civilians, the strategicuse of violence, counterinsurgency strategy,and the roots of individual participation in violent collective action. Each is ripe for the

concerted application of contract theory andmechanism design and insights from behav-ioral economics and industrial organization.

Sixth, we argue that researchers ought totake a more systematic approach to under-standing war’s economic consequences. Anepisode of civil conict, not its absence, isthe norm in most countries, and that warmay be a nation’s most important histori-cal event. Yet what those effects imply forlong-run economic development is unclear.This article also attempts to bring a unify-ing growth theory framework to the study of war’s economic legacies. The bulk of existingevidence focuses on war’s impacts on factors

of production—population and capital—andnds that rapid recovery along these dimen-sions is possible. War’s impacts on humancapital (including education, nutrition, andhealth), however, are often more persistent.Like the “causes” literature, research into“consequences” is beginning to benet frombetter micro-level data and greater use of experimental or quasi-experimental varia-tion. Viewed through the lens of economicgrowth theory, however, there remain moregaps than solid conclusions in our under-standing of postwar recovery. Both theory and evidence are weakest in assessing theimpact of civil war on the fundamental driv-ers of long-run economic performance—institutions, technology, and culture—eventhough these may govern whether a society recovers, stagnates, or plunges back into war.

Finally, in pursuit of all these objectives,much is to be gained from collecting new

data. We conclude our review with recentexamples and priorities for data development.

We share a title with a useful recent survey by Collier and Anke Hoefer (2007) but havedifferent goals. Collier and Hoefer focusin-depth on a set of core macroeconomicquestions. Our piece brings in a broaderrange of research questions and approaches,including an overview of the large conictliterature in political science, and a critical




but hopeful view of the new applied micro-economic work in conict, probably the sin-gle most promising research frontier in our view. We also focus on the theoretical andeconometric limitations of existing work. Webelieve a number of the arguments in thisarticle are novel or have never before beenassembled in a single place, including thediscussion of specic directions for futureresearch.

The rest of the paper is organized as fol-lows. On civil war causes, section 2 surveystheoretical advances and section 3 covers thelarge empirical literature. Section 4 tacklesthe growing literature on civil war’s economic

consequences. The nal section summarizeskey lessons and policy implications, and sug-gests strategies to sustain intellectual prog-ress in this emerging eld.

2. Theories of Armed Conict

Newspaper reports, historical accounts,and econometric work overow withexplanations for conict: ancient hatredsincite violence; oil wealth breeds separat-ism; trade shocks trigger insurrections;income inequality leads to class warfare.Surveying the vast literature on civil war,one feels caught in a complex web of rootand proximate causes (not to mention endo-geneity). In this context, the principal con-tribution of formal economic theory hasbeen to clarify and systematize this tangleof material explanations. Models from

both economics and political science havereduced varied accounts of civil war onsetto a few common logics, each of which canbe approximated in a parsimonious framework of self-interested, wealth-maximizinggroups or individuals. We rst review theseminal theories of civil war, then otherinuential branches of the theoretical lit-erature, and wrap up this section with our views on promising directions.

2.1 Insurrection as Competition for Resources

Models of armed conict depart from theassumptions of standard economic theory in at least three ways: property rights areneither well-dened nor automatically pro-tected, contracts between parties cannotbe enforced, and rulers can be replaced by means other than the ballot box. In this law-less setting, predation and defense are alter-natives to directly productive activities.

The contest model, the workhorse of theformal conict literature, originated withTrygve Haavelmo (1954), and was popu-

larized by Jack Hirshleifer (1988; 1989),Michelle R. Garnkel (1990), and StergiosSkaperdas (1992). It considers two competingparties, a rebel group and a government, andanalyzes each side’s allocation of resources toproduction versus appropriation; Garnkeland Skaperdas (2007) summarize the permu-tations and mechanics of two-party contestmodels embedded in a general equilibriumframework. While production is modeledin the standard manner, appropriation ismodeled using a “contest success function” where inputs (e.g., guns, G) translate intoa probability of ghting side 1 winning, p1,and consuming the opponent (side 2’s) eco-nomic production in addition to their own.Following Hirshleifer (1989), the most com-monly used formulation in theoretical appli-cations is presented in equation 1, where G1 refers to side 1’s weapons, G2 refers to 2’s weapons, and m captures the effectiveness of

weaponry in determining the victor:

(1) p1 (G1, G2) = G1

m ________ G1

m + G2 m

.

Contest models boast at least one robustprediction: the odds of winning increase with the relative effectiveness of that side’s




ghting technology. Technology is denedbroadly in this literature, including any fac-tor that inuences effectiveness, from skillfulrevolutionary leaders, to access to rearmsand training, rugged terrain, or bases on for-eign soil. As we will see below, this predic-tion receives broad empirical support in thesuccess of rebel movements, contributing tothe popularity of the contest approach.

Contest models often treat rebels and rul-ers as unitary actors. Hershel I. Grossman(1991) departs slightly, considering the caseof a single ruler and many citizens, eachof whom can either produce or predate.5 Grossman’s move from unitary actors to

representative households (assumed unableto coordinate their activities) does not greatly change the conclusions of the contest model,but it does highlight the importance of theindividual participation problem: armedgroup leaders must motivate citizens to sol-dier for their side. One immediate insight isthat participation in soldiering rises as theopportunity cost of ghting falls.

These models thus predict that poverty lowers individual incentives for maintainingorder, as soldiering increases with the rela-tive returns to ghting versus production.Can this prediction account for the cross-country correlation between poverty andcivil war? In fact, the theoretical connectionbetween income and armed civil conict isnot so clear cut. In contest models the win-ning party consumes the resources of boththe state and the losers. On the one hand,the greater the national wealth (whether

from taxes, assets like natural resources, orexternal transfers), the more there is to ghtover and thus, in standard formulations, thegreater the equilibrium effort devoted toghting rather than producing (e.g., Garnkeland Skaperdas 2007; Grossman 1999). Yet

5Another approach considers a rebel leader who com-petes with the incumbent for citizen support (Grossman1999).

the absence of resources—natural or oth-erwise—makes production less individually attractive than ghting, but also means thereis a smaller pie to ght over. James D. Fearon(2007) notes that these opposing wealtheffects cancel out in some cases: if state reve-nues are drawn entirely from taxes on citizenincomes, then income could have no effecton equilibrium levels of conict. Positive ornegative income effects could result, though,if utility or revenue collection has a nonlinearfunctional form.

Ernesto Dal Bó and Pedro Dal Bó (2004)model these potentially opposing effects ina two-sector model of the economy. In the

capital-intensive sector, an income shockincreases the value of controlling the state without increasing wages and the opportu-nity cost of ghting; the opposite is true of a shock to the labor-intensive sector. Thusin the rst case conict risk increases, whilein the second it falls. Timothy J. Besley andTorsten Persson (2008a; forthcoming) usea related framework to model the impactof import and export commodity priceson government revenues and rents, as wellas on labor incomes. They conclude thatterms of trade volatility in either directionmay stimulate repression and armed con-ict: increasing import prices increase con-ict risk by suppressing the real wage whilehigher export prices lead to greater conictrisk by boosting the size of the governmentrevenue pie tempting armed groups. A nota-ble aspect is the authors’ attempt to linkthese sharp theoretical predictions to cross-

country evidence, and we return to theirempirical ndings below. Both papers alsosuggest that the distribution of income and wealth—whether across individuals or sec-tors—is central in explaining the economicincentives for rebellion.6 Civil war seems

6 See Roland Benabou (2000) and Abhijit V. Banerjeeand Esther Duo (2003) on inequality and economic per-formance across countries.




more likely when state wealth is easily appro-priated or divorced from the citizenry, as with some natural resource wealth and for-eign aid ows. We revisit this issue below.

2.2 Why Fight? Information Asymmetryand Incomplete Contracting

One drawback of the typical contest modelis that insurrection is never fully deterred;arming and ghting always occur in equilib-rium. There is typically no decision to ght:arming and ghting are one and the same.This prediction of ever-present conict isunsatisfying since political competition overpower and resources is ubiquitous while vio-

lent conict is not. Thus we turn to the deter-minants of compromise (and its breakdown).Creating and arming organizations is

costly and wars are destructive and risky.Thus a fundamental question is why warsever occur at all. If the competing groups arerational, both should prefer a bargained solu-tion to destructive conict.

The possibility of bargaining under thethreat of violence is embedded in leadingtheories of political and institutional develop-ment.7 Acemoglu and Robinson (2001, 2006),for instance, develop a model of elites com-peting with the poor for control of the state.Elites accommodate the poor by extendingthe voting franchise in periods when the poorcan credibly threaten to revolt, and there isno violent conict on the equilibrium path.Carles Boix (2003) develops a related model, where conict outbreaks depend on shiftsin the military capacity of a revolutionary

7 Bargaining models of conict proceed from micro-economic theories of bargaining where parties have theoption of resorting to costly conict if bargaining breaksdown (see John Kennan and Robert Wilson 1993 for acomprehensive survey). Union–rm wage negotiation andpretrial legal settlement in wealthy countries have beenthe two most studied cases. Conict models, however, donot assume that contracts will be enforced once signed,further complicating the negotiation. Barry R. Posen(1993) presents a theoretical international relations per-spective on ethnic conict.

challenger. Fearon (1995) famously outlinedthree reasons why bargaining could fail, lead-ing to inter-state war. First, leaders may notalways behave rationally—decisions mightbe based on emotion, or leaders may not fully calculate benets and risks (bounded ratio-nality). Second, leaders may be fully rationalbut not internalize the full cost of conictbecause of political agency problems. Third,leaders might be rational and internalizecosts, but nd war unavoidable nonetheless.

Almost all theoretical work focuses onthis third case. Fearon highlights threemechanisms consistent with “rational war”:(i) asymmetric information, including pri-

vate information about military strength,and the strategic incentive to misrepresentit to potential opponents;8 (ii) commitment

problems, especially the inability of the par-ties to commit to deals in the absence of athird-party enforcer; and (iii) issue indivis-

ibilities, whereby some issues do not admitcompromise. We will follow the literatureand focus attention on the rst two.9

2.2.1 Information Asymmetries

War can occur when one side overesti-mates its ability to win, or underestimatesits opponent’s strength (Powell 2002). Butasymmetric information is generally insuf-cient cause for war. After all, if both partieshave an incentive to make a deal, they shouldalso have incentives to gather informationand communicate their strengths (Fearon1995). For asymmetries to cause war among

8 This argument appears to originate with Dagobert L.Brito and Michael D. Intriligator (1985).

9 Issue indivisibilities are considered a relatively minorexplanation in most cases. However, Ron E. Hassner(2003) argues that the indivisibility of sacred spaces may be one reason for the persistence of conict betweenIsraelis and Palestinians or Hindus and Muslims in India.Robert Powell (2006) argues, however, that indivisibilitiesare merely a special case of the commitment problem; if commitment were possible, both sides would prefer a lot-tery that awards the contested indivisible prize with thesame odds as ghting, thus avoiding war’s destruction.




rational actors, accurate disclosure of infor-mation must also be impaired. An incentive tomisrepresent one’s strength is the most com-monly theorized mechanism, such as whena state exaggerates its strength and engagesin (inefcient) war in order to deter futureopponents from insurrection.10 To take aninterstate war case as an example, SaddamHussein’s exaggeration of Iraq’s stock of weap-ons of mass destruction in 2002 could be seenas an effort to mislead opponents and deterinvasion—an effort that, nevertheless, failed.

Such accounts are plausible but likely offeronly half an explanation. For one, relativemilitary strength should reveal itself quickly

on the battleeld. Information problems thusprovide a particularly poor account of themany prolonged civil conicts (Fearon 2004;Powell 2006).

Most models also assume just two actors.Civil wars are seldom so simple. Joan Estebanand Ray (2001) develop a multiplayer con-test, where each has imperfect informationabout the others’ costs of conict. With fouror more players, Pareto-improving socialdecision making is impossible and conictensues. Thus information asymmetries may be even more hazardous than the basic two-player models would suggest. Ray (2009)identies another rational route to conict when players are many, developing a modelof coalition formation under multiple threatsthat shows that conict may be unavoidableeven in a world with complete contracts.Societies divide along multiple lines—by class, geography, religion, or ethnicity—and

while society can arrange a set of transfers

10 For instance, Barbara F. Walter (2006) shows thatethnic groups are more likely to seek self-determinationif a government has acquiesced to earlier autonomy demands by other groups. A government that takes suchfuture externalities into account might nd it worthwhileto ght a costly war today to prevent secession, thus pos-sibly heading off future conicts. The relationship alsosuggests that war may not always be accurately modeledas a two-player game.

that avoids a conict along any one division,it may be impossible to nd an arrangementthat simultaneously prevents conicts alongall divisions simultaneously.11

Also promising are recent attempts to inte-grate asymmetric information with othertheoretical mechanisms. Sylvain Chassangand Gerard Padro-i-Miquel’s (2008a, 2008b,2009) work incorporates such asymmetry into a contest model employing a globalgames logic.12 Their key insight is that tran-sient economic shocks increase the immedi-ate incentives to ght but not the discountedpresent value of victory. The model thusimplies that in dire economic circumstances

groups predate upon one another since they have less to lose than in periods where thereturns to production are higher. Yet con-ict is also possible in better economictimes as asymmetric information on the trueeconomic conditions, and rst-strike advan-tages on the battleeld, combine to gener-ate mutual fears of preemptive attacks. Theframework is notable for its testable predic-tions: armed conicts should follow negativeeconomic shocks; higher and less volatilenational incomes are associated with lessconict; and expected future income growthreduces the risk of war today.13

11 Motivated by warlord politics in Tajikistan andGeorgia, Jesse Driscoll (2008) models bargaining betweena president and multiple challengers.

12Global games are associated with the work of HansCarlsson and Eric van Damme (1993) and Stephen Morrisand Hyun Song Shin (1998).

13Chassang and Padro-i-Miquel show that there are

always economic conditions severe enough that civil con-ict breaks out. In their notation, the size of the nationaleconomic pie in period t is θt ∈ (0, ∞); c is the fraction of production destroyed in a civil war; P is the odds that aghting side will prevail if conict breaks out; V V is thecontinuation value if that side prevails in the war; V P is thecontinuation value of the most peaceful Subgame PerfectNash Equilibrium; and δ is the time discount rate. They show that peace is only sustainable if θt [1 − 2P(1 − c)] ≥ δ [PV V − V P]. Since the right hand side of the expressionis strictly positive for plausible parameter values, there isalways an economic shock sufciently bad (close to zero)that violates this inequality.




Dal Bó and Powell (2009) also show thatasymmetric information can lead to war inthe context of pervasive commitment prob-lems without relying on global games. Thegovernment has better information on its wealth than challengers. As in Chassang andPadro-i-Miquel, they show that governmentattempts to buy off the opposition (and avoidconict by offering them a share of the pie)fail in periods of lower economic activity,since the challenger fears that the govern-ment is low-balling them. Commitment prob-lems could further restrict the government’sability to secure peace by incorporating thechallenger into a power-sharing government

(since this would provide the challenger witha stronger position for future aggression).Sandeep Baliga and Tomas Sjostrom (2004)develop a related imperfect commitmentmodel, in which private information abouteach ghting side’s propensity to arm canlead to arms races with probability close toone.

2.2.2 Commitment Problems andIncomplete Contracting

The most intriguing theories of civil warfocus on the cases where credible commit-ments to peace or redistribution cannot bemade even with complete information—that is, at least one side faces an incentive torenege once a settlement is reached (Walter1997). Such circumstances include military scenarios with a rst-strike advantage, andinstances where waging war today can pre-

vent one’s opponent from gaining military strength in the future.

Powell (2006) shows formally that eachof these commitment problems is rootedin a single phenomenon: large shifts in thefuture distribution of power. For a leadingexample, consider a temporarily weak gov-ernment that is attempting to “buy off” astrong rebel group with transfers to securepeace. When the state returns to relative

strength—perhaps because of a reboundin economic activity, foreign aid or com-modity revenues—it will be tempted torenege on its earlier bargain, thus limitingthe amount it can credibly promise to therebel group today. If this time-consistentbut more modest transfer is less than whatthe rebels can gain by ghting today, they will wage war now to lock in the highestpossible payoff.

Similarly, a commitment problem arises when one party can permanently alterthe strategic balance of power by waging war now (Garnkel and Skaperdas 2000;Michael McBride and Skaperdas 2007;

Powell 2006). If going to war weakens oreven eliminates a rebel group for all time,the state will gain a peace dividend since itno longer needs to spend on arms to deterfuture conict. Thus the state has reasonto wage bloody but short conicts if peacedeals are not credible.14

The commitment problem directly sug-gests that civil war is more likely to occur when there are limits to conict resolutionand contract enforcement. Since formal legaland state institutions presumably help toenforce commitments intertemporally, soci-eties with weak government institutions andfew checks and balances on executive powershould empirically be those most likely toexperience violent civil conict (e.g., Fearonand Laitin 2003; Eliana La Ferrara andRobert H. Bates 2001; Skaperdas 2008). This

14 This approach suggests that the likelihood of waris affected by each side’s valuation of the future versusthe present. The risk of future retaliation—the “shadow of the future”—should deter sides from conict. Whenfuture returns depend on present success on the battle-eld, however (i.e., rst-strike advantages), the shadow the future can increase the incentives for conict. It canalso lead to war if a peace settlement is costly in terms of expected future defensive arming. Greater intertemporaldiscounting increases the likelihood of war in the rst caseand decreases it in the second and third (McBride andSkaperdas 2007; Skaperdas and Constantinos Syropoulos1996).




relationship may partially explain the wide-spread occurrence of lengthy civil wars insub-Saharan Africa, a region notorious for its weak state capacity and limited legal infra-structure (Bates 2001; Bates 2008; Jeffrey Herbst 2000).

Yet weak institutions and the absence of athird-party enforcer alone are not sufcientcause for civil conict. The theory impliesthat conict is at least twice conditional: rston weak institutions, and second, on futureshifts in relative power across the ghtingsides. Future empirical models must beginto take these issues into account more seri-ously in testing.

In terms of policy, the theory suggeststhat enforcement of contracts by the inter-national community can potentially substi-tute for weak domestic institutions (Walter1997). Interventions might include armedpeacekeepers, the provision of guaranteednancial transfers to rebels by outside inter-national agencies, and the threat of pun-ishment (including trade sanctions, assetfreezes, and bombing) if either side renegeson the peace deal.

External interventions could also havethe opposite effect, however, and prevent anongoing war from reaching a credible peaceagreement. For instance, the recent pros-ecution of Charles Taylor (former warlordand President of Liberia) and indictmentof Joseph Kony (head of Uganda’s Lord’sResistance Army) by international courtscould make postwar power-sharing deals forrebels less credible in the future, and thus

extend current civil wars if the rebels haveno guarantee that putting down arms willshield them from prosecution in The Hague.On the other hand, the possibility that aninternational indictment could be droppedappears to have been one of the primary incentives for Kony to agree to a ceasere andbegin negotiating peace in the rst place. Wediscuss the scattered empirical evidence oninternational interventions below.

2.3 The Microfoundations of GroupConict

Contest models and rationalist theoriesof civil war rely upon groups behaving asunitary actors, strong assumptions consid-ering the well-known problem of collectiveaction (Mancur Olson 1971). To understandthe causes of war, we must also understandhow groups form, cohere, and persuade theirmembers to risk their lives.

2.3.1 Civil War and the ParticipationProblem

Classic solutions to the collective action

problem use “selective incentives” to motivate participation, with material and pecuni-ary incentives the focus of most models (e.g.,Grossman 1999). Such incentives include

wages, opportunities to loot, promises of future reward, or physical protection fromharm. Economic inequality provides a pos-sible motive for conict to the extent thatseizure of the state brings material gains tothe victors (Fearon 2007).

A literature on agrarian revolutions in the1960s and 1970s argues instead that inequal-ity motivates participation in rebellion notfor private gain, but because it generatesfrustration over inequality or the destabiliza-tion of traditional social systems (James C.Davies 1962; Ted Robert Gurr 1971; Jeffery M. Paige 1975; Roger D. Petersen 2001;James C. Scott 1976). By these accounts,poverty, income inequality, and unmet eco-nomic expectations may indeed be the root

causes of conict, but the more proximateexplanations are better described as griev-ances. Rather than deny material motiva-tions, these accounts provide an alternativeset of mechanisms for individual participa-tion in rebellion.

Related nonmaterial incentives are thoughtto be common within armed groups. Severalstudies argue that a leader’s charisma, groupideology, or a citizen’s satisfaction in pursuing




justice (or vengeance) can also help solve theproblem of collective action in rebellion (e.g.,John E. Roemer 1985; Elisabeth Jean Wood2003b). These unconventional incentiveshave typically been the subject of sociology (Amitai Etzioni 1975; Olson 1971). A conve-nient way of modeling such sentiments in arational framework is as “goods” of inherent value that individuals consume by ghting.15

What these micro-level approaches oftenignore, however, is that ghting is not theonly means by which individuals and groupscan pursue political and economic change. Walter (2004) argues that the absence of a nonviolent means for achieving change

is also often necessary to incite rebellion.Nonviolent political alternatives could beincorporated into the decision frameworkfacing citizens, leaders, and armed groups,generating testable predictions about therelationship between political institutionsand the likelihood of civil conict.16

Such diverse selective incentives—pecuni-ary or not—are easily embedded in a princi-pal–agent framework. A leading example isScott Gates (2002), who models how rebelleaders can use material incentives along-side ethnic appeals to motivate citizens to join and exert effort in the rebellion (i.e., tosatisfy the participation and incentive com-patibility constraints). His model emphasizeshow incentives and methods of recruitment vary with ease of supervision; the greater thedistance—whether geographic or social—between the leader and the recruit, the more

15 The approach closely parallels two literatures. Oneis a branch of the voting literature that suggests the col-lective action problem inherent in participation in demo-cratic elections (where the odds of affecting the outcomeare innitesimal but voting has concrete costs) is overcomeby the value some individuals place on the act of votingitself (Amrita Dhillon and Susana Peralta 2002; Timothy J. Feddersen 2004). A literature on revolution emphasizesthe inherent value individuals place on retaliation againstan unjust state (Frantz Fanon 1961).

16 Matthew Ellman and Leonard Wantchekon (2000) isa useful step in this direction.

difcult are supervision and punishment, andthe more likely that material incentives (e.g.,looting) will need to be offered to recruits tosecure their cooperation.

Threats and punishments can also be usedas selective incentives. Coercive recruitmentis especially common in African insurgen-cies where, in the absence of a shared socialbasis for mobilizing rural support, rebel lead-ers resort to the only tool at their disposal(Thandika Mkandawire 2002). Michael Suk-Young Chwe (1990) and Bernd Beber andChristopher Blattman (2008) model the use of coercion and pain in a principal-agent setting,and identify the conditions (and agent types)

where it is optimal for armed group leadersto threaten pain instead of offering rewards.This consumption approach to nonpecuniary incentives is analytically convenient and yieldsuseful insights, such as the rationale for usingcoercion on low productivity recruits (espe-cially children). Yet these models are unlikely to capture the complex individual motivationsunderlying participation in armed groups,however, and thus constitute an importantarea for further research.

2.3.2 The Formation of CompetingCoalitions

The models reviewed assume that rebeland government groups exist and are actively engaged in combat. They do not tackle theissue of how competing groups form and why they cohere. An emerging literature basedon the noncooperative theory of endogenouscoalitions explores the distributional basis

of group formation. These models typically assume that group action is more efcientthan individual action, providing citizens with an incentive to join forces. These mod-els also allow for conict within each groupover the distribution of their joint surplus,conict that can be costly for the individ-ual. Stable groups are those that have low-cost mechanisms for distributing the gains,such as property rights norms. The size of




stable groups depends on the relative effec-tiveness of groups at managing both inter-group and intra-group conict (e.g., FrancisBloch, Sántiago Sánchez-Pagés, and RaphaelSoubeyran 2006; Garnkel 2004). Thisapproach is a promising source of micro-foundations for the commitment problemsdiscussed above, since the institutions thatallow for within-group cooperation may alsomitigate intergroup conict.

Relaxing the unitary actor assumptioncould also expand the range of rationalexplanations for armed conict. Informationproblems within groups could lead to bar-gaining breakdowns (just as was the case for

asymmetric information across groups). Fieldgenerals have incentives to mislead civilianleaders about the capability of their military forces if they hope to keep the ghting goingfor longer than citizens would like (to keepmilitary budgets at high levels, for instance).

Alternatively, the possibility that groupsmight split could exacerbate commitmentproblems: signing a peace deal with a rebelgroup leader is of limited value if hard-linersare able to secede and continue ghting.17 Theexistence of splinter factions may explain thereluctance of ghting sides to enter into peacetalks and cause such talks to fail. StephenJohn Stedman (1997) argues that the greatestrisk for peace negotiations comes from such“spoilers”: “leaders and parties who believethat peace emerging from negotiations threat-ens their power, worldview, and interests, anduse violence to undermine attempts to achieveit” (p. 5). Ray (2007; 2009) and Acemoglu,

Georgy Egorov, and Konstantin Sonin (2009)discuss coalitional stability when such devia-tions and counterdeviations can occur.

17 For example, Stathis N. Kalyvas (2000) argues thatinternal divisions between moderates and radicals withinAlgeria’s Islamist FIS party, and the moderates’ inability to make binding policy commitments to reassure anti-Islamist elements of the national army, contributed to theoutbreak of civil war there after FIS won the 1991 nationalelection.

2.3.3 Ethnic Groups and Conict

Ethnic nationalism is popularly viewed as the leading source of group cohesion and (by extension) intergroup civil conict; of 709minority ethnic groups identied around the world, at least 100 had members engage in anethnically based rebellion against the stateduring 1945 to 1998 (Fearon 2006). But why do ethnic groups themselves form, cohere,and sometimes engage in such violence? Afull review of the literature on the formationof ethnicity and ethnic conict is beyondthe scope of this paper, but an outline of themain ideas merits discussion.18

“Primordialist” arguments stress thedeep cultural, biological or psychologicalnature of ethnic cleavages, whereby con-ict is rooted in intense emotional reac-tions and feelings of mutual threat (DonaldL. Horowitz 1985). Economic models thatassume individuals prefer to mingle withco-ethnics (or share political preferences)might be construed as primordialist innature (Alesina, Reza Baqir, and WilliamEasterly 1999; Alesina and La Ferrara 2000;Esteban and Ray 1999). There are clearparallels to the models of group formationdiscussed above: co-ethnic preferences canaugment intragroup mechanisms of commu-nication and cooperation, while interethnicanimosities may exacerbate information andcommitment problems.19

18 For overviews of ethnic mobilization and violencesee Laitin (2007), Fearon (2006), and Bates (2008). Onethnic divisions and economic performance, see AlbertoAlesina and La Ferrara (2005) and Pranab Bardhan (1997;2004).

19Alternatively, as with the grievances discussed above,ethnic violence might have inherent utility value. In theextreme case, we could even reject the rationalist assump-tion entirely that opposing ethnic groups prefer to reach apeaceful solution. However, we believe the goals of formaleconomic theory here should go beyond simply assumingthat a taste for violence drives civil conict, to uncover thedeeper economic, political and social factors at play.




Even if ethnic identities are not primordialand inter-ethnic animosities are absent, eth-nicity may still facilitate strategic coordinationand enforcement. Ethnic groups often exhibitdense social networks and low cost informa-tion and sanctioning, and may have identi-able characteristics that allow outsiders tobe excluded from public goods (FrancescoCaselli and Wilbur John Coleman 2006;Fearon and Laitin 1996; Edward Migueland Mary Kay Gugerty 2005). Fearon andLaitin (1996) show that better within-groupcohesion can facilitate peace deals betweenethnic groups. Alesina and La Ferrara (2005)also speculate that ethnically homogenous

groups possess a production advantage thataugments their incentives to associate. Bates(1986) argues that shared language and cus-toms facilitate organization. Finally, Estebanand Ray (2008) suggest that ethnic allianceshave a distinct advantage over class alli-ances in mobilizing for conict. While classand ethnic groups both possess shared socialidentities, only ethnic groups exhibit within-group economic inequality: inequality allowsthe rich to supply conict capital (e.g., guns)

while the poor supply conict labor.Finally, “modernist” theories stress that

ethnic conict arises when groups excludedfrom social and political power begin toexperience economic modernization (Bates1986; Ernest Gellner 1983)—a situation thatparallels Powell’s (2006) account of shifts infuture power leading to bargaining break-downs today.

2.4 Challenges and Areas for Further Work

Most real-world disputes are settled, evenamong antagonistic ethnic groups. Thus thetheoretical apparatus described above isplausible: conict is rooted in endemic com-petition for resources across groups, withbargained solutions occasionally breakingdown because of commitment or informationproblems. Persuasive though this frameworkmay be in many circumstances, there remain

many challenges and areas for further theo-retical investigation.

2.4.1 Disentangling Competing Accounts

Existing formal theories of conict yieldfalsiable predictions, but few articulate theprecise empirical tests that would distinguishamong alternative mechanisms. Income volatility is one example. In the theories weconsider above, a negative aggregate incomeshock is associated with an increase in armedconict in various models, including thosethat emphasize the diminished opportunity costs of soldiering (Gates 2002, Chassang andPadro-i-Miquel 2009), weaker state repres-

sive capacity (Fearon and Laitin 2003), or therole of asymmetric information (Chassangand Padro-i-Miquel 2008a). Meanwhile, anegative aggregate income shock is associated with a decrease in conict risk in models thatstress capturing the state and its revenues asa prize (e.g., Garnkel and Skaperdas 2007;Grossman 1999). Finally, income volatil-ity in either direction could inhibit crediblebargaining and commitments if it is associ-ated with rapid shifts in power across groups(Powell 2006), or gives rise to worse informa-tion about current economic conditions (DalBó and Powell 2009). Few theories modelmore than one of these dynamics or identify the empirical predictions that will adjudicateamong competing accounts.

One exception is Dal Bó and Dal Bó (2004), who distinguish between shocks to differenteconomic sectors. This theory is an improve-ment over single-sector models, yet even so,

alternative mechanisms and interpretationsare still possible. For instance, if higher cap-ital-intensive good prices fail to increase con-ict, it might be because greater state capacity (associated with higher government revenue)dominates the state-as-prize effect. If civil war is the result of a bargaining breakdown,there are good theoretical reasons to believethat events such as price shocks have differ-ential effects on civil conict depending on




the local institutional setting, the number of already existing armed groups, and the futureshifts in power across political groups likely to result. Future theoretical work should fol-low the lead of this paper in helping us pin-point the empirical patterns that distinguishbetween alternative mechanisms.

2.4.2 Understanding Grievances

At present, the economic motivations forconict are better theorized than psychologicalor sociological factors. Individual preferencesin existing models typically include only mate-rial rewards and punishments. One implica-tion is that we have not derived the falsiable

predictions that distinguish between materialand non-material theoretical accounts. Thus we cannot discard non-economic explana-tions for conict. Take the role of economicinequality, for example. The unequal distri-bution of resources can generate materialincentives for a relatively poor group to seizecontrol of the state. More than one historicalaccount, however, emphasizes citizens’ emo-tional and ideological outrage over inequality as a prime motivation for engaging in violentcollective action.20 While the reduced-formprediction that greater economic inequality leads to armed conict is unchanged in eithercase, the relationship could be interpreted asevidence of either “greed” (economic motiva-tions) or “grievance.”

Christopher Cramer (2002) critiques theconict literature for its tendency to use suchreduced-form empirical relationships to but-tress economic interpretations, arguing that

the underlying relationships between eco-nomic, social, and psychological factors are farmore complex. He stresses Antonio Gramsci’s(1971) denition of “economism”: presenting

20 Barrington Moore (1993), for instance, has arguedthat Nazi fascism and anticapitalist rhetoric stirredanger in German peasants over the perceived control of resources by a supposedly hostile Jewish elite. A similardynamic was at work during the anti-Tutsi genocide inRwanda in 1994 (Scott Straus 2006).

causes as immediately operative that in factonly operate indirectly, and thus overstatingproximate causation. Understanding thesecomplex relationships is crucially importantfor preventing armed conicts. Innovative ways of modeling and measuring individualpolitical grievances are required to makeprogress on this agenda. Yet in the end, ourmeasures may fail to capture the relevant vari-ation. Grievances are uid and the case litera-ture points to the evolution of identities andnorms during wartime (Wood 2003a).

Recent behavioral and experimental eco-nomic research argues that notions of fairnessand grievance are salient in individual deci-

sion making. There is growing lab evidencethat individuals have a taste for punishingsocial norms violations and are willing toincur nontrivial private costs to do so. This willingness to punish unfair behavior appearsto have neural-physiological underpinnings(Dominique J.-F. de Quervain et al. 2004) andis consistent with preferences for equity (Gary Charness and Matthew Rabin 2002; ErnstFehr and Klaus M. Schmidt 1999). Jung-KyooChoi and Samuel Bowles (2007) argue thataltruistic preferences favoring one’s in-groupmay have conferred an evolutionary advan-tage. Such within-group social preferencescould reduce the local collective action prob-lem inherent in mobilizing armed groups by lowering the cost of sanctioning free riders.21

21 The taste for violence may differ from the taste forpunishing others monetarily, but the experimental eco-nomics literature has not to our knowledge carried out

similar research on individual preferences for inicting violence on others. Given the inherent human subjectsissues, observational data on perpetrators of violence isa more promising avenue here. Sociological and psycho-logical understandings of interpersonal violence contrastssharply with the rational choice approach we emphasize inthis article. Sociologist Randall Collins (2008) argues thatmost interpersonal and combat violence is characterizedby a short and confused belligerent “haze”: actors are emo-tionally overwhelmed with tension and fear, and violenceis perceived as the resolution of this fear. Meanwhile,some public health and psychology evidence suggests thatmuch violence is “shame-induced” (James Gilligan 2000).




2.4.3 Disaggregating Institutions

The commitment problem is a persuasiveexplanation for civil war. Unfortunately, wehave a poor understanding of the specicpolitical and legal institutions capable of enforcing commitments and facilitating com-promise between competing groups. Sometheories emphasize the importance of mar-ket promotion and tax levying (Besley andPersson 2008b), and others property rightsand the rule of law (Garnkel 2004). Stillothers emphasize the role of internationalinstitutions and the threat of external intervention, others the internal legitimacy of

the state (e.g., the rule of a minority ethnicgroup, whether in Tutsi Rwanda or apartheidSouth Africa, could be particularly destabi-lizing). Meanwhile, Powell (2006) empha-sizes institutions that help manage rapidshifts in power, an example of which mightbe the ability of elites to extend or retractthe democratic franchise, as in Acemogluand Robinson (2001, 2006). The “institu-tions” concept needs to be better disaggre-gated and tested to be useful in the civil warliterature.

Barry R. Weingast (1997), Bates (2008)and Bates, Avner Greif and Smita Singh(2002) argue that the incentives and con-straints facing leaders are crucial, and inparticular that rulers loot the state whenthe long-term costs of doing so are low.Institutions shape these costs as well asthe ruler’s time horizon and discount rate.Paradoxically, institutions that extend a

ruler’s horizon, such as the elimination of term limits or the weakening of politicalcompetition, may increase the incentives forsupporting political order, so the ruler canextract rents over a longer time. Conversely,Bates (2008) argues that international pres-sure for African states to democratize inthe 1990s increased disorder, since it short-ened leaders’ horizons at the same time thatforeign aid ows were reduced; with few

institutional checks on their power, Africanrulers had incentives to predate.

These arguments meld with a growingcomparative politics literature on state fail-ure and “warlordism” in the late twentiethcentury. In a study of civil war in Liberia,Sierra Leone, and Guinea, Amos Sawyer(2004, 2005) emphasizes the large spoilsfrom power combined with the absenceof checks and balances on the executive asthe primary cause of war in those nations. William Reno (1999) also examines the inter-nal dynamics of “warlord states” in SierraLeone, Nigeria, the Democratic Republic of Congo, and Liberia. Like Bates, Reno argues

that, in the presence of resource wealth, a weakened state, less foreign aid, and pres-sures for economic liberalization, strongmenfound it optimal to deinstitutionalize thestate and formal bureaucratic mechanisms infavor of a parallel “shadow state” under theirown control.

2.4.4 The Conduct and Organization of Civil War

Another important area of study is theconduct of rebellion, investigating what fac-tors and initial conditions inuence a group’sformation, recruitment strategies, ghtingtactics, and internal organization. One goalis to describe the logic of civil war and vio-lence—a reaction to the view, popularizedby journalism and some international rela-tions scholars, that the brutal violence thatcharacterizes much of modern civil warfare isa product of illogical barbarism unrestrained

by economic, political or social structures(e.g., Mary Kaldor 1999; Robert D. Kaplan1994).

Rebel groups are large, self-sustainingindigenous organizations in societies whereeffective organizations (including privaterms) are rare; understanding the glue thatholds them together should be a top researchpriority. One strand of recent research appliescontract theory to theories of recruitment.




Some of these, already discussed above, focuson how armed groups motivate recruits toght (e.g., Beber and Blattman 2008; Gates2002, 2004). Weinstein (2005b, 2007) devel-ops a theory linking a rebel group’s social andeconomic endowments to its composition andtactics. He argues that groups rich in mate-rial resources are ooded with opportunistic joiners with little commitment to the civilianpopulation, while armed organizations withideological “resources,” like a strong sense of common identity, tend to attract more com-mitted soldiers.22

Armed group cohesion is the subject of alarge body of work in military sociology and

history. Rather than focusing on economicincentives, this literature emphasizes thepowerful role of group socialization and socialidentity in generating solidarity, commitment,and a willingness to risk one’s life (see Paul D.Kenny 2008 for a review). Inuential organi-zational devices include the creation of new identities among recruits and unit solidar-ity (Richard A. Gabriel and Paul L. Savage1979; Edward A. Shils and Morris Janowitz1948) and systems of command and control(Robert Sterling Rush 1999, 2001; Martin Van Creveld 1982). This emphasis on organi-zation-level dynamics in state militaries con-trasts with the emphasis on individual-levelmotives often used to explain participation innonstate groups; we believe both literaturescould gain from an exchange of perspectives.Two such crossovers are Francisco GutiérrezSanín (2008), who examines Colombian para-military and guerrilla groups, and Wood

(2008), who discusses how armed groups have

22 The contrast between the Revolutionary UnitedFront (RUF), funded through diamond mining and smug-gling, and the community-supported Civilian DefenseForces (CDF) in Sierra Leone’s recent civil war provides an illustration of this divergence. L. Alison Smith,Catherine Gambette and Thomas Longley (2004) show that the RUF was much more likely to commit humanrights abuses against civilians than the CDF. See DavidKeen (2005) for a careful discussion of the Sierra Leonecivil war.

constructed identities and recongured socialnetworks in El Salvador, Peru, Sierra Leone,and Sri Lanka.

Fearon (2007) asks why we tend to see thesustained survival of many small and lightly-armed guerrilla groups, each with littlechance of capturing political power (Congo,Sudan and Uganda are countries where thishas been true). He constructs a contest suc-cess function with decreasing returns to scalefor rebels over some size range—in other words, above some size, each additional rebelincreases the probability the rebel group isdetected, denounced, or destroyed by thegovernment, and this effect outweighs the

ghting benets of greater size (at least up tosome point).23 Powell (2007) is perhaps thebest articulated formal attempt to get insidethe black box of armed groups’ ghting strat-egies. He models optimal military spendingacross potential targets (e.g., cities or ghtingunits) by a government fearing rebel attack,and is able to decompose such spending intoa defensive effect, a deterrence effect, and acost effect.

Finally, other models help to explain rebel violence directed at civilians.24 Jean-PaulAzam (2002, 2006) formalizes a strategiclogic whereby an armed group engages inlooting to reduce the returns to non-military labor effort for potential recruits (thus mak-ing them more likely to join the group), whilesimultaneously generating spoils to rewardexisting recruits. The logic of violenceagainst civilian populations is the subject of a growing literature in political science (see

Kalyvas 2006 for a review). In work basedon a comparative study of irregular civil wars (i.e., guerrilla wars) in the past century,Kalyvas (2006) argues that rival sides prefer

23 The sensitivity of results to such functional formassumptions calls out for more research investigating themicro-foundations of contest success functions.

24 See Straus (2007) for a review of the related litera-ture on the perpetrators of violence.




to use selective rather than indiscriminate violence to punish “defectors,” or civilianenemies and informers. In the absence of information, both sides rely on local collabo-rators to denounce defectors. Kalyvas arguesthat selective violence—including violencerelated to private, not political motives—ismost widespread in zones where each sideholds signicant force but lacks full control.25 Finally, recent studies examine the use of sexual violence by armed groups (Dara Kay Cohen 2008; Wood 2006; 2009).

This collection of theories just scratchesthe surface of the recruitment of ght-ers and organization of civil warfare. This

area remains one of the most promisingand understudied areas in the literatureon conict, and is ripe for the applicationof advances in contract theory, corporatenance, behavioral economics and industrialorganization (Jean Tirole 1988, 2006). New evidence to motivate and test these theoriesis discussed in section 3.

2.4.5 Departures from the Rational Model

As we discuss below, existing empiricalmodels of conict have limited explana-tory and predictive power. We can draw atleast three possible conclusions from theirrelatively weak performance. First, thedeterminants of war could be understood within standard rational choice frameworksbut simply difcult to measure. In this caseour prime focus as researchers should be toimprove data and measurement. To someextent this is already happening.

Second, war could have idiosyncraticcauses, attributable to chance, singular cir-cumstances, or unsystematic “irrational’behaviors by leaders, encompassing errorsin decision making, personality defects,

25 In contrast, according to Kalyvas, the most heav-ily contested zones are likely to be relatively peacefulbecause denunciations will be deterred by the likelihoodof immediate retribution.

and so forth (Erik Gartzke 1999). Suchan account is not inconsistent with formaltheory. Models are seldom intended to bedeterministic but rather to describe generaltendencies. Civil war outbreak is a relatively rare event and thus it is conceivable thatthe basic formal logic is right but at leastsome civil wars are in fact costly mistakes.Indeed, the historical literature is replete with leaders’ passions, fallibility, and ide-ology; historians often attribute war andpeace to the attributes of individuals likeHitler or Gandhi. Some possibilities forincorporating these issues into formal mod-els already exist. For instance, uncertainty

over whether an opposition leader is an“irrational” type would affect strategies inmodels of asymmetric information.

Third, wars could have determinants thatare outside the simple rational framework,but systematically so. Some obvious explana-tions are still consistent with rational models,such as a leader’s failure to internalize thefull social costs of war—a possibility raisedby Fearon (2004) and recently modeled by Matthew O. Jackson and Massimo Morelli(2007) in the context of inter-state war. They show it only takes one “biased” leader (in thesense that their returns to war differ fromtheir citizens’) for war to break out betweenopposing states.

A related possibility is that leaders are vul-nerable to systematic errors in decision-mak-ing, such as overestimating their chance of winning (overcondence), time-inconsistentpreferences, or other types of predictable

“irrational” behavior. In such cases, formalmodels of conict may be fertile ground forapplication of advances in theoretical behav-ioral economics. Efforts to incorporate psy-chological factors and misperception intointernational relations theory include RobertJervis (1976) and Jack S. Levy (1997), but,to the best of our knowledge, these insightshave yet to be applied to formal models of civil war.




New empirical evidence suggests thatpolitical leaders often do matter. BenjaminF. Jones and Benjamin A. Olken (2009)compare successful to failed assassinationattempts, and nd that the unexpected assas-sination of leaders tends to ename low-scale conicts and diminish high-intensity conicts. Similarly, the unexpected death of rebel leader Jonas Savimbi is widely viewedas the event that directly ended Angola’s war,so much so that Massimo Guidolin and LaFerrara (2007) use his death in an event study of war’s termination on diamond company stock returns. While we think that economictheory should probably refrain from pinning

too much on personalities, the econometricevidence just cited means that leadershipcannot be entirely ignored.

This example on the role of leaders sug-gests that certain determinants of conictoutside standard models are observable andtestable, and thus could be a basis for new theory. Furthermore, an important possibil-ity seldom discussed in the recent formaltheoretical literature is that complex, unsys-tematic, and difcult to observe forces may greatly inuence the outbreak of war, andmake a single general economic theory of civil war impossible to craft. The need forintellectual humility is taken for granted by many civil war scholars (e.g., Cramer 2007).Even if observable structural factors remainimportant, the existence of other inuences will complicate empirical testing, especially in the statistical analysis of relatively rareevents. We now turn to the evidence, where

the implications of these and other estima-tion challenges are discussed.

3. Evidence on the Causes of Conict

The correlates of war are by now well-established. Civil war is more likely to occurin countries that are poor, are subject to neg-ative income shocks, have weak state insti-tutions, have sparsely populated peripheral

regions, and possess mountainous terrain.Ultimately, empirical work should aim todistinguish which of the competing theoreti-cal mechanisms best explain the incidence,conduct, and nature of civil war, but this goalis still far from being realized.26 We havelimited evidence on the relative inuenceof the commitment problems and informa-tion asymmetries so central to formal theory.In many cases it is still not clear which of the above correlates actually cause war and which are merely symptoms of deeper prob-lems, and we have yet to solve the puzzlesof participation, collective action, and groupcohesion laid out above.

3.1 Cross-Country Evidence

Cross-country regressions dominate theconict literature, and no discussion of civil war empirics is complete without referenceto the seminal contributions of Collier andHoefer (1998; 2004) and Fearon and Laitin(2003). Collier and Hoefer ignited interestamong economists—and heated disagree-ment among scholars in other elds—witha simple argument: political grievancesare universal but the economic incentivesto rebel are not, and these latter factorsare often decisive. Scores of papers havefollowed in their footsteps, most sharing ahandful of traits: a dependent variable thatindicates war onset or incidence, proxy vari-ables representing possible causes of conict,and a regression-based test of these compet-ing determinants.

Collier and Hoefer broadly root their

empirical model in a contest model of con-ict, and nd several variables with robust,positive correlations with conict incidence.First, slow current economic growth is

26 Other recent reviews of the empirical literature,often spanning economics and political science, includeCollier and Hoefer (2007), Macartan Humphreys(2003), Patricia Justino (2007), Kalyvas (2007), NicholasSambanis (2002), and Wood (2003a).




associated with conict, as is the proportionof natural resources in total exports. Higherlevels of secondary school attainment in thepopulation, in contrast, are associated with alower risk of civil war. Meanwhile, a country’sethnic fractionalization, income inequality,and democracy are not statistically signi-cant predictors of conict risk conditionalon these other factors. Collier and Hoeferconclude that economic forces, primarily theability to organize and nance a rebellion(as captured in their economic growth andschooling variables, and the ability to exploitnatural resources) most strongly predict whether civil war occurs.

Fearon and Laitin (2003) take a closely related cross-country approach. Their coreregression, while not derived explicitly fromtheory, became the standard formulation formost cross-country work that followed. Itresembles the logit specication in equation2, where the dependent variable, ONSET it, isan indicator for the onset of civil war in coun-try i in year t; CW it is an indicator for theincidence of civil war (which equals one inonset and all active war years); y

i,t−1is lagged

per capita income; and X it is a vector of K population, geographic, political controls(including democracy measures) and social variables (including ethnic and religiousfractionalization):

(2) ONSET it = Λ(β 0 + β 1 CW i,t−1

+ β 2 y i,t−1 + X ′ it β K + εit ).

Like Collier and Hoefer, Fearon andLaitin nd, rst, that conditions favor-ing insurgency, like rough terrain, increasethe likelihood of civil war, and second, thatproxies for political “grievances” (e.g., eth-nic and cultural diversity) have little predic-tive power. Yet Fearon and Laitin also arguethat proxies for state institutional capacity and strength—most importantly, per capitaincome—are robust predictors of civil war.

They conclude that war is engendered by weak central governments and environmen-tal conditions favoring insurgents.

How do these two papers reach differentconclusions with similar data and economet-ric techniques? Most importantly, the twostudies attach different interpretations tokey variables like per capita income. Collierand Hoefer link it to the opportunity costsfacing potential rebels, while Fearon andLaitin emphasize its correlation with statecapacity. Yet neither of these two “pure”interpretations is entirely justied given theevidence at hand. The link between incomelevels and armed conict is theoretically

complex, and ner-grained data—say, onincomes that revert to the state versus thecitizenry, or actual longitudinal measuresof state capacity—is required to distinguishbetween these interpretations.

Second, the authors differ in how they code civil wars. Sambanis (2004) examinesthe four competing datasets and nds vemain differences: (i) in thresholds of violencerequired to be dened as a civil war; (ii) thedenitions of war beginnings and endings;(iii) in their treatment of ‘internationalized’civil war (where there is some involvementby outside parties); (iv) in their treatment of related forms of conict (e.g. communal vio-lence or state repression); and (v) the under-lying data sources they draw from.27 BothCollier and Hoefer and Fearon and Laitinexamine conict onset, albeit with differentdenitions of war. Other options, however,

27 The coding of civil wars and conicts remains prob-lematic. Increases in conict intensity are generally notcaptured, except by the 25 and 1,000 death thresholdsin the PRIO/Uppsala data. Some civil “conicts” may include low level violence that by other criteria would notbe considered a war (e.g., crackdowns by federal policeon drug gangs). Cramer (2007) reviews other challenges,including: a large-country reporting bias (where it is easierto meet the battle death threshold); the emphasis on battledeaths (omitting, for instance, civilian killings, refugeemovements, and state repression); and difculties of dif-ferentiating conict lulls from true termination of conict.




include conict incidence (including all yearsof war in the analysis) or conict duration.These three approaches can be applied toat least four different civil war datasets, tocreate twelve possible dependent variables.These datasets do not always agree on thecoding of war and correlation coefcientsacross datasets range from 0.96 down to 0.42(with an average of 0.68).28

A third source of inconsistent results liesin the somewhat ad hoc empirical modelstypically used. In this way the cross-country conict literature mirrors the earlier debateassessing the causes of economic growth, where there was also little agreement on the

correct econometric specication (e.g., RossLevine and David Renelt 1992). Authors vary in their use of annual versus ve-year peri-ods, corrections for time dependence, thetreatment of ongoing war years, the appro-priate estimator for rare events, the use of country xed effects, and so forth.29

Fourth, estimates are sensitive to theexplanatory variables employed. Hegre andSambanis (2006) test the sensitivity of esti-mates to changes in the conditioning set,using the approach popularized in XavierSala-i-Martin (1997), and identify a few robust correlates of civil war onset: low percapita income, slow income growth, rough

28 See Sambanis (2004). These correlations of conictonset are likely to exaggerate differences in conict inci-dence (since datasets may disagree on the exact year of conict initiation). This difculty and inconsistency in

pinpointing onset will frustrate attempts to relate conictonset to time-varying explanatory variables more thantime-invariant ones. The four most common datasets thatSambanis explores are: the Correlates of War (COW) (J.David Singer and Melvin Small 1994), Fearon and Laitin’s(2003) dataset, PRIO/Uppsala (Gleditsch et al. 2002), andSambanis’s dataset (2004).

29 On time-dependence and dynamics, see NathanielBeck and Jonathan N. Katz (2004) and Beck, Katz, andRichard Tucker (1998). On logistic regression with rareevents data, see Gary King and Langche Zeng (2001).For other issues, see the discussion in Håvard Hegre andSambanis (2006).

terrain, large population size,30 recent politi-cal instability, small government militaries,and war-prone neighbors. Yet many of these variables are plausibly endogenous, biasingother estimates in unknown directions.

Finally, the country level of analysis hasinherent limitations. Individual- and group-level conict factors, such as poverty orethnic hostility, are imperfectly tested atthe national level (Sambanis 2004). In suchcases, cross-country evidence (or the absenceof evidence) should be regarded with cau-tion. As we discuss in the next section, micro-level data is likely to yield more convincinganswers to the fundamental theoretical

questions. In our view, other analytical tools,from case studies to historical analysis, alsoremain useful.

3.1.1 Recent Cross-Country Empirical Advances

Recent cross-country research focuseson improving causal identication andmeasurement.

The search or exogeneity. The correla-tions of civil conict with both low incomelevels and negative income shocks are argu-ably the most robust empirical patterns inthe literature cited above, but the directionof causality remains contested. Even theuse of lagged national income growth (as inearlier studies) does not eliminate this con-cern, since the anticipation of future politi-cal instability and conict can affect current

30 The nding on population size suggests a possiblelink between population pressure—with its resultingresource scarcity and environmental degradation—andcivil conict, a theory that dates back to at least Malthus(for a review see Thomas F. Homer-Dixon 1999; ColinH. Kahl 2006). This population pressure hypothesis isrelated to the hypothesis that poverty increases armedcivil conict risk, where rapid population growth could beone driver of lower per capita income. The link betweenpopulation and conict is complex, however, and couldrepresent several causal factors (as well as systematic mea-surement error arising from the coding of battle deaths).See Gleditsch (1998).




investment behavior and thus living stan-dards.31 Another way of saying this is thatthere are likely to be permanent xed differ-ences between countries that are correlated with their income levels, economic growthrates, and civil war.

To address this concern, several papersseek to isolate exogenous variation in income.In sub-Saharan Africa, where most house-holds rely on rain-fed agriculture, fallingrainfall and drought cause large reductionsin income. Miguel, Shanker Satyanath andSergenti (2004) use annual rainfall growth asan instrument for income growth. The secondstage estimation equation is as follows, where

CW it is civil conict prevalence (or onset insome specications) in country i in year t, g it denotes per capita income growth, X it isa vector of population, geographic, politicalcontrols and social variables, α i is a country xed effect (capturing time invariant char-acteristics that relate to violence, growth, orboth), and δ i year t denotes a country-specictime trend:

(3) CW it = αi + X ′ it β + γ 0git

+ γ 1 g i,t−1 + δ i year t + εit .

They use annual country rainfall growthrates (current and lagged one year) as instru-mental variables for the per capita incomegrowth terms in equation 3. There is a rea-sonably strong rst stage relationship in thesub-Saharan Africa sample, but it is weakerin other world regions, where much less

economic activity relies on rain-fed agri-culture, making Africa the natural region

for the application of this approach. In theIV specication they nd that a 5 percentdrop in income growth increases the likeli-hood of a civil conict in the following yearby up to 10 percentage points, or nearly onehalf. Antonio Ciccone (2008) reaches thesame conclusion in a modied specicationusing log rainfall rather than rainfall growthas the key explanatory variable. This maineffect is not substantially dampened in coun-tries with stronger democratic institutions,greater ethno–linguistic fractionalization, oroil exporters.

This analysis highlights the role thatincome shocks play in generating armed

conict in Africa. Unfortunately, thiseconometric strategy once again does notallow the authors to denitively pin downa unique causal mechanism: rainfall shocksmay provoke conict because they lowerthe opportunity cost of ghting among ruralpopulations (those most affected by weathershocks), or because crop failure also reducesgovernment revenues and state capacity, orboth.32

Price shocks provide an alternative meansto study the income–conict relationship.Here the evidence is mixed. For instance,Besley and Persson (2008a) exploit interna-tional commodity price movements to inves-tigate civil war causes. Consistent with thepredictions of their theoretical framework(summarized above), rising import prices leadto greater conict, which they argue is due toa drop in real wages. Export price increasesare also associated with increased civil war

prevalence, since growing government rev-enue makes seizing the state increasingly

31 For a discussion of this theoretical point, seeChassang and Padro-i-Miquel (2007).

32 There may also be other violations of the exclusionrestriction unrelated to economic factors, for instance if rainfall directly affects the costs of ghting. Moreover, theauthors only study one type of economic shock, mainly

affecting the rural sector; variation in national incomeinduced by changes in industrial production or foreignaid could conceivably have different impacts. Future workshould also examine the possibility that droughts lead to violence between settled and nomadic groups, a salientissue missed in the existing civil war data.




attractive.33 However, Samuel Bazzi andBlattman (2010) reexamine the effect of tradeshocks on civil war using an expanded andmore disaggregated database of internationalcommodity prices and country trade shares.They suggest that previous results showing arelationship between trade shocks and politi-cal instability (e.g., Brückner and Ciccone2007; Angus S. Deaton and Ronald I. Miller1995; Besley and Persson 2008a, forthcom-ing) are sensitive to the denition of conictand to specication. Unlike rainfall, theseshocks show a less consistent relation to con-ict, whether they are experienced mainly by farmers (i.e., agricultural commodities),

the government (minerals and energy), or inthe aggregate. Nor do trade shocks robustly predict political instability in Africa wheninteracted with governance quality, ethnicfractionalization, or other common explana-tory variables, casting doubt on the oft-citedrelationship between trade volatility and civilconict, and with it the causal effect of low incomes on conict. This question—why massive trade and income shocks do not seemto systematically destabilize regimes—is animportant topic of further research, demand-ing better data, theoretical specication, andcase studies.

More important than generating any singleresult, however, these papers illustrate the

33 Markus Brückner and Ciccone (2007) use terms of trade shocks (driven by commodity price movements) asan instrument for national income. They nd a large effectof adverse income shocks on conict risk among undemo-

cratic African countries. This nding differs from Miguel,Satyanath, and Sergenti (2004), who do not nd any sta-tistically signicant interactions between income shocksand political institutions. Commodity price shocks, whileexogenous, are again plausibly not a valid instrument forincome; these shocks could affect conict via governmentinstability (due to collapsing revenues) or by heighten-ing inequality. In this case, a reduced form approach isless vulnerable to bias than the instrumental variable one(Bazzi and Blattman 2010). Political instability in evenmoderate-sized commodity producers could affect the world price, making commodity price shocks less exog-enous than rainfall.

advantage of quasi-experimental economet-ric approaches for distinguishing correlationfrom causation. Indeed, future cross-country empirical work should achieve more crediblecausal inference by focusing on a single, orsmall number of, exogenous conict deter-minants and plausible instruments for themrather than running horse races betweenmany endogenous variables.

More detailed and theoretically moti-vated measurement. Recent developmentsin the literature on natural resources andconict illustrate the value of better mea-surement.34 David K. Leonard and Straus(2003) emphasize the importance of enclave

production, which has little connection to theproductivity of most citizens (and thereforemay be less vulnerable to civil war violence).More accurate data have been compiled onoil production and reserves (Humphreys2005), while others have done the same forprimary and secondary diamond deposits(Elisabeth Gilmore et al. 2005), and mineralrents (Kirk Hamilton and Michael Clemens1999). Ross (2006) nds that these new andimproved measures of underlying hydrocar-bon and diamond deposits are strongly asso-ciated with more civil conict while oldernatural resource measures (based on actualproduction or exports) show less robust cor-relations. These ndings are consistent withthe contest model’s prediction that insurgen-cies ourish in resource rich regions becauseof the existence of more rents to ght overand the availability of easy nance, which arealso echoed by some case studies (Philippe

Le Billon 2001, 2005; Ross 2004a).There remains a need for better measures

of political grievances, institutional quality and even poverty. Consider political griev-ances rst. Much has been made of the weakcross-country association between armedconict and grievance proxies, including

34 For a review of the literature, see Michael L. Ross(2004b, 2006) and Humphreys (2005).




economic inequality and ethnic fraction-alization (e.g., Hegre and Sambanis 2006;Laitin 2007). This weak association is sur-prising given the robust negative relationshipbetween economic performance and somesocial divisions, as well as popular percep-tions of their centrality in driving conict(Alesina and La Ferrara 2005; Alesina andRoberto Perotti 1996; Easterly and Levine1997). However, if risk factors like inequality and ethnic fragmentation are measured withconsiderable error, or if their relationship toconict is conditional on particular institu-tional or historical contexts, then we shouldnot be surprised that their statistical asso-

ciation with conict is weak. A similar casecould be made about the existing and gener-ally crude measures of state capacity.

A more fundamental concern is thatthe existing proxies are theoretically inap-propriate. National income per capita, forinstance, may not capture the relevantaspects of poverty that drive ghting, suchas the proportion of rural male youth livingclose to subsistence income. Most measuresof ethnic and religious divisions are usedprincipally because they are straightforwardto calculate, rather than because they aretheoretically convincing. Indices of ethnicfractionalization have been questioned as ameaningful proxy for ethnic tensions (e.g.,Daniel N. Posner 2004a, 2004b). Here wehave again seen some progress in measure-ment. Esteban and Ray (1994, 1999) proposethat a bimodal distribution of preferencesor resources—“polarization”—is linked

to greater conict risk. Jose G. Montalvoand Marta Reynal-Querol (2005) createan empirical measure of polarization andnd support for Esteban and Ray’s theory: while fractionalization is not correlated withcivil conict, polarization predicts civil warincidence. More recently, measures of ethnicdominance—effectively indicators of minor-ity ethnic rule—have been explored; Lars-Erik Cederman and Luc Girardin (2007)

nd that minority ethnic rule is associated with increased risk of war, although onceagain this result may not be robust (Fearon,Kimuli Kasara, and Laitin 2007).

Another area of measurement concern isincome inequality. Some case studies sug-gest that ‘horizontal’ inequality—inequality that coincides with ethnic or other politically salient cleavages—is a particularly impor-tant driver of civil conict (Sambanis 2005;Frances Stewart 2001). Yet more work isnecessary to code these inequalities, as theexisting data remains fragmented and its sen-sitivity unexplored (Marie L. Besancon 2005;Gurr and Will H. Moore 1997; Gudrun Østby

2005). Even with better measures, it remainsdifcult to say whether it is the extent of inequality or its context (factors such as statestrength or the ideological climate) that mat-ter most (Cramer 2003, 2007).

The nding that many civil conictsare fought partially along ethnic linesalone is insufcient to make the case thatethnic-based grievances are driving the ght-ing. An alternative explanation, for example,is that the costs of organizing a rebellion(and collective action more generally) aresimply lower within ethnically homogeneousgroups. Heightened ethnic tension during acivil war might then be a result of the ght-ing rather than its cause.

Both the cross-country and case evidencehighlight the susceptibility of states with weak institutions to civil war. In particular,partly democratic societies (called anocra-cies in political science) have emerged as

prime incubators of civil conict. By thisargument, violent collective action occursbecause dissidents are free enough to orga-nize but nonviolent political activism is typi-cally ineffective (Fearon and Laitin 2003;Hegre et al. 2001).

Yet recent work suggests such ndingsmust be taken with caution. For instance,democracy and anocracy measures, com-monly based upon the Polity IV dataset




(Monty G. Marshall and Keith Jaggers 2006),explicitly use civil war and political violencein the coding of the data, thus mechanically correlating democracy and conict by deni-tion (James Raymond Vreeland 2008). Thesendings highlight the need for better mea-sures of state institutions and less reliance onexisting data. The importance of goverencepersists, however, even after accounting forthe endogenous coding of the Polity IV data.Goldstone et al. (2010) use conict forecast-ing to show that regime type is among themost robust predictors of civil war onset:regimes with restricted competition andsome repression of political participation

(anocracies) exhibit the highest relative riskof war, especially those regimes classied aspartly democratic but factional (in that thereare polarized competing blocs). Their resultsnd robust support in the case literature (e.g.,Sawyer 2004, 2005). While predictive ratherthan explanatory, such exercises emphasizethe importance of investigating the institu-tions–conict link further.

Reviewing the case literature, Sambanis(2005) suggests several possibilities await-ing empirical exploration: considering new versus established democracies separately;the mass popular inclusiveness of politicalinstitutions; the geographic concentration of power; and the degree of state control over acountry’s geographic periphery. Leonard andStraus (2003) also emphasize the importanceof direct taxation and institutions of personalrule.35 Several of these institutional charac-teristics have yet to be carefully dened and

measured; where they exist, moreover, they have not been tested against the alternatives.

Finally, while a degree of measurementerror in both dependent and indepen-dent variables is an unavoidable hazard of

35 In the former case, Leonard and Straus attempt todevelop a measure of state strength for Africa using directtaxes relative to national income in the early independenceperiod. Such worthwhile efforts should be extended.

cross-country work, few of the papers wereviewed weigh its consequences on theirestimates. The implications of measure-ment error ought to be discussed with thesame attention as endogeneity concerns.Fortunately, instrumental variables estima-tion addresses attenuation bias due to clas-sical measurement error, and this is onepromising way forward.

Integration with case studies. Historical-political analysis is the most time-honoredapproach to the study of civil conict. New efforts to integrate case analysis with cross-country statistical work look to build on thestrengths and minimize the weaknesses of

both approaches.The clearest example comes from the study of peacekeeping interventions. Ironically, thestudy of peace has given us some of the bestevidence we have for the rationalist rootsof war. Michael W. Doyle and Sambanis(2000, 2006) review the evidence on UnitedNations peacekeeping missions and nd thatthey are associated with a higher likelihoodof peace two years after the end of the war. Virginia Page Fortna (2004a, 2008) exam-ines the duration of peace with and with-out peacekeepers and reaches a similarconclusion. Both recognize the limitationsof econometric analysis when missions areselective and heterogeneous. Fortna shows,however, how nearly all observable determi-nants of peacekeeping interventions point tothe U.N. selecting the hardest, rather thaneasiest, cases and thus if anything her analy-sis may be underestimating peace-building

effectiveness.More revealing, however, is the insight

their cases bring to theories of conict.Doyle and Sambanis, and Fortna, concludethat interventions are effective becausepeacekeepers (i) change the economic incen-tives of the armed groups away from warfare;(ii) monitor and enforce compliance with thepeace agreement; and (iii) facilitate commu-nication between sides, reducing information




asymmetries. Agreeing to a foreign interven-tion, furthermore, is a means for both sidesto credibly signal a commitment to peace.If keeping the peace requires that externalactors resolve information asymmetries andcommitment problems, it seems likely thattheir absence contributed to war in the rstplace. The evidence is far from conclusive,but unlike most cross-country regressions,these case-based studies specically grapple with rationalist theories of war.

Multi-country case studies are also gen-erating new hypotheses and illuminatingsome of the causal dynamics driving civilconict (Cynthia J. Arnson and I. William

Zartman 2005; e.g., Collier and Sambanis2005a, 2005b; Fearon and Laitin 2005; Walter and Jack Snyder 1999). Generalizableor not, a single case can illustrate possiblecausal mechanisms, generate new hypoth-eses for testing, and stimulate innovativedata collection. While this case literatureis diverse and impossible to summarize infull, a number of inuential patterns andmechanisms stand out, including: the con-ict-provoking effects of commodity priceshocks on fragile economies (a claim withonly mixed cross-country empirical back-ing); the central role of external nancingto sustain insurgencies (including providingcross-border territory for camps, markets forextracted resources, and military aid); thepervasiveness of earlier state repression; per-sistent ethnic or elite class dominance; andthe emergence of insurgencies in peripheralregions where central government control is

weak.36

36 For instance, Annalisa Zinn (2005) studies theNigerian case, where many factors contributed to civil warrisk but there was no full-blown conict in the 1980s and1990s. The federal structure of Nigeria may have helpeddiffuse ethnic rivalry at the center. This argument echoesthat of Horowitz (1985) who, in his seminal contributionto the study of ethnic conict, argues for federalism asan institutional reform that changes the locus of politi-cal conict from the center to an increasingly large set of smaller conicts in different federal states.

Beyond borders. Another promis-ing direction is investigating civil conictcauses beyond the nation-state. One of themore novel approaches is taken by Andreas Wimmer and Brian Min (2006), who usexed geographic territories as the unit of analysis (rather than the more recent nation-state) over two centuries. They suggest thatthe likelihood of civil and interstate warshas been highest during the two massiveinstitutional transformations that shapedthe modern world: the nineteenth century incorporation of most of Africa and Asiainto European empires, and mid-twentiethcentury formation of nation-states in those

regions. Many wars, they argue, have beenfought to determine states’ governing struc-tures, and so are most likely to occur whenthese institutional principles are in ux dueto external geopolitical forces.37

More commonly, researchers lookingbeyond borders explore spillovers (or “conta-gion”) from neighboring countries. Hegre andSambanis (2006) nd that war in a geograph-ically contiguous country is a robust predic-tor of armed civil conict. Kristian SkredeGleditsch (2007) nds that the presence of trans-boundary ethnic groups increases con-ict risk, while having stronger democraciesin the region and more interregional tradeare both associated with less civil war. IdeanSalehyan and Gleditsch (2006) provide evi-dence for another potential source of con-ict contagion: refugees. Refugee ows canease arms smuggling, expand rebel socialnetworks, and provide a new pool of rebel

recruits. Looking beyond borders may alsochange our perspective on the role of ethnic“grievances.” Diasporas, whether in neigh-boring countries or farther aeld, driven by

37 Note that Alesina, Easterly, and Janina Matuszeski(2006) nd that articial states created largely during thenineteenth century wave of European colonization are nomore likely to experience civil or interstate warfare thanother countries.




ethnic or religious sentiments often play amajor role in rebel nance.

The empirical salience of these and otherinternational issues in driving domestic civilconicts (including the role of foreign aid,Cold War interventions, and cross-borderraids) highlights an important limitation of the existing theoretical work on armed con-ict causes, namely its almost exclusive focuson the internal armed groups’ decision of whether or not to ght. This is an importantdirection for future formal theoretical work,and will likely draw heavily on the existinginternational relations literature.

Confict duration and termination.

Researchers have also studied war dura-tion and termination.38 For instance, Fearon(2004), proceeds inductively, sorting cases by length and looking for salient patterns. Hends that short wars are disproportionately initiated by coups and popular revolutions, orarose from the breakup of the former SovietUnion and anticolonial wars, all of which seekpolitical control of the central government.Meanwhile, autonomy-seeking peripheralregion insurgencies and “sons-of-the-soil”movements (fought by the local majority against in-migrants) tend to last much longer.

Researchers commonly use a proportionalhazard model to analyze conict duration,employing a variety of economic and social variables to assess the role of greed, griev-ance and other factors in the length of civil wars (e.g., Collier, Hoefer, and MansSoderbom 2004). Others have introducedmore sophisticated methods, including

competing risk models (Karl R. de Rouenand David Sobek 2004). One nding is thatethnically fragmented and polarized coun-tries experience longer conicts (Sambanisand Ibrahim A. Elbadawi 2000; Reynal-Querol and Montalvo 2007). David E.Cunningham (2006) nds that conicts are

38 Roy Licklider (1993) provides an inuential early collection of case studies.

longer where multiple groups (“veto players”)must approve a settlement because thereare fewer mutually acceptable agreements,information asymmetries are more acute,and shifting alliances create incentives tohold out, complicating negotiations.

These duration analyses have been use-ful but suffer from many of the same chal-lenges as the onset and incidence literature:divergent results using different datasets;endogenous explanatory variables; and heroicinterpretations of proxy variables. Nonlinearhazard models also come with additionalidentication assumptions, and thus are sensi-tive to measurement error, repeated and con-

temporaneous conict events, and the unit of time used (Gates and Håvard Strand 2004). A typology o confict. Researchers have

analyzed civil war as a single phenomenon by assumption, leading some political scientiststo ask whether the heterogeneity in types of civil war should be explicitly incorporated intoempirical models. Are the 1967 Biafran sepa-ratist conict in Nigeria, Nepal’s Maoist insur-gency, and the long-running insurgency inColombia all examples of the same phenom-enon, or should we study them separately?

In response, several papers have begunto explore new civil war “typologies.” Somehave segregated wars by scale, distinguish-ing between “conicts” of 25 to 1,000 battledeaths per year, versus “wars” of more than1,000 battle deaths (Gleditsch, et al. 2002).Others, like Sambanis (2001), explore whether“identity” (i.e., ethnic and religious) wars havedifferent causes than ”nonidentity” wars.

Kalyvas (2005, 2007) and Laia Balcells andKalyvas (2007) suggest an alternative typology based on war origins and conduct, identifyingfour main classes: “conventional wars” (featur-ing regular armies and dened front lines);“symmetric nonconventional wars” with regu-lar armies ghting peripheral or rural insur-gencies; “symmetric irregular wars,” foughtbetween weak national armies and insur-gents; and the least common, “urban wars.”




Such subclassications are difcult to teststatistically, as they may only increase the volatility and sensitivity of empirical results(especially because it subdivides an already uncommon event into smaller subcategories where there may not be statistical power todemonstrate the validity of a typology). Tothe extent that rare events limit the statisti-cal power of such analysis, typology research will need to lead with theory or case studiesrather than regression analysis.

A further concern is that a generally accepted approach (and theoretical justica-tion) for subclassication will prove elusive;the type of civil conict that occurs—for

instance, a center-seeking versus autonomy-seeking civil conict—is endogenous to staterepressive capacity, rebel organizationalcompetence, underlying political grievances,and the likelihood of military success usingthat strategy relative to others.

Cramer (2007) challenges civil warscholars further, asking why civil wars areanalyzed as phenomena distinct from otherforms of political violence—communal riots,state massacres, and coups d’état. Neitherthe theoretical nor empirical case has beensettled for how to most usefully classify political violence into different categories.Moreover, the lines between wars, conicts,coups and communal violence are sometimesambiguous, potentially leading to errors of measurement. Exploring these categoricalassumptions is an interesting area for futureanalysis, one that could soon be more convincingly tackled with new data on non-

state armed conicts and “one-sided” state violence (Kristine Eck and Lisa Hultman2007; Human Security Report Project 2008;UCDP 2008).39

39 The Uppsala Conict Data Program (UCDP)denes non-state armed conict as the use of armedforce between two organized groups, neither of which is the government of a state, which results inat least twenty-ve battle-related deaths in a year.

3.1.2 Further Challenges and Paths Ahead for Cross-Country Empirical Work

Despite the empirical difculties, we donot believe that the cross-country regres-sion should be abandoned entirely. But thepath forward looks different than the onealready traveled. Existing empirical modelsare too rarely rooted in formal economic the-ories of conict, regression functional formsare too often ad hoc, the selection of prox-ies is driven by the variables easily at hand(or online), and their inclusion justied by informal arguments. As noted above, there isgood reason to believe that the relationships

between civil conict and income shocks,ethnic diversity and political grievancesshould be conditional ones, evident primarily when interacted with other contextual vari-ables, and the theorizing and testing of thesepotential interactions is a logical next step forcross-country research.

As this literature continues to advance,there are a handful of best practices to main-tain. First, relentless robustness and speci-cation checking. Second, a focus on causalidentication via the use of a single or smallnumber of exogenous instrumental variables.Third, the generation of new data on conictrisk factors and triggers, including bettermeasures of political grievance and pov-erty among key population subgroups, and various dimensions of state institutions andcapacity. Fourth, where measurement errorpersists, explicit attention to its ramications.

Although deriving policy implications is

not the main goal of this survey, there aresome implications of this literature worthspeculating about. The empirical relation-ship between violence and low and falling

They dene one-sided violence as the use of armedforce by the government or by a formally organizedgroup against civilians which results in at least twenty-ve deaths in a year, excluding extrajudicial killings(UCDP 2008).




incomes found in the cross-country literaturesuggests that implementing insuranceschemes to protect poor societies fromnegative income shocks could be fruitful inreducing the risk of civil conict. A numberof authors have recently proposed reformsto the design of foreign aid and to nationalagricultural policies to help blunt aggregateincome shocks and thus help avoid futurerounds of bloodshed (Collier and Hoefer2002). One possibility is expanded regionaldrought insurance for farmers. A variant isforeign aid contingent on objective conictrisk indicators (e.g., weather)—what Miguel(2007) calls “rapid conict prevention sup-

port”—to bolster local economic conditions when the risk of political violence is high.40

This example illustrates the potential valueresearch could have in informing policies toprevent civil war. Yet while we observe apoverty–conict link in the data, too littleis known about the precise identity of theactual perpetrators and organizers of vio-lence, so the question of which poor to tar-get with assistance to head off violence has yet to be decisively answered. It also remainsan open question whether interventions thatchange incentives for government and rebelleaders would be more cost-effective thanefforts to target the pocketbooks of potentialrebel recruits. A clearer understanding of rebel recruitment and organization is neces-sary for such assessments. These issues (andothers) are beginning to be addressed in theemerging applied microeconomic researchon civil war.

3.2 Micro-Level Empirical Evidence on theCauses of Civil War

The analysis of household and regional datais a growing, and perhaps the most promis-ing, new direction of empirical research.Three questions have been of greatest inter-est so far: (i) the roots of individual participa-tion in armed groups; (ii) the role of internalgeography in inuencing where and whencivil conicts are fought; and (iii) the organi-zation and conduct of conict.

3.2.1 The Decision to Rebel

Individuals are the natural unit of analy-sis for understanding how armed groupsmobilize civilians to ght and contributeresources to their cause. The black boxassumptions made in theoretical modelson group cohesion and origins need better justication. In response, the issue of col-lective action is the subject of a growingempirical literature.

The largest body of evidence comes fromcase studies of twentieth century rebel-lions. Several offer evidence consistent withmodels of self-interested actors seeking tomaximize material payoffs. For example,Mark Irving Lichbach (1994; 1995) illus-trates how successful social movementsoffer selective material incentives to joiners.Samuel L. Popkin (1979; 1988) nds thatpolitical entrepreneurs developed mecha-nisms to directly reward peasant rebellionin Vietnam. Weinstein (2007) illustrates

40 Targeting this aid toward the social groups most likely to participate in armed violence—for example, by fund-ing temporary job creation for unemployed young men, orcrop insurance for farmers, increasing the opportunity costof ghting in lean years—and in bolstering state capacity might be most effective in preventing armed civil conictsfrom occurring in war prone countries, most importantly in sub-Saharan Africa. Several African countries, mostnotably Botswana, have already successfully implementedsimilar national drought insurance programs includingpublic works employment, and these could serve as models

(Theodore R. Valentine 1993); Mick Moore and VishalJadhav (2006) discuss a related large-scale rural public works employment program successfully implementedin Maharashtra, India. Burke et al. (2009) argue that cli-mate change may increase civil war risk in Africa by over50 percent to 2030 and that this would make the need forinsurance against weather shocks even more pressing. SeeCullen S. Hendrix and Sarah M. Glaser (2007), MichaelKevane and Leslie Gray (2008), and David B. Lobell et al.(2008) for a range of perspectives on the issue of weather,climate change and conict in sub-Saharan Africa.




how in Mozambique, Sierra Leone, andPeru rebel ghters were remunerated vialooting of civilian property and drug sales.Material incentives may also be non-pecu-niary. Where violence against civilians iscommonplace, joining an armed grouphas in many cases been a path to relativesafety (Jeffrey Goodwin 2006; Kalyvas andMatthew Kocher 2007; Lichbach 1995; T.David Mason and Dale A. Krane 1989). Theprestige associated with martial successmay also be valued in and of itself.

Yet echoing our discussion above, materialincentives are not always a factor in the individual decision to ght, leading some scholars

to instead argue that moral, ideological, orethnic grievances mainly facilitate collectiveaction. Scott (1976) and Wood (2003b) arguethat moral outrage led people to rebel againstdeprivation during economic moderniza-tion in Southeast Asia, and over governmentabuses in El Salvador, respectively. In neithercase were selective material incentives appar-ent.41 Another literature documents how ethnic and social identities have been usedto identify, reward, and sanction free-riders,thereby providing selective social incentivesto participate (Moore 1993; Elinor Ostrom1990; Petersen 2001; Weinstein 2007).

A small but growing number of recentpapers employ within-country regional datato explore the factors that predict violence andrebellion, and most nd strong associations with local economic conditions. In Indonesia,Patrick Barron, Kai Kaiser and MennoPradhan (2004) nd positive correlations

between village-level communal violence

41 Some recent work on terrorism reaches similar con-clusions. Alan B. Krueger and Jitka Maleckova (2003)claim terrorists’ primary motive is passionate support fortheir cause and feelings of indignity or frustration, andthat poverty and education play a secondary role. Buenode Mesquita’s (2005) formal model allows terrorist orga-nization leaders to sort recruits by quality, possibly lead-ing to the positive selection of terrorists documented by Krueger and Maleckova.

and local unemployment, economic inequal-ity and natural disasters. Using data gatheredfrom newspaper reports, Daniel L. Chen(2007) nds that areas of high baseline religi-osity experienced more social violence in theaftermath of the Indonesian nancial crisis. InNepal, S. Mansoob Murshed and Gates (2005)nd a strong correlation between district-levelcivil war deaths and low living standards.Using the same conict outcome measure inseventy-ve Nepalese districts, Quy-Toan Doand Lakshmi Iyer (2007) nd that conictintensity is strongly and positively related tothe presence of mountainous and forestedterrain, as well as higher local poverty and

lower literacy rates, but is only weakly relatedto caste diversity. Karen Macours (2008) usesdifferent data to argue for another dimensionto Nepalese recruitment: Maoist insurgentsappear to have targeted the districts with thefastest recent growth in income inequality forrecruitment.42

These studies are informative and pio-neering but many suffer from challenges of data quality and endogeneity (limitationsthe authors are sometimes the rst to note).Further data selection bias worries are intro-duced when conict data are assembled from Western, English-language news reports.Moreover, individual motivations and deci-sions are difcult to infer from district-levelaggregate data; just as in the cross-country literature, there is too often a tendency tomake deep behavioral claims from simplecross-sectional correlations. The locationof ghting might also reect armed groups’

strategic considerations (i.e., the locationof important government military targets)

42 One of the drawbacks to reviewing a burgeoningliterature is that it is sometimes difcult for the outsider(and reviewer) to readily reconcile contrasting resultsfrom different dataset and papers on the same country.The variety of recruitment and violence patterns found inNepal alone is testament to this fact, and we look forwardto more cross-dataset comparisons and weighing of alter-native hypotheses for Nepal, as well as other conicts.




rather than the underlying socioeconomicconditions in those areas. Finally, thereremains the possibility of reverse causality;for example, in a single cross-section, conictcould contribute to poverty directly, as wellas be driven by poverty itself. Even paneldata is not immune to these concerns, due tothe economic changes driven by anticipatedfuture conict or other omitted variables.Nevertheless, the subnational approach is auseful step forward.

A recent study by Oeindrila Dube andJuan F. Vargas (2008) overcomes some of these concerns, employing exogenous priceshocks and detailed panel data on civil vio-

lence—guerilla and paramilitary attacks,clashes with government military, and civil-ian casualties—across over one thousandColombian municipalities. Consistent withtheir theoretical model (which builds onDal Bó and Dal Bó 2004), they nd thatan increase in the international price of Colombia’s leading labor-intensive exportcommodity, coffee, signicantly reduces violence in coffee-producing regions, whilean increase in the international price of animportant capital-intensive export good,petroleum, increases violence in regions withoil reserves and pipelines. In an important validation of their theoretical model, they then use rural household surveys to show that the positive coffee shock affects labormarket outcomes in the hypothesized way,boosting rural incomes and thus presumably raising the opportunity cost of participatingin rebellion. A limitation of the study is its

lack of data on actual individual recruitmentinto rebel groups or paramilitaries.

To the extent the patterns observed inIndonesia, Nepal and especially Colombiaare causal, the most likely interpreta-tion is that higher individual opportunity costs lower the probability of participa-tion in armed groups. We are hopeful thatincreased use of innovative micro datasets will yield a more complete view of the pov-

erty–conict relationship and clarify theexact mechanisms.43 For instance, it remainsunclear empirically whether it is usually thepoorest who actually ght in rebel groups,and none of these studies tells us by whatmeans collective action problems are over-come in forming and running armed groups.

A handful of individual- and household-level studies are beginning to answer thesecentral questions. Working in post-genocideRwanda, Philip Verwimp (2003, 2005) builta panel dataset based on a pre-genocide agri-cultural survey sample. He nds that bothpoor wage workers and land renters were dis-proportionately represented among genocide

perpetrators, and that they appear to havebeen motivated by interest in the property of landlords (who were disproportionately

victims), suggesting a class-based interpre-tation.44 In Sierra Leone, Humphreys and Weinstein (2008) collect post-war data oncombatants and noncombatants from thesame villages, and nd that retrospective pov-erty measures (e.g., mud housing), low educa-tion, and rebel promises of material rewardsare robustly correlated with recruitment intoarmed groups. However, proxies for politi-cal exclusion, such as supporting a nationalopposition political party, did not predictparticipation. Unexpectedly, they also ndthat the empirical determinants of voluntary and forcible recruitment were similar, sug-gesting that rebel leaders might be employ-ing selective rewards and punishments stra-tegically. Alternatively, it could point to thelimitations of postwar self-reported data

on the rebel participation decision, sincerespondents have strong incentives to lieabout the nature of their recruitment and

43 For example, Justino (2009) surveys the emergingmicro-level evidence and suggests that household poverty interacts with vulnerability (or risk of exposure) to vio-lence in complex ways.

44 Catherine André and Jean-Philippe Platteau(1998) nd related results. See also Straus (2006) on theRwandan Genocide.




wartime behaviors (claiming abduction evenif they in fact volunteered to ght), to escapesocial disapproval or even legal prosecution.

Taken together, these studies suggest thatmaterial incentives are inuential in drivingkilling even in the most brutal civil wars andin genocide—supposedly the quintessentialact of irrational hatred. Proxies for politicalgrievances perform far more poorly at pre-dicting individual behavior than economicfactors in these cases. Existing data on politi-cal grievances are admittedly quite coarseand may not adequately account for contextspecicity, but this provides evidence againstthe view that political grievances are always

decisive determinants of participation inarmed groups. We again return to the needfor detailed micro-data—ideally, individualpanel data collected comparably across coun-tries—on political attitudes and grievances,incomes and labor market opportunities, as well as local government budgets and secu-rity capacity in order to more conclusively disentangle competing explanations for par-ticipation in civil conict.

The above examples also call attention tothe importance of measuring the incentivesoffered by armed organizations. A robustassessment of competing explanations forrebel recruitment would require data onthe individual characteristics of rebel par-ticipants and non-participants, as well asthe recruitment “offers” received by bothtypes of individuals, both offers taken andthose refused. These data will obviously beextremely challenging to collect, especially

retrospective data given the high and selec-tive mortality experienced during wartime,and will likely require greater coordinationbetween researchers, governments andhumanitarian aid donors.

3.2.2 Internal Geography

Like recruitment, geographic patterns of conict within states are best explored usingsub-national data. To this end, researchers

in organizations like the International PeaceResearch Institute of Oslo (PRIO) and theUniversity of Uppsala have begun to con-struct and analyze sub-national conict data-sets. Early results are largely consistent withthe existing cross-country and case evidenceon the role of geography. Halvard Buhaugand Jan Ketil Rød (2006), for instance, dis-aggregate conict and geographic country data into 100 kilometer by 100 kilometergrids within Africa, and nd that separatistconicts are more likely to occur in sparsely populated regions near national borders, atgreater distances from the capital (wherepolitical control by the central government is

likely costlier), and in the vicinity of petroleumelds, where the rents of political power forsecessionists are presumably highest.45 Theexistence of easily lootable resources in thecontext of a bitterly poor society also drove violence in Sierra Leone’s war: there are sig-nicantly more armed clashes within chief-doms containing greater diamond wealth(John Bellows and Miguel 2009). Joshua D.Angrist and Adriana D. Kugler (2008) simi-larly nd that an increase in world cocaineprices led to increased civil conict violencein coca growing regions of Colombia relativeto other parts of the country.

3.2.3 The Organization and Conduct of Warfare

Another direction is empirical research onthe ghting factions themselves. Theoreticalprogress in this area accelerated some years

ago, and we now have a host of theories of

45 A closely related approach examines the geographicfeatures of civil wars, such as the size of the conict zoneor the distance from the capital. Buhaug and Gates (2002)build a database of battle sites and nd that wars of seces-sion and wars with an ethnic or religious dimension tendto fought in the peripheries. Larger conict zones areassociated with border zones, the presence of naturalresources, and peripheral conicts. While interesting, itremains to be seen what such associations can add to atheoretical understanding of civil war.




rebel and terrorist organization, several of which we reviewed above, as well as exten-sive case evidence.46

Systematic evidence on armed group orga-nization and action, however, has laggedbehind. Economists and political scien-tists have begun to conduct surveys of ex-combatants in Burundi, Colombia, Liberia,Sierra Leone, and Uganda (Jeannie Annanet al. 2008; Ana M. Arjona and Kalyvas2008; Humphreys and Weinstein 2004; EricMvukiyehe, Cyrus Samii, and GwendolynTaylor 2008; James Pugel 2007), some still works in progress. These surveys explore whothe combatants were, where they came from,

and their prewar experiences. Comparisonsacross armed groups, or to civilians, suggestmotivations for joining an armed group orcommitting particular acts of war.

Such data have limitations, however: they are self-reported; they are based on a sam-ple of survivors; and they can seldom movebeyond descriptive analysis of who joinsand why. Isolating exogenous variation inrecruitment, tactics, or exposure to violenceis crucial for drawing rm conclusions. Forinstance, to understand the determinants of civilian abuse, Humphreys and Weinstein(2006) construct military unit-level mea-sures of discipline and civilian abusivenessthat are exogenous to the individual respon-dent. However, the sample size is such that,unfortunately, there are seldom more thanone or two observations per unit. Beber andBlattman (2008) use new data on combatants,and exogenous constraints on rebel recruit-

ment, to understand the logic of coercivechild recruitment in northern Uganda.Unpopular and short of funds, the rebelLord’s Resistance Army had just one meansof gaining recruits: abduction, followed by the constant threat of punishment against

46 For case studies on African guerrilla movements,see Christopher S. Clapham (1998) and Morten Bøås andKevin C. Dunn (2007).

new recruits. The likelihood of receiving arearm and self-reported dependability wasincreasing in age, while loyalty and lengthof stay fall in age. From a rebel perspective,the intermediate age group—young adoles-cents—were the most attractive recruits,suggesting that coercion and child soldieringgo hand in hand.

Deliberately indiscriminate violenceagainst civilians may be another source of exogenous variation. Kalyvas (2006) docu-ments 100 studies and 45 historical cases where state violence against noncombatantsprovoked greater insurgent violence as aresponse. In a recent micro-empirical study,

Jason Lyall (2009a) examines the effect of Russia’s purposefully random shelling of Chechen villages on insurgent activity. Thispaper, one of the few to arguably study ran-dom violence, comes to the opposite conclu-sion: insurgent attacks in the village and itsneighbors decline after shelling.47 However,he is reluctant to draw conclusions about how violence in a particular place translates intolong-term outcomes there, and the generality of this result is unclear.

Counterinsurgency is a topic of major cur-rent interest in the wake of the U.S.-led warsin Iraq and Afghanistan. Military analystsand commanders have written extensively on theories and lessons learned in the eld(e.g., Richard L. Clutterbuck 1966; DavidGalula 1964; H. R. McMaster 2008; John A.Nagl 2002; David H. Petraeus 2006; Kalev I. Sepp 2005).48 Researchers have begun toinvestigate the efcacy of recent U.S. coun-

terinsurgency operations. Berman, Shapiroand Joseph H. Felter (2008) nd that the

47 Lyall (2009b) compares “sweep” operations by Russian and pro-Russian Chechen forces during the warin Chechnya (2000–2005). Co-ethnics appear to be betterat counterinsurgent operations: comparing sweeps in pairsof similar settlements, insurgent attacks increase afterRussian sweeps but decline after Chechen operations.

48 Stephen Biddle et al. (2008) review military counter-insurgency research in political science.




disbursement of small-scale reconstructionfunds by U.S. eld military commanders inIraq is correlated with lower levels of insur-gent attacks. Of course, selection on unob-served traits, or regression to the mean,could be driving this apparent effect; for thisreason, a number of experimental and quasi-experimental evaluations of counter-insur-gency spending are presently underway.49

Some recent work nds that armed groupsrespond strategically to new information.Radha Iyengar and Jonathan Monten (2008)develop data on insurgent attacks and mediacoverage in Iraq, and nd an “embolden-ment” effect of new information about U.S.

withdrawal intentions on the pace of insur-gent attacks. They use this evidence to show that insurgent organizations are sophisticatedstrategic actors, but while doing so also illus-trate the existence of asymmetric informa-tion between the warring parties. This workadvances our understanding of the empiricalrelevance of the information asymmetriesthat are so prominent in theoretical work.

3.2.4 Next Steps

There are four main limitations to this new applied micro-empirical literature. First, thenecessary datasets are expensive, hard-won,and often require a mix of luck and ingenuity.Hence they are too few in number. Second,sufcient attention has often not been paidto measurement issues, research design,and econometric identication. Too many researchers have rushed to collect micro-data without adequately preparing a research

design in advance, or testing the assump-tions required for causal inference. Third, itremains to be seen whether and how micro-level results will test conict theories. In

49 Of note is the U.S. Defense Department’s MinervaResearch Initiative, which in cooperation with a con-sortium of university researchers, is supporting rigorousimpact evaluations of development and military pro-grams to build peace in Iraq, Afghanistan, the Caucasus,Philippines, and elsewhere (Princeton University 2009).

particular, while counter-insurgency studieshave inherent military value, it is not alwaysclear how they relate to broader theoreticaldebates. Fourth, it remains to be seen how micro-level insights from one war general-ize to other contexts, and thus can be use-ful for policymakers elsewhere. At this early stage, this is primarily an argument in favorof increasing the number of micro-empiricalstudies on the causes and conduct of civil war.

4. Economic Legacies of Civil Conict

People living in zones of war are maimed,killed, and see their property destroyed.

They may be displaced, or prevented fromattending school or earning a living. A grow-ing empirical literature estimates the magni-tude of these effects of war on later income,poverty, wealth, health, and education.50 Each of these outcomes has implicationsbeyond the individual, however. To the extentthat these costs are borne unequally acrossgroups, conict could intensify economicinequality as well as poverty. The destruction(and deferred accumulation) of both humanand physical capital also hinder macroeco-nomic performance, combining with any effects of war on institutions and technology to impact national income growth.

Understanding the economic legacies of conict is also important to the design of postconict recovery. If war itself furtheraggravates factors that enhance the risk of civil conict—poverty, inequality, and socialdiscord—then it could partially account for

war reoccurrence.Indeed, the aggregate effects of armed

conict, and its threat, are considerable.Dani Rodrik (1999) argues that outbreaksof social conict are a primary reason why

50 Justino (2007, 2009) also surveys this emergingliterature. Many of the datasets and working papers arebeing shared via research groups such as the Householdsin Conict Network (http://www.hicn.org).




national economic growth rates lack per-sistence and why so many countries haveexperienced a growth collapse since the mid-1970s. A number of cross-country growthstudies link measures of political instability to large negative effects on national savings,investment, income and growth.51 ValerieCerra and Sweta Chaman Saxena (2008) ndthat output declines six percent in the imme-diate aftermath of a civil war. Quantitativecase evidence supports this cross-country relationship: Alberto Abadie and JavierGardeazabal (2003) nd that terrorist vio-lence in the Basque region of Spain has sig-nicantly reduced economic growth there

relative to neighboring regions. The effect onpoverty can be dramatic. In Rwanda, 20 per-cent of the population moved into poverty following the genocide (Justino and Verwimp2006). Civil wars may also have negativegrowth spillovers on neighboring countries(James C. Murdoch and Sandler 2004).

An economic growth theory frameworkis useful for analyzing the consequences of conict. If conict affects economic perfor-mance, it must be because it affects a fac-tor of production (physical capital, labor, orhuman capital), the technology, institutions,and culture that augment these factors, orprices (e.g., costs of capital). The growthframework also claries the possible natureof the impacts, not only on income levelsand economic growth in equilibrium, butalso out-of-equilibrium dynamics such as thespeed of convergence.

The framework we use to organize our

discussion is based on neoclassical mod-els of growth with human capital (e.g.,Robert E. Lucas 1988; N. Gregory Mankiw,David Romer, and David N. Weil 1992).

51 See Robert J. Barro (1991), Alesina et al. (1996),Alesina and Perotti (1996), and Svensson (1998). Thepolitical instability-growth relationship may be partly endogenous but some argue that the association is l ikely topersist even after better accounting for this bias (KwabenaGyimah-Brempong and Thomas L. Traynor 1999).

Alternative frameworks, however, can gener-ate radically different predictions regardingthe likely impact of violence—and, in par-ticular, the destruction of capital—on eco-nomic performance. To illustrate, a one-timeshock to capital has no effect on equilibriumincome or growth in a neoclassical model,but persistent effects are possible in poverty trap, endogenous growth, and vintage capi-tal models (e.g., Costas Azariadis and AllanDrazen 1990; Barro and Sala-i-Martin 2003;Simon Gilchrist and John C. Williams 2004).

The relative degree of physical and humancapital destruction also matters: recovery could be faster under highly asymmetric

destruction—say, extensive physical capitaldestruction when human capital remainslargely intact—since the relative abundanceof one type of capital raises the marginalproduct of the scarce type, spurring oninvestment. Barro and Sala-i-Martin (2003,p. 246) describe this “imbalance effect” in aone-sector endogenous growth model withphysical and human capital, with the follow-ing production function:

(4) Y = AK α(Lh)1−α,

where A, K , and L have the usual inter-pretations as technology, physical capital,and workers, respectively, h denotes aver-age worker human capital, and total humancapital is H = Lh. They examine the case where the K / H ratio deviates from its steady state value of α /(1 − α), for instance due to war damage; capital investments are irre-

versible; and there are adjustment costs tocapital accumulation. When adjustmentscosts for human capital accumulation aregreater than for physical capital investment, which seems plausible empirically, the dis-proportionate loss of human capital in warresults in slower economic growth andrecovery than the destruction of physicalcapital, during the transition back to steady state growth.




Given the proliferation of plausible theo-retical perspectives, empirical evidence isessential. Yet assessing the economic conse-quences of civil war is complicated by a centralidentication problem: war-torn countriesare different than peaceful ones (as detailedin sections 2 and 3). Poor postwar economicperformance could reect the declining eco-nomic conditions that contributed to armedconict in the rst place, in addition to any direct impacts of war. Similar endogene-ity concerns arise in assessing impacts ongovernment performance and institutions,and even individual-level outcomes (to theextent different types of individuals are tar-

geted for violence or recruitment into armedgroups). The existing empirical literature onpostwar economic recovery is only begin-ning to seriously address these issues, and asa result the conclusions we can draw aboutthe consequences of war, and the appropri-ate postwar policy responses, are more lim-ited than in the case of civil conict causesdiscussed above.

4.1 Physical Capital and Investment

Evidence from interstate wars suggeststhat the postwar evolution of physicalcapital often behaves as predicted by theneoclassical model, namely, rapid recov-ery to equilibrium levels. One set of stud-ies examines the impact of U.S. bombingon later outcomes at the city or regionallevel. Although they generally lack detailedinformation on local physical capital lev-els, in Japan (Donald R. Davis and David

E. Weinstein 2002) and Germany (StevenBrakman, Harry Garretsen, and MarcSchramm 2004) in World War II, citiesthat were heavily bombed quickly recoverin population back to prewar trends, suchthat 20 to 25 years postwar city populationsare indistinguishable from cities that wereleft untouched by bombing. In the Vietnam War, which combined external interventionand a civil war, Miguel and Gerald Roland

(2006) nd similarly rapid local populationrecovery from bombing.

These cross-region results echo the con-sensus from the cross-country literature onthe rapid recovery of postwar economies (A.F. K. Organski and Jacek Kugler 1977, 1980;Adam Przeworski et al. 2000). Indeed, arecent study of the output response to alter-native crises—including currency crises,banking crises, civil war, and sudden shifts inexecutive power—nds that while civil warscause the steepest short-run fall in output(six percent on average), only in the case of civil war does output rebound quickly, recov-ering half of the fall within a few years, while

output drops are more persistent for nancialcrises (Cerra and Saxena 2008). While suchevent studies conceal a great deal of hetero-geneity in experiences, and suffer from obvi-ous omitted variable bias concerns, none of these results appear consistent with poverty trap models of economic growth such asthose recently advanced by Jeffrey D. Sachs(2005).

Nevertheless, there are reasons to becautious in generalizing these experiences.These studies cannot rule out the possibil-ity that the economic devastation causedby civil war prevent some countries fromachieving durable peace. Countries with suc-cessful postwar economic recovery are alsomore likely to collect systematic economicdata, introducing possible selection bias: war-torn countries where the economy andinstitutions have collapsed (e.g., Congo andSomalia) lack good data, while those that

recover (Vietnam) have data. This couldbias the cross-country estimates of war’seconomic impacts towards zero as the mostdestructive wars exit the sample. Civil warsare also often localized and fought with smallarms and munitions, so they do not necessar-ily see the large-scale destruction of capitalcaused by bombing, creating some separa-tion between the evidence we have and thecontexts of greatest interest.




Yet even in civil conicts without large-scale bombing, capital can sometimes bedepleted in devastating ways. First, house-hold assets may be stolen or destroyed.Mozambicans, for instance, are thoughtto have lost 80 percent of their cattle stockduring their civil war (Tilman Bruck 1996), while many in northern Uganda lost allof their cattle, homes and assets (Annan,Blattman, and Roger Horton 2006; RobertGersony 1997); cattle and other farm assetsoften represent most of a rural household’ssavings. As of yet, however, there is still lim-ited systematic panel data on the implica-tions of such asset loss on long-run household

welfare. Second, countries at war are likely to see massive ight of mobile forms of capi-tal, since foreign assets offer higher relativereturns at lower risk (Collier 1999; Collier,Hoefer, and Catherine Pattillo 2004). Thesame factors could lead to such low levels of new investment that the existing capital stockquickly deteriorates.52

The neoclassical growth model predic-tion that the capital stock should returnto its steady state level once the ghtingstops—implying relatively high returnsand rates of investment that decline as theequilibrium is approached—supposes thatunderlying institutions and technology arelargely unaffected by the ghting, and thatmilitary spending, the returns to capitalinvestment and the cost of capital similarly return to prewar levels. Yet any political oreconomic uncertainty following war is likely to decrease expected returns, increase rela-

tive risk, and possibly shorten investment

52 Rising military spending can also crowd out govern-ment infrastructure projects and other public goods. A World Bank report estimates that average military spend-ing in poor countries rises from 2.8 percent of nationalincome in peace to 5 percent at war (Collier et al. 2003).Cross-country evidence suggests that such military spend-ing is growth-retarding due to the shift away from produc-tive investment (Norman Loayza, Malcolm Knight, andDelano Villanueva 1999).

horizons, thus reducing investment and rais-ing the cost of capital. Collier (1999) arguesthat adverse effects on the cost of capital aresometimes persistent empirically.

Foreign nancial aid and other interna-tional interventions could play an importantrole in rebuilding infrastructure and replen-ishing household assets in these cases. Thereis also anecdotal evidence from countrieslike Sierra Leone and Liberia that the role of the international community was decisive inshifting expectations about future conict risk(Collier 2007). Collier and Hoefer (2002)suggest that increased foreign aid is likely to reduce civil conict risk, and nd some

modest reductions in the likelihood of con-ict for aid recipients, working through thechannel of faster economic growth. Yet thenonrandom placement of both civil conictsand foreign aid means we cannot necessar-ily interpret these statistical relationships ascausal. Joppe De Ree and Eleonora Nillesen(2006), however, examine aid disbursementsand civil conict risk in sub-Saharan Africausing an instrumental variable strategy thatexploits exogenous changes in donors’ overallforeign aid budgets, and they nd that a 10percent increase in aid to an African coun-try reduces conict risk by 6 percent. Takentogether, these two studies suggest that postwar foreign aid may play a key role in solidi-fying the transition to peace.53

Future advances could come from disag-gregating such postwar aid ows and inves-tigating the role of specic activities (such aspeacekeeping) on reducing risk and promot-

ing economic recovery. Recently Collier, LisaChauvet and Hegre (2008) attempt just sucha cost–benet analysis of alternative aid andmilitary interventions. Such calculations are

53An emerging literature examines the role that postwar demilitarization programs (usually funded by aiddonors) could play in securing the peace, although thereis little quantitative evidence on the effectiveness of theseprograms.




highly dependent on parameters for which wecurrently have limited data (e.g., the growtheffects of avoiding a civil war, the selectionof countries into different interventions) andthus have to be interpreted with caution.

4.2 Life, Labor, and Human Capital

Wars kill and maim people, both directly and indirectly through famine and disease.Conict victims are overwhelmingly civil-ians, and indirect deaths are seen dispropor-tionately among the poor, women, childrenand the elderly.54

The short-run impact of war is clearly disas-trous, but there is mixed evidence on how long

the economic effects on human capital andquality of life persist. In the study of Vietnambombing mentioned above, local living stan-dards and human capital levels also convergedrapidly across regions after the war, leavingfew visible economic legacies twenty-ve yearslater (Miguel and Roland 2006). This empiri-cal nding echoes the cross-country literatureshowing rapid post-war economic recovery and argues against poverty trap type models.An innovation is the attempt to address theendogeneity of bombing. Miguel and Rolandinstrument for bombing intensity using dis-tance from the arbitrarily settled North-South Vietnamese border (on the 17 th parallel northlatitude).55 A limitation of this paper—andothers that examine differences across sub-national units over time—is its inability tocredibly estimate the aggregate national eco-nomic impact of war damage; for that, cross-country studies may be more convincing.

A new and rapidly growing microeconomicliterature nds more persistent negative war

54 Wars are thought to have directly caused 269,000deaths and 8.44 million disability-adjusted life-years(DALYs) in 1999 alone, with twice again this number of deaths and DALYs estimated in 1999 due to the linger-ing effects of wars between 1991 and 1997 (Hazem AdamGhobarah, Paul Huth, and Bruce Russett 2003, 2004).

55 Districts located near the border were subject to moreghting, cross-border raids, artillery shelling, and bombing.

impacts on individual human capital, espe-cially in African cases. Using panel dataon child nutrition, Harold Alderman, JohnHoddinott and Bill Kinsey (2006) nd that young children who suffered from war-related malnutrition in Zimbabwe are sig-nicantly shorter as adults, which may affect their lifetime labor productivity. In arelated paper, Tom Bundervoet, Verwimp,and Richard Akresh (2009) exploit varia-tion in the timing of armed clashes in theBurundi civil war to estimate impacts onchild nutrition, and nd that children wholived in a war-affected region have sharply lower height-for-age than other children,

with an average drop of roughly 0.5 standarddeviations. Turning to a Central Asian set-ting, adolescent Tajik girls whose homes were destroyed during that civil war are lesslikely to obtain secondary education, again with likely adverse effects on later wagesand life chances (Olga Shemyakina 2006).The validity of these studies, all of which usedifference-in-differences methods, relies onthe assumption of similar underlying humandevelopment trends in the war-affected andpeaceful regions of these countries, some-thing that is challenging to convincingly establish with the limited time horizons of most datasets.56 Moreover, as in the bombingstudies, these studies may underestimate war’s overall impacts to the extent that eventhose in largely peaceful regions were alsoadversely affected by civil war disruptions.

Turning to combatants, it appears that theinterruption of human capital accumulation is

one of the most pervasive impacts of military service. Studies of U.S. and European veter-ans of the Vietnam and Second World Warsnd large and persistent falls in earnings

56An unexpected spillover effect of war on the humancapital of neighboring countries comes from Montalvo andReynal-Querol (2007) who, using civil wars as an instru-mental variable, argue that for each 1,000 refugees thereare between 2,000 and 2,700 additional cases of malariain the refugee-receiving country.




and higher mortality (Angrist 1990, 1998;Angrist and Krueger 1994; Norman Hearst,Thomas B. Newman, and Stephen B. Hulley 1986; Guido Imbens and Wilbert van derKlaauw 1995). These patterns are echoed by new evidence from developing countries. Forinstance, Blattman and Annan (forthcoming)and Annan et al. (2009) use exogenous varia-tion in rebel recruitment methods—namely,near-random forced recruitment in ruralUganda—to estimate its impact on adoles-cents and young adults. These conscripts aremore likely to have persistent injuries, accu-mulate less schooling and work experience,are less likely to be engaged in skilled work,

and earn lower wages as adults (especially males). Psychological trauma and commu-nity rejection, meanwhile, are concentratedin the small minority that experienced themost violence.57 The conclusion that emergesis that military experience is a poor substi-tute for civilian education and labor marketexperience. In settings where a large share of youth actively participate in ghting, aggre-gate economic impacts could be quantita-tively important.

This emerging applied microeconomic lit-erature only scratches the surface of the rangeof possible civil war impacts on the economy and society. More evidence is required onthe educational, employment, and healthimpacts of conict on armed group partici-pants and civilians, including internally dis-placed people. The leading question is not whether wars harm human capital stocks, butrather in what ways, how much, for whom,

and how persistently—all crucial questionsfor understanding war’s impacts on economicgrowth and inequality, as well as priorities forpostconict assistance. To our knowledge, no

57A somewhat different pattern is observed by Humphreys and Weinstein (2006, 2007), who nd thatincreases in Sierra Leone ghters’ exposure to violenceare correlated with lower postwar community acceptance,but also nd that violence has little correlation with labormarket success.

rigorous evidence yet exists on which typesof programs are most effective at overcom-ing war’s adverse legacies on human capital.58

4.3 War, Institutions, and Society

The steady state to which a postconictsociety returns is a function of the fundamen-tal determinants of growth: technology, insti-tutions, and social organization. The rapidreturn to prewar levels of labor and capital inGermany, Japan, and Vietnam noted abovesuggests that these determinants were notdiminished by war (or, if they were, they likewise recuperated quickly).

Unfortunately, we have little systematicquantitative data with which to rigorously judge claims about the evolution of institu-tions during and after civil wars. A sizableliterature has sought to identify the spe-cic institutional factors that matter mostfor economic growth—including property rights (Acemoglu, Simon Johnson, andRobinson 2001), social capital and cohesion(Stephen Knack and Philip Keefer 1997),rational bureaucracies, and work ethics, toname a few—but which of these are affectedby civil war (not to mention how much andunder what circumstances) remains a mat-ter of speculation. The social and institu-tional legacies of conict are arguably themost important but least understood of all war impacts.

58Other potentially important topics awaiting system-atic empirical analysis include: the role of war-related

emigration (especially of the skilled) on later economicgrowth, the general equilibrium effects of death andemigration on labor markets, and civil war’s effects onthe prices of land, capital, and labor. Abbey Steele (2007)includes a review of the determinants of population dis-placement, while Florence Kondylis (2008a, 2008b) pres-ents quasi-experimental evidence on some of its economicimpacts. To the extent that the sudden death of sizableshares of the working age adult population affects rela-tive prices, fertility and investment decisions, civil warcould have impacts on living standards reminiscent of theHIV/AIDS epidemic in sub-Saharan Africa along the linesargued in Alwyn Young (2005).




The historical evidence (described above)that war enables the development of capablegovernment institutions in Europe may notgeneralize to civil war cases. In the threecountries that experienced U.S. bombing,the wars were fought against (largely) for-eign armies, and hence could rally citizensand renew government motivation and legiti-macy. In civil war, government may loselegitimacy, while victors and vanquished(and victims) are condemned to coexist in thesame society, potentially exacerbating politi-cal and social divisions.

Yet even the waging of internal war neednot always be uniformly destructive to insti-

tutions. The effort to control the nation’speripheries, and the extension of nationalcontrol down to the community level areessential state responsibilities. Successfulstates do so through a variety of means,including the use of force (Tilly 1982;Max Weber 1965). Hence internal warfarecould hypothetically generate state-build-ing rather than institutional disintegra-tion. Yoweri Museveni’s violent takeoverof the Ugandan state, for instance, hingedin part on his ability to organize citizencouncils from the village up to the nationallevel—councils that became the basis forpostwar administration and (especially by regional standards) a relatively strong state(Weinstein 2005a)

Indeed, there is some cross-country evi-dence that wars that end in outright mili-tary victory for one ghting side lead to amore stable peace and possibly stronger

state institutions (Fortna 2004b; MonicaDuffy Toft 2008), although once again theomitted variables correlated with outrightmilitary victories make interpretation of this pattern difcult. A draw, a negotiatedagreement, or a ceasere is, by this logic,an unstable equilibrium bound to unravel.This belief has led some scholars to arguethat the international community should“give war a chance” (Herbst 2000; Edward

N. Luttwak 1999). In an indication of how unsettled the literature currently is, anotherrecent case study and statistical analysisindicates that aggressive peacekeeping leadsto more lasting peace (Fortna 2008; HumanSecurity Report Project 2008; Sambanis2007). Easterly (2008) argues against this view, however, pointing out that there areas many examples of failed foreign interven-tions (e.g., Darfur, Democratic Republic of Congo, Somalia) as successes (Sierra Leone).

One reason why research has produced few denitive answers so far may be that there isno simple, general relationship between civil

war and institutions; impacts may depend

on why a war started in the rst place, how it is fought, and how it ends.59 The unpack-ing of these complex relationships is perhapsthe most pressing area for future empiricalresearch in this area. While the war termina-tion literature tends to focus on the durabil-ity of peace agreements, a useful next step isto focus on patterns of institutional changeduring and postconict. While the analysisof these relationships is clearly full of omit-ted variable bias concerns, the search forinnovative research designs should continue.Beber (2008), for instance, uses the higherlikelihood of mediated bargaining occurringduring the summer months (when foreignpoliticians and diplomats are freer to brokerdeals) to identify the gains to mediation ininterstate conicts.

A nal perspective on the institutional andsocial impacts of war comes from an empiri-cal literature on postwar political participa-

tion and accountability. Of the preliminary data available, some of the results are quitecounterintuitive, especially an apparentcausal link between war violence and pro-ductive citizenship postwar. A recent micro-study nds that war victimization increaseslater individual political mobilization and

59 The literature on war termination in particular is vast, and is outside the scope of this review.




participation in local collective action inSierra Leone, which the authors interpretas a result of the psychological legacies of individual violence exposure (Bellows andMiguel 2006, 2009). Former combatantsin Uganda were also more likely to voteand become local leaders (Blattman 2009).Likewise, psychologists nd that the victimsof violence are in general resilient (Ann S.Masten 2001), and that exposure has even ledto greater political activism among groupssuch as Jewish Holocaust survivors (DevoraCarmil and Shlomo Breznitz 1991) andPalestinian victims of bombardment (Raija-Leena Punamäki, Samir Qouta, and Eyad El

Sarraj 1997). Together, these ndings beginto challenge the notion that the motivations,institutions, and social norms that promotelocal level collective action are necessarily harmed by civil war.60

4.4 Remaining Challenges

Viewed through the lens of economicgrowth models, the existing empirical liter-ature on civil war impacts still looks spotty.Macroeconomic studies indicate that theshort-run output effects of armed conictcan be large, but more work is needed toexamine the underlying effects on factorsof production and relative prices. The early signs suggest that population and physicalcapital can fully recover after war, perhapsas quickly as within two decades. Thatrecovery, however, appears to be contingenton the preservation, or even the improve-ment, of political stability and institutions,

as was the case in Japan, Germany, and Vietnam. Yet what the key institutions are,and which domestic policies and external

60 Yet ndings are far from uniform. Miguel, SebastianM. Saiegh, and Satyanath (2008) argue that civil warsmay shape national socio-cultural norms toward violence.They nd that European soccer league players from coun-tries with histories of civil war commit signicantly more violent yellow and red card fouls (conditional on playercharacteristics).

interventions can help maintain their sta-bility, are still poorly understood. Eventheoretically compelling patterns could bespurious, driven by omitted variables ratherthan causal impacts.

The microeconomic literature is even lesssystematic at present, although in our view itholds great promise. Many factors appear tobe adversely affected by civil war in at leastsome cases. Those who participate in wars, orsimply live through them, often suffer frompersistent injuries, lose out on education,and see a permanent decline in their pro-ductivity and earnings. But understanding which impacts are more profound and per-

sistent than others; which disproportionately strike the poor; and how those effects can becontained by local institutions and economicpolicies is still largely unexplored.

Without rm answers to these questions,policymakers and foreign aid donors haveoften taken a scattershot approach to postwar programs. The subject of post-conictrecovery policy is vast and is largely outsidethe scope of this review, but most of that lit-erature comes in the form of best practicessummaries, case studies, and other litera-ture produced by international aid organiza-tions, governments, and NGOs. Academicresearch remains limited, and where it exists,tends to focus on high-level analysis (e.g.,the relationship between aggregate foreignaid and national economic growth) and sois largely unhelpful to those seeking specicprogrammatic solutions. Given the many possible omitted variables involved in the

timing of foreign interventions, related toboth domestic and international political fac-tors, establishing the causal impact of armedintervention on long-run political and eco-nomic outcomes has been elusive.

An obvious answer is to call for more datacollection, continued searching for naturalexperiments or actual eld experiments, andmore rigorous impact evaluations of postcon-ict programs. Judging by the rise of research




organizations like the Households in ConictNetwork (HiCN), increased funding by the World Bank’s Development Research Group,and the growing literature, this call for moreand better analysis is already beginning to beanswered.

5. Discussion and Future Directions

Armed conict is nally moving into theresearch mainstream in development eco-nomics. This article has attempted to survey this ourishing interdisciplinary eld,describe its more robust ndings, and pointthe way forward in a way that is useful for

both those new to the eld as well as thosealready actively working within it.Some of the core insights are worth re-

stating here. First, there has been consider-able progress in the formal modeling of thepolitical economy of civil war during thepast two decades, with insights on the individual decisions, institutional features, andeconomic conditions that promote violentconict. Commitment problems—eitheracross the two sides to a conict, or amongfactions within a ghting side—are currently viewed as the leading rationalist theoreticalexplanation for civil war, especially for long-duration civil wars, although certain typesof information asymmetries may also play a role. Disentangling the relative contribu-tions of the various commitment problemsand information asymmetries proposed inthe theoretical literature is a top priority forempirical research. Developing new expla-

nations—possibly challenging the currentmodeling assumptions of unitary armedgroups, or even rationality—is also likely tobe fruitful.

Second, a variety of theoretical modelspredict that low incomes, weak state insti-tutions, and social divisions may contributeto the onset of civil wars, and these issueshave been the focus of most empirical treat-ments. The most robust empirical nding in

the existing literature is that economic con-ditions—both low income levels and slow growth rates—contribute to the outbreakof civil wars and conicts in less developedcountries. This nding has found support atboth the cross-country and the micro levels,although the correct interpretation of thesepatterns in terms of underlying theoreticalmechanisms remains contested. A smallerliterature suggests that economic factors aredecisive in driving individual participation inarmed groups. However, the theoretical andempirical conict literatures have too oftenrun along parallel paths, informing eachother, yes, but seldom directly intersecting;

greater efforts need to be made to identify and test the precise empirical implications of the leading theoretical frameworks.

In contrast, the empirical evidence thatsocial divisions, political grievances, andresource abundance are drivers of violenceremains weaker and more controversial.The existing literature tends to measurenon-material factors crudely, and empiricaltests rarely attempt to capture the nuancesof a social phenomenon as complex as civil war, making it impossible to decisively rejectthat nonmaterial factors are playing somerole. Further research using better data isneeded to rmly settle the question of whatrole political grievances play in driving civilconicts.

There is also an emerging literature on theeconomic legacies of war. At this point themacro literature and newer micro literaturehave produced somewhat contradictory nd-

ings, although they can potentially be recon-ciled by appealing to the divergent outcomesthey consider. The macro literature focuseson physical capital, economic growth, andpopulation, while the micro literaturemainly on human capital. Recall that stan-dard economic growth models, including theBarro and Sala-i-Martin (2003) frameworkdescribed in section 4, predict that the loss of human capital will have more lasting adverse




economic growth consequences than thedestruction of physical capital. Future workmust clarify how the nature of the conict(internal versus international) as well as thepolitical, social, and institutional contextaffects long-run economic growth, and justas importantly, must more seriously addressthe many omitted variables that could simul-taneously drive the outbreak of wars andaffect postwar economic recovery. The neo-classical economic growth framework use-fully highlights the most important gap inour knowledge: the impacts of internal waron institutions, technology, social norms andculture. Progress on these issues is critical for

crafting appropriate postwar recovery poli-cies, a major economic policy issue in con-ict-prone regions, including sub-SaharanAfrica.

Throughout this discussion, a key lessonthat emerges is the important role that new data sources have played in enabling researchprogress. The development of the PRIO/ UCDP civil conict database has propelledthe cross-country conict literature forward.Disaggregated data on U.S. bombing patternsallowed Davis and Weinstein (2002) to carry out their seminal study on Japan. The increas-ing number of longitudinal household-leveldatasets in less developed countries havemade the new micro studies on war impactspossible. Some of this data collection hasrequired remarkable ingenuity and courageon the part of the investigators, notably, thecollection of data on civilians and combatantsfrom ongoing or recently concluded wars.

Yet much more work remains to be done. Inour view, a major goal of civil war researchers within both economics and political sciencein the coming years should be the collectionof new data, especially extended panel micro-data sets of economic conditions and oppor-tunities. Ideally, these efforts would also becoordinated, publicly shared and compa-rable, in a similar fashion, say, to the WorldBank’s Living Standards Measurement Study

(LSMS) or the Demographic and HealthSurveys (DHS) program.61

Data collection is of course inherently difcult in “hot” conict zones. But evenin many postconict settings where condi-tions are closer to normal, statistical agen-cies simply return to the status quo of survey instruments, and fail to valuably collect ret-rospective conict experience data. To illus-trate from the authors’ own experiences inLiberia, Sierra Leone, and Uganda, neitherthe government statistical agencies nor theinternational donors nancing reconstruc-tion there had plans to systematically includequestions on war experiences, victimiza-

tion, or participation in the national censusor other representative household surveysconducted at conict’s end. Where nationalsurveys of war experiences were conducted,they tended to focus exclusively on combat-ants. Closer cooperation among governmentdata collection agencies, development orga-nizations, and researchers will be requiredfor the systematic and comparable dataneeded to make further progress.

A few specic data collection directionsappear particularly promising. First, moredetailed information on rebel organizationand decision making would be useful. Somecivil conicts, like those in Congo and Sudan,feature a dozen or more active armed groups.Pooling data from several such settings couldallow for a relatively large sample analysis of armed groups—or “cross-rebel regressions”—establishing patterns useful to applied theo-rists working on rebel organizations. While

the proliferating number of surveys of ex-combatants (described earlier) will improvesuch analysis, so far these datasets have beenselective: most focus on non-state actors and

61 For LSMS, see http://go.worldbank.org/IPLXWMCNJ0. For DHS, http://www.measuredhs.com. The insti-tutional basis for such coordination may already exist inthe Households in Conict Network (http://www.hicn.org).




nearly all on Africa. The risk is that our under-standing of civil war will be driven by a subsetof conicts rather than a more globally repre-sentative sample. New data from Afghanistan,Colombia, Indonesia, Iraq, and Nepal (dis-cussed above) is starting to ll in the gaps.

Second, and perhaps easiest to collect, arefollow-up household surveys in post-conictsettings integrating retrospective informa-tion on a wide range of economic behav-iors and experiences during the war, alongthe lines of the data work in Bundervoet, Verwimp, and Akresh (2009) in Burundi,Bellows and Miguel (2006, 2009) in SierraLeone, Blattman and Annan (forthcom-

ing) in Uganda, and Verwimp (2005) inRwanda. The Burundi and Rwanda surveysdeserve special mention. In both cases theauthors identied a prewar national house-hold survey, located the original (archived)surveys, and tracked down the sample house-holds again after the war. Such intellectualentrepreneurship should be expanded andrewarded in the profession.

Third, we need to improve measures of political attitudes and grievances and testtheir association with actual behaviors. Oneexample stands out. James Habyarimana etal. (2007; forthcoming) identify theoretically distinct mechanisms that link ethnic diver-sity to trust, cooperation and public goodsprovision, and then run experimental gamesto compare the explanatory power of the dis-tinct mechanisms in a representative sampleof 300 subjects from a Kampala, Uganda,slum neighborhood. Experimental econom-

ics lab research in other developing coun-tries, especially in conict and postconictsocieties, could shed light on the individualdecision to participate in violence or onthe resolution of collective action problems within armed groups.

Fourth, at the macro level, we encour-age the development and synthesis of dataon additional forms of political instability and violence. Political repression gures

prominently in theories of conict and coop-eration (e.g., Acemoglu and Robinson 2006;Besley and Persson 2009, forthcoming),and yet we have a limited sense of its use oreffectiveness. As noted above, the distinctionbetween civil war and these other phenom-ena has been asserted rather than demon-strated. If we are interested in the strugglebetween groups for national power, it is notobvious that we should ignore coups; com-munal violence could similarly shed light onparticipation in violent collective action.62

Civil wars and conicts arguably inictmore suffering on humanity than any othersocial phenomenon. Now they are emerg-

ing as central to many countries’ politicalevolution and possibly as key impedimentsto global development. We hope this article will promote the incorporation of these top-ics into graduate and undergraduate coursesin both economics and political science, andstimulate further research. As we’ve discov-ered, and to recast Lucas’s famous phrase,once you start thinking about civil war, it’shard to think about anything else.

References

Abadie, Alberto, and Javier Gardeazabal. 2003. “TheEconomic Costs of Conict: A Case Study of theBasque Country.” American Economic Review,93(1): 113–32.

Acemoglu, Daron, Georgy Egorov, and KonstantinSonin. 2009. “Dynamics and Stability of Constitu-tions, Coalitions, and Clubs.” Unpublished.

Acemoglu, Daron, Simon Johnson, and James A.Robinson. 2001. “The Colonial Origins of Com-parative Development: An Empirical Investigation.” American Economic Review, 91(5): 1369–1401.

Acemoglu, Daron, and James A. Robinson. 2001. “A

Theory of Political Transitions.” American Eco- nomic Review, 91(4): 938–63.

Acemoglu, Daron, and James A. Robinson. 2006. Eco- nomic Origins of Dictatorship and Democracy.Cambridge and New York: Cambridge University Press.

62 New data are already emerging on leadership tran-sitions (Henk E. Goemans, Gleditsch, and GiacomoChiozza 2009) and on one-sided violence, nonstate armedconict, coups, and human rights abuses (Human Security Report Project 2008).




Alderman, Harold, John Hoddinott, and Bill Kinsey.2006. “Long Term Consequences of Early Child-hood Malnutrition.” Oxford Economic Papers, 58(3):450–74.

Alesina, Alberto, Reza Baqir, and William Easterly.1999. “Public Goods and Ethnic Divisions.” Quar-

terly Journal of Economics, 114(4): 1243–84.Alesina, Alberto, William Easterly, and Janina Matusz-eski. 2006. “Articial States.” National Bureau of Economic Research Working Paper 12328.

Alesina, Alberto, and Eliana La Ferrara. 2000. “Partic-ipation in Heterogeneous Communities.” Quarterly Journal of Economics, 115(3): 847–904.

Alesina, Alberto, and Eliana La Ferrara. 2005. “Eth-nic Diversity and Economic Performance.” Journalof Economic Literature, 43(3): 762–800.

Alesina, Alberto, Sule Özler, Nouriel Roubini, andPhillip Swagel. 1996. “Political Instability and Eco-nomic Growth.” Journal of Economic Growth, 1(2):189–211.

Alesina, Alberto, and Roberto Perotti. 1996. “IncomeDistribution, Political Instability, and Investment.”European Economic Review, 40(6): 1203–28.

André, Catherine, and Jean-Philippe Platteau. 1998.“Land Relations under Unbearable Stress: RwandaCaught in the Malthusian Trap.” Journal of Eco- nomic Behavior and Organization, 34(1): 1–47.

Angrist, Joshua D. 1990. “Lifetime Earnings and the Vietnam Era Draft Lottery: Evidence from SocialSecurity Administrative Records.” American Eco- nomic Review, 80(3): 313–36.

Angrist, Joshua D. 1998. “Estimating the Labor MarketImpact of Voluntary Military Service Using SocialSecurity Data on Military Applicants.” Economet-

rica, 66(2): 249–88.Angrist, Joshua D., and Alan B. Krueger. 1994. “Why Do World War II Veterans Earn More than Nonvet-erans?” Journal of Labor Economics, 12(1): 74–97.

Angrist, Joshua D., and Adriana D. Kugler. 2008.“Rural Windfall or a New Resource Curse? Coca,Income, and Civil Conict in Colombia.” Review of Economics and Statistics, 90(2): 191–215.

Annan, Jeannie, Christopher Blattman, Khristo-pher Carlson, and Dyan Mazurana. 2008. “Survey of War Affected Youth: Phase I & II Codebook.”Unpublished.

Annan, Jeannie, Christopher Blattman, and RogerHorton. 2006. “The State of Youth and Youth Pro-

tection in Northern Uganda: Findings from theSurvey for War Affected Youth.” Kampala, Uganda:UNICEF.

Annan, Jeannie, Christopher Blattman, DyanMazurana, and Khristopher Carlson. 2009. “Womenand Girls at War: ‘Wives’, Mothers, and Fighters inthe Lord’s Resistance Army.” Households in ConictNetwork Working Paper 63.

Arjona, Ana M., and Stathis N. Kalyvas. 2008. “Pre-liminary Results of a Survey of Demobilized Com-batants in Colombia.” Unpublished.

Arnson, Cynthia J., and I. William Zartman, eds. 2005.Rethinking the Economics of War: The Intersection

of Need, Creed, and Greed. Washington, D.C.: Woodrow Wilson Center Press; Baltimore: JohnsHopkins University Press.

Azam, Jean-Paul. 2002. “Looting and Conict betweenEthnoregional Groups: Lessons for State Formationin Africa.” Journal of Conict Resolution, 46(1):

131–53.Azam, Jean-Paul. 2006. “On Thugs and Heroes: Why Warlords Victimize Their Own Civilians.” Econom- ics of Governance, 7(1): 53–73.

Azariadis, Costas, and Allan Drazen. 1990. “ThresholdExternalities in Economic Development.” Quarterly Journal of Economics, 105(2): 501–26.

Balcells, Laia, and Stathis N. Kalyvas. 2007. “Warfarein Civil Wars.” Unpublished.

Baliga, Sandeep, and Tomas Sjostrom. 2004. “ArmsRaces and Negotiations.” Review of Economic Stud-

ies, 71(2): 351–69.Banerjee, Abhijit V., and Esther Duo. 2003. “Inequal-

ity and Growth: What Can the Data Say?” Journal of

Economic Growth, 8(3): 267–99.Bardhan, Pranab. 1997. “Method in the Madness? APolitical-Economy Analysis of the Ethnic Conictsin Less Developed Countries.” World Development,25(9): 1381–98.

Bardhan, Pranab. 2004. Scarcity, Conicts, and Coop-eration: Essays in the Political and InstitutionalEconomics of Development. Cambridge and Lon-don: MIT Press.

Barro, Robert J. 1991. “Economic Growth in a CrossSection of Countries.” Quarterly Journal of Eco- nomics, 106(2): 407–43.

Barro, Robert J., and Xavier Sala-i-Martin. 2003. Eco- nomic Growth, Second edition. Cambridge and

London: MIT Press.Barron, Patrick, Kai Kaiser, and Menno Pradhan.2004. “Local Conict in Indonesia: Measuring Inci-dence and Identifying Patterns.” World Bank Policy Research Working Paper 3384.

Bates, Robert H. 1986. “Modernization, Ethnic Com-petition, and the Rationality of Politics in Contem-porary Africa.” In Governing Black Africa, ed.Mario E. Doro and Newell M. Schultz. New Yorkand London: Africana Publishing Company.

Bates, Robert H. 2001. Prosperity and Violence: ThePolitical Economy of Development. New York andLondon: Norton.

Bates, Robert H. 2008. When Things Fell Apart: State

Failure in Late-Century Africa. Cambridge andNew York: Cambridge University Press.Bates, Robert H., Avner Greif, and Smita Singh. 2002.

“Organizing Violence.” Journal of Conict Resolu- tion, 46(5): 599–628.

Bazzi, Samuel, and Christopher Blattman. 2010.“Commodities and Conict.” Unpublished.

Beber, Bernd. 2008. “International Mediation of Mili-tarized Conicts: A New Dataset and Instrumental- Variable Approach.” Unpublished.

Beber, Bernd, and Christopher Blattman. 2008. “TheIndustrial Organization of Rebellion: The Logic of Forced Labor and Child Soldiering.” Unpublished.




Beck, Nathaniel, and Jonathan N. Katz. 2004.“Time-Series-Cross-Section Issues: Dynamics.”Unpublished.

Beck, Nathaniel, Jonathan N. Katz, and RichardTucker. 1998. “Taking Time Seriously: Time-Series-Cross Section Analysis with a Binary Dependent

Variable.” American Journal of Political Science,42(4): 1260–88.Bellows, John, and Edward Miguel. 2006. “War and

Institutions: New Evidence from Sierra Leone.” American Economic Review, 96(2): 394–99.

Bellows, John, and Edward Miguel. 2009. “War andLocal Collective Action in Sierra Leone.” Journal of Public Economics, 93(11–12): 144–57.

Benabou, Roland. 2000. “Unequal Societies: IncomeDistribution and the Social Contract.” AmericanEconomic Review, 90(1): 96–129.

Berman, Eli, and David D. Laitin. 2008. “Religion,Terrorism and Public Goods: Testing the ClubModel.” Journal of Public Economics, 92(10–11):

1942–67.Berman, Eli, Jacob N. Shapiro, and Joseph H. Fel-ter. 2008. “Can Hearts and Minds Be Bought? TheEconomics of Counterinsurgency in Iraq.” NationalBureau of Economic Research Working Paper14606.

Besancon, Marie L. 2005. “Relative Resources:Inequality in Ethnic Wars, Revolutions, and Geno-cides.” Journal of Peace Research, 42(4): 393–415.

Besley, Timothy J., and Torsten Persson. 2008a. “TheIncidence of Civil War: Theory and Evidence.”National Bureau of Economic Research WorkingPaper 14585.

Besley, Timothy J., and Torsten Persson. 2008b. “Wars

and State Capacity.” Journal of the European Eco- nomic Association, 6(2–3): 522–30.

Besley, Timothy J., and Torsten Persson. 2009.“Repression or Civil War?” American EconomicReview, 99(2): 292–97.

Besley, Timothy J., and Torsten Persson. Forthcoming.“The Logic of Political Violence.” Econometrica.

Biddle, Stephen, Wendy Brown, Jeffrey C. Isaac,Stathis N. Kalyvas, and Douglas A. Ollivant. 2008.“The New U.S. Army/Marine Corps Counterinsur-gency Field Manual as Political Science and PoliticalPraxis.” Perspectives on Politics, 6(2): 347–60.

Blattman, Christopher. 2009. “From Violence to Vot-ing: War and Political Participation in Uganda.”

American Political Science Review, 103(2): 231–47.Blattman, Christopher, and Jeannie Annan. Forth-coming. “The Consequences of Child Soldiering.”Review of Economics and Statistics.

Bloch, Francis, Sántiago Sánchez-Pagés, and RaphaelSoubeyran. 2006. “When Does Universal Peace Pre- vail? Secession and Group Formation in Conict.”Economics of Governance, 7(1): 3–29.

Bøås, Morten, and Kevin C. Dunn, eds. 2007. AfricanGuerrillas: Raging Against the Machine. Boulder:Lynne Rienner Publishers.

Bohlken, Anjali Thomas, and Ernest Sergenti. 2008.“Economic Growth and Ethnic Violence: An

Empirical Investigation of Hindu–Muslim Riots inIndia.” Unpublished.

Boix, Carles. 2003. Democracy and Redistribution.Cambridge; New York and Melbourne: CambridgeUniversity Press.

Brakman, Steven, Harry Garretsen, and Marc Sch-

ramm. 2004. “The Strategic Bombing of GermanCities During World War II and Its Impact on City Growth.” Journal of Economic Geography, 4(2):201–18.

Brito, Dagobert L., and Michael D. Intriligator. 1985.“Conict, War, and Redistribution.” American Polit- ical Science Review, 79(4): 943–57.

Bruck, Tilman. 1996. “The Economic Effects of War.”Unpublished.

Brückner, Markus, and Antonio Ciccone. 2007.“Growth, Democracy, and Civil War.” UniversitatPompeu Fabra Economics Working Paper 1053.

Bueno de Mesquita, Ethan. 2005. “The Quality of Ter-ror.” American Journal of Political Science, 49(3):

515–30.Bueno de Mesquita, Ethan. 2008. “Terrorist Fac-tions.” Quarterly Journal of Political Science, 3(4):399–418.

Buhaug, Halvard, and Scott Gates. 2002. “The Geog-raphy of Civil War.” Journal of Peace Research,39(4): 417–33.

Buhaug, Halvard, and Jan Ketil Rød. 2006. “LocalDeterminants of African Civil Wars, 1970–2001.”Political Geography, 25(3): 315–35.

Bundervoet, Tom, Philip Verwimp, and RichardAkresh. 2009. “Health and Civil War in RuralBurundi.” Journal of Human Resources, 44(2):536–63.

Burke, Marshall B., Edward Miguel, Shanker Satya-nath, John Dykema, and David B. Lobell. 2009.“Warming Increases the Risk of Civil War in Africa.”Proceedings of the National Academy of Sciences,106(49): 20670–74.

Carlsson, Hans, and Eric van Damme. 1993. “GlobalGames and Equilibrium Selection.” Econometrica,61(5): 989–1018.

Carmil, Devora, and Shlomo Breznitz. 1991. “PersonalTrauma and World View—Are Extremely StressfulExperiences Related to Political Attitudes, ReligiousBeliefs, and Future Orientation?” Journal of Trau-

matic Stress, 4(3): 393–405.Caselli, Francesco, and Wilbur John Coleman. 2006.

“On the Theory of Ethnic Conict.” National Bureauof Economic Research Working Paper 12125.Cederman, Lars-Erik, and Luc Girardin. 2007.

“Beyond Fractionalization: Mapping Ethnicity ontoNationalist Insurgencies.” American Political Sci-ence Review, 101(1): 173–85.

Cerra, Valerie, and Sweta Chaman Saxena. 2008.“Growth Dynamics: The Myth of Economic Recov-ery.” American Economic Review, 98(1): 439–57.

Charness, Gary, and Matthew Rabin. 2002. “Under-standing Social Preferences with Simple Tests.”Quarterly Journal of Economics, 117(3): 817–69.

Chassang, Sylvain, and Gerard Padro-i-Miquel.




2007. “Strategic Risk, Civil War and Intervention.”Unpublished.

Chassang, Sylvain, and Gerard Padro-i-Miquel. 2008a.“Conict and Deterrence under Strategic Risk.”National Bureau of Economic Research WorkingPaper 13964.

Chassang, Sylvain, and Gerard Padro-i-Miquel. 2008b.“Mutual Fear and Civil War.” Bureau for Researchand Economic Analysis of Development WorkingPaper 165.

Chassang, Sylvain, and Gerard Padro-i-Miquel. 2009.“Economic Shocks and Civil War.” Quarterly Jour- nal of Political Science, 4(3): 211–28.

Chen, Daniel L. 2007. “Islamic Resurgence and Social Violence during the Indonesian Financial Crisis.” InInstitutions and Norms in Economic Development,ed. Mark Gradstein and Kai A. Konrad, 179–200.Cambridge and London: MIT Press.

Choi, Jung-Kyoo, and Samuel Bowles. 2007. “TheCoevolution of Parochial Altruism and War.” Sci-

ence, 318(5850): 636– 40.Chwe, Michael Suk-Young. 1990. “Violence inIncentives: Pain in a Principal–Agent Model.”Northwestern University Center for MathematicalStudies in Economics and Management Science Dis-cussion Paper 871.

Ciccone, Antonio. 2008. “Transitory Economic Shocksand Civil Conict.” Unpublished.

Clapham, Christopher S., ed. 1998. African Guerril-las. Bloomington: Indiana University Press.

Clutterbuck, Richard L. 1966. The Long Long War:Counterinsurgency in Malaya and Vietnam. New York: Praeger.

Coghlan, Benjamin, Pascal Ngoy, Flavien Mulumba,

Colleen Hardy, Valerie Nkamgang Bemo, Tony Stewart, Jennifer Lewis, and Richard Brennan.2007. “Mortality in the Democratic Republic of Congo: An Ongoing Crisis.” http://www.theirc.org/ sites/default/les/migrated/resources/2007/2006-7_congomortalitysurvey.pdf.

Cohen, Dara Kay. 2008. “The Role of Female Combat-ants in Armed Groups: Women and Wartime Rapein Sierra Leone (1991–2002).” Unpublished.

Collier, Paul. 1999. “On the Economic Consequencesof Civil War.” Oxford Economic Papers, 51(1):168–83.

Collier, Paul. 2007. The Bottom Billion: Why the Poor-est Countries Are Failing and What Can Be Done

About It. Oxford and New York: Oxford University Press.Collier, Paul, Lisa Chauvet, Håvard Hegre. 2008. “The

Security Challenge in Conict-Prone Countries.”Copenhagen: Copenhagen Consensus Center.

Collier, Paul, V. L. Elliott, Håvard Hegre, Anke Hoef-er, Marta Reynal-Querol, and Nicholas Sambanis.2003. Breaking the Conict Trap: Civil War andDevelopment Policy. Washington, D.C.: World Bank;Oxford and New York: Oxford University Press.

Collier, Paul, and Anke Hoefer. 1998. “On EconomicCauses of Civil War.” Oxford Economic Papers,50(4): 563–73.

Collier, Paul, and Anke Hoefer. 2002. “Aid, Policy and Peace: Reducing the Risks of Civil Conict.”Defence and Peace Economics, 13(6): 435– 50.

Collier, Paul, and Anke Hoefer. 2004. “Greed andGrievance in Civil War.” Oxford Economic Papers,56(4): 563–95.

Collier, Paul, and Anke Hoefer. 2007. “Civil War.”In Handbook of Defense Economics, Volume 2,Defense in a Globalized World, ed. Todd Sandler,and Keith Hartley, 711–40. Amsterdam and Oxford:Elsevier, North-Holland.

Collier, Paul, Anke Hoefer, and Catherine Pattillo.2004. “Africa’s Exodus: Capital Flight and the BrainDrain as Portfolio Decisions.” Journal of AfricanEconomies, 13(Supplement 2): ii15–54.

Collier, Paul, Anke Hoefer, and Mans Soderbom.2004. “On the Duration of Civil War.” Journal of Peace Research, 41(3): 253–73.

Collier, Paul, and Nicholas Sambanis, eds. 2005a.Understanding Civil War: Evidence and Analysis,

Volume 1: Africa. Washington, D.C.: World Bank.Collier, Paul, and Nicholas Sambanis, eds. 2005b.Understanding Civil War: Evidence and Analysis,Volume 2: Europe, Central Asia, and Other Regions. Washington, D.C.: World Bank.

Collins, Randall. 2008. Violence: A Micro-sociologicalTheory. Princeton and Oxford: Princeton University Press.

Cramer, Christopher. 2002. “Homo Economicus Goesto War: Methodological Individualism, RationalChoice and the Political Economy of War.” WorldDevelopment, 30(11): 1845–64.

Cramer, Christopher. 2003. “Does Inequality CauseConict?” Journal of International Development,

15(4): 397–412.Cramer, Christopher. 2007. Violence in DevelopingCountries: War, Memory, Progress. Bloomington:Indiana University Press.

Cunningham, David E. 2006. “Veto Players and Civil War Duration.” American Journal of Political Sci-ence, 50(4): 875–92.

Dal Bó, Ernesto, and Pedro Dal Bó. 2004. “Workers, Warriors, and Criminals: Social Conict in GeneralEquilibrium.” Brown University Department of Economics Working Paper 2004-10.

Dal Bó, Ernesto, and Robert Powell. 2009. “A Modelof Spoils Politics.” American Journal of Political Sci-ence, 53(1): 207–22.

Davies, James C. 1962. “Toward a Theory of Revolu-tion.” American Sociological Review, 27(1): 5–19.Davis, Donald R., and David E. Weinstein. 2002.

“Bones, Bombs, and Break Points: The Geography of Economic Activity.” American Economic Review,92(5): 1269–89.

Deaton, Angus S., and Ronald I. Miller. 1995. Inter- national Commodity Prices, Macroeconomic Per- formance, and Politics in Sub-Saharan Africa.Princeton: Princeton University, Department of Economics, International Finance Section.

De Quervain, Dominique J.-F., Urs Fischbacher, Valerie Treyer, Melanie Schellhammer, Ulrich




Schnyder, Alfred Buck, and Ernst Fehr. 2004. “TheNeural Basis of Altruistic Punishment.” Science,305(5688): 1245–58.

De Ree, Joppe, and Eleonora Nillesen. 2006. “Aiding Violence or Peace? The Impact of Foreign Aid on theRisk of Civil Conict in Sub-Saharan Africa.” Utre-

cht School of Economics Discussion Paper 06-09.De Rouen, Karl R., Jr., and David Sobek. 2004. “TheDynamics of Civil War Duration and Outcome.” Journal of Peace Research, 41(3): 303–20.

Dhillon, Amrita, and Susana Peralta. 2002. “EconomicTheories of Voter Turnout.” Economic Journal,112(480): F332–52.

Do, Quy-Toan, and Lakshmi Iyer. 2007. “Poverty,Social Divisions and Conict in Nepal.” HarvardBusiness School Working Paper 07-065.

Doyle, Michael W., and Nicholas Sambanis. 2000.“International Peacebuilding: A Theoretical andQuantitative Analysis.” American Political ScienceReview, 94(4): 779–801.

Doyle, Michael W., and Nicholas Sambanis. 2006.Making War and Building Peace: United NationsPeace Operations. Princeton and Oxford: PrincetonUniversity Press

Driscoll, Jesse. 2008. “Inside the Leviathan: Coup-Proong after State Failure.” Unpublished.

Dube, Oeindrila, and Juan F. Vargas. 2008. “Commod-ity Price Shocks and Civil Conict: Evidence fromColombia.” Harvard University Center for Interna-tional Development Graduate Student and Postdoc-toral Fellow Working Paper 14.

Easterly, William. 2008. “Foreign Aid Goes Military!” New York Review of Books, 55(19).

Easterly, William, and Ross Levine. 1997. “Africa’s

Growth Tragedy: Policies and Ethnic Divisions.”Quarterly Journal of Economics, 112(4): 1203– 50.Eck, Kristine, and Lisa Hultman. 2007. “One-Sided

Violence against Civilians in War: Insights from New Fatality Data.” Journal of Peace Research, 44(2):233–46.

Ellman, Matthew, and Leonard Wantchekon. 2000.“Electoral Competition under the Threat of Politi-cal Unrest.” Quarterly Journal of Economics, 115(2):499–531.

Esteban, Joan, and Debraj Ray. 1994. “On the Mea-surement of Polarization.” Econometrica, 62(4):819–51.

Esteban, Joan, and Debraj Ray. 1999. “Conict and

Distribution.” Journal of Economic Theory, 87(2):379–415.Esteban, Joan, and Debraj Ray. 2001. “Social Decision

Rules Are Not Immune to Conict.” Economics of Governance, 2(1): 59–67.

Esteban, Joan, and Debraj Ray. 2008. “On the Salienceof Ethnic Conict.” American Economic Review,98(5): 2185–2202.

Etzioni, Amitai. 1975. A Comparative Analysis of Complex Organizations: On Power, Involvementand Their Correlates. New York: Free Press.

Fanon, Frantz. 1961. The Wretched of the Earth. New York: Grove Press, 2004.

Fearon, James D. 1995. “Rationalist Explanations for War.” International Organization, 49(3): 379–414.

Fearon, James D. 2003. “Ethnic and Cultural Diver-sity by Country.” Journal of Economic Growth, 8(2):195–222.

Fearon, James D. 2004. “Why Do Some Civil Wars

Last So Much Longer than Others?” Journal of Peace Research, 41(3): 275–301.Fearon, James D. 2006. “Ethnic Mobilization and Eth-

nic Violence.” In The Oxford Handbook of PoliticalEconomy, ed. Barry R. Weingast and Donald A. Wittman, 852– 68. Oxford and New York: OxfordUniversity Press.

Fearon, James D. 2007. “Economic Development,Insurgency, and Civil War.” In Institutions and Eco- nomic Performance, ed. Elhanan Helpman, 292–328. Cambridge and London: Harvard University Press.

Fearon, James D., Kimuli Kasara, and David D. Laitin.2007. “Ethnic Minority Rule and Civil War Onset.”

American Political Science Review, 101(1): 187–93.Fearon, James D., and David D. Laitin. 1996. “Explain-ing Interethnic Cooperation.” American PoliticalScience Review, 90(4): 715–35.

Fearon, James D., and David D. Laitin. 2003. “Ethnic-ity, Insurgency, and Civil War.” American PoliticalScience Review, 97(1): 75–90.

Fearon, James D., and David D. Laitin. 2005. “Ran-dom Narratives.” Unpublished.

Feddersen, Timothy J. 2004. “Rational Choice Theory and the Paradox of Not Voting.” Journal of EconomicPerspectives, 18(1): 99–112.

Fehr, Ernst, and Klaus M. Schmidt. 1999. “A Theory of Fairness, Competition, and Cooperation.” Quarterly

Journal of Economics, 114(3): 817–68.Ferguson, Niall. 2002. The Cash Nexus: Economicsand Politics from the Age of Warfare through the Age of Welfare, 1700–2000. New York: Basic Books.

Fogg, Piper. 2007. “A New Standard for MeasuringDoctoral Programs. Faculty Productivity Index:Overall Rankings.” Chronicle of Higher Education,53(19): A8.

Fortna, Virginia Page. 2004a. “Does PeacekeepingKeep Peace? International Intervention and theDuration of Peace after Civil War.” InternationalStudies Quarterly, 48(2): 269–92.

Fortna, Virginia Page. 2004b. Peace Time: Cease-Fire Agreements and the Durability of Peace. Princeton

and Oxford: Princeton University Press.Fortna, Virginia Page. 2008. Does PeacekeepingWork? Shaping Belligerents’ Choices after CivilWar . Princeton and Oxford: Princeton University Press.

Gabriel, Richard A., and Paul L. Savage. 1979. Crisis in Command: Mismanagement in the Army. New York: Hill and Wang.

Galula, David. 1964. Counterinsurgency Warfare:Theory and Practice. Westport, Conn.: Praeger.

Gambetta, Diego. 2005. “Can We Make Sense of Sui-cide Missions?” In Making Sense of Suicide Mis- sions, ed. Diego Gambetta, 259–300. Oxford and




New York: Oxford University Press.Garnkel, Michelle R. 1990. “Arming as a Strategic

Investment in a Cooperative Equilibrium.” Ameri-can Economic Review, 80(1): 50–68.

Garnkel, Michelle R. 2004. “On the Stability of GroupFormation: Managing the Conict Within.” Conict

Management and Peace Science, 21(1): 43–68.Garnkel, Michelle R., and Stergios Skaperdas. 2000.“Conict without Misperceptions or IncompleteInformation: How the Future Matters.” Journal of Conict Resolution, 44(6): 793–807.

Garnkel, Michelle R., and Stergios Skaperdas. 2007.“Economics of Conict: An Overview.” In Hand- book of Defense Economics, Volume 2, Defense ina Globalized World, ed. Todd Sandler, and KeithHartley, 649–710. Amsterdam and Oxford: Elsevier,North-Holland.

Gartzke, Erik. 1999. “War Is in the Error Term.” Inter- national Organization, 53(3): 567–87.

Gates, Scott. 2002. “Recruitment and Allegiance: The

Microfoundations of Rebellion.” Journal of ConictResolution, 46(1): 111–30.Gates, Scott. 2004. “Recruiting Child Soldiers.” Paper

presented at the Center for the Study of Civil War Workshop on Techniques of Violence in Civil War,Oslo.

Gates, Scott, and Håvard Strand. 2004. “Modeling theDuration of Civil Wars: Measurement and Estima-tion Issues.” Paper presented at the Joint Session of Workshops of the European Consortium of PoliticalResearch, Uppsala.

Gellner, Ernest. 1983. Nations and Nationalism.Ithaca and London: Cornell University Press.

Gersony, Robert. 1997. “The Anguish of Northern

Uganda: Results of a Field-Based Assessment of theCivil Conicts in Northern Uganda.” Report submit-ted to the United States Embassy, Kampala.

Ghobarah, Hazem Adam, Paul Huth, and Bruce Rus-sett. 2003. “Civil Wars Kill and Maim People—Longafter the Shooting Stops.” American Political Sci-ence Review, 97(2): 189–202.

Ghobarah, Hazem Adam, Paul Huth, and Bruce Rus-sett. 2004. “The Post-war Public Health Effects of Civil Conict.” Social Science and Medicine, 59(4):869–84.

Gilchrist, Simon, and John C. Williams. 2004. “Transi-tion Dynamics in Vintage Capital Models: Explain-ing the Postwar Catch-up of Germany and Japan.”

Federal Reserve Bank of San Francisco WorkingPaper 2004-14.Gilligan, James. 2000. “Violence in Public Health and

Preventive Medicine.” Lancet, 355(9217): 1802–04.Gilmore, Elisabeth, Nils Petter Gleditsch, Päivi Lujala,

and Jan Ketil Rød. 2005. “Conict Diamonds: ANew Dataset.” Conict Management and Peace Sci-ence, 22(3): 257–72.

Gleditsch, Kristian Skrede. 2007. “TransnationalDimensions of Civil War.” Journal of Peace Research,44(3): 293–309.

Gleditsch, Nils Petter. 1998. “Armed Conict and theEnvironment: A Critique of the Literature.” Journal

of Peace Research, 35(3): 381–400.Gleditsch, Nils Petter, Peter Wallensteen, Mikael

Eriksson, Margareta Sollenberg, and Håvard Strand.2002. “Armed Conict 1946–2001: A New Dataset.” Journal of Peace Research, 39(5): 615–37.

Goemans, Henk E., Kristian Skrede Gleditsch, and

Giacomo Chiozza. 2009. “Introducing Archigos:A Dataset of Political Leaders.” Journal of PeaceResearch, 46(2): 269–83.

Goldstone, Jack A., Robert H. Bates, David L. Epstein,Ted Robert Gurr, Michael Lustik, Monty G. Mar-shall, Jay Ulfelder, and Mark Woodward. 2010. “AGlobal Model for Forecasting Political Instabil-ity.” American Journal of Political Science, 54(1):190–208.

Goodwin, Jeffrey. 2006. “A Theory of Categorical Ter-rorism.” Social Forces, 84(4): 2027–46.

Gramsci, Antonio. 1971. Selections from the Prison Notebooks. London: Lawrence and Wishart.

Grossman, Herschel I. 1991. “A General Equilib-

rium Model of Insurrections.” American EconomicReview, 81(4): 912–21.Grossman, Herschel I. 1999. “Kleptocracy and

Revolutions.” Oxford Economic Papers, 51(2):267–83.

Guidolin, Massimo, and Eliana La Ferrara. 2007.“Diamonds Are Forever, Wars Are Not: Is Con-ict Bad for Private Firms?” American EconomicReview, 97(5): 1978–93.

Gurr, Ted Robert. 1971. Why Men Rebel. Princetonand Oxford: Princeton University Press.

Gurr, Ted Robert, and Will H. Moore. 1997. “Ethno-political Rebellion: A Cross-Sectional Analysis of the1980s with Risk Assessments for the 1990s.” Ameri-

can Journal of Political Science, 41(4): 1079–1103.Gutiérrez Sanín, Francisco. 2008. “Telling the Differ-ence: Guerrillas and Paramilitaries in the Colom-bian War.” Politics and Society, 36(1): 3–34.

Gyimah-Brempong, Kwabena, and Thomas L. Traynor.1999. “Political Instability, Investment and Eco-nomic Growth in Sub-Saharan Africa.” Journal of African Economies, 8(1): 52–86.

Haavelmo, Trygve. 1954. A Study in the Theory of Economic Evolution. Amsterdam: North-Holland.

Habyarimana, James, Macartan Humphreys, DanielN. Posner, and Jeremy M. Weinstein. 2007. “Why Does Ethnic Diversity Undermine Public GoodsProvision?” American Political Science Review,

101(4): 709 –25.Habyarimana, James, Macartan Humphreys, DanielN. Posner, and Jeremy M. Weinstein. Forthcoming.The Co-Ethnic Advantage: Diversity and theImpediments to Collective Action. New York: Rus-sell Sage Foundation.

Hamilton, Kirk, and Michael Clemens. 1999. “Genu-ine Savings Rates in Developing Countries.” WorldBank Economic Review, 13(2): 333– 56.

Hanson, Matthew A. 2007. “The Economics of Road-side Bombs.” College of William & Mary Depart-ment of Economics Working Paper 68.

Harbom, Lotta, and Peter Wallensteen. 2007. “Armed




Conict, 1989–2006.” Journal of Peace Research,44(5): 623–34.

Hassner, Ron E. 2003. “‘To Halve and to Hold’: Con-icts over Sacred Space and the Problem of Indivis-ibility.” Security Studies, 12(4): 1–33.

Hearst, Norman, Thomas B. Newman, and Stephen B.

Hulley. 1986. “Delayed Effects of the Military Drafton Mortality: A Randomized Natural Experiment.” New England Journal of Medicine, 314(10): 620–24.

Hegre, Håvard, Tanja Ellingsen, Scott Gates, and NilsPetter Gleditsch. 2001. “Toward a Democratic CivilPeace? Democracy, Political Change, and Civil War,1816–1992.” American Political Science Review,95(1): 33–48.

Hegre, Håvard, and Nicholas Sambanis. 2006. “Sen-sitivity Analysis of Empirical Results on Civil WarOnset.” Journal of Conict Resolution, 50(4):508–35.

Hendrix, Cullen S., and Sarah M. Glaser. 2007.“Trends and Triggers: Climate, Climate Change

and Civil Conict in Sub-Saharan Africa.” PoliticalGeography, 26(6): 695–715.Herbst, Jeffrey. 2000. States and Power in Africa:

Comparative Lessons in Authority and Control.Princeton: Princeton University Press.

Hirshleifer, Jack. 1988. “The Analytics of ContinuingConict.” Synthese, 76(2): 201–33.

Hirshleifer, Jack. 1989. “Conict and Rent-SeekingSuccess Functions: Ratio vs. Difference Models of Relative Success.” Public Choice, 63(2): 101–12.

Homer-Dixon, Thomas F. 1999. Environment, Scar-city, and Violence. Princeton and Oxford: PrincetonUniversity Press.

Horowitz, Donald L. 1985. Ethnic Groups in Conict.

Berkeley and London: University of California Press.Human Security Report Project. 2008. “HumanSecurity Brief 2007.” Vancouver: Human Security Research Group.

Humphreys, Macartan. 2003. “Economics and Vio-lent Conict.” Program on Humanitarian Policy andConict Research Working Paper.

Humphreys, Macartan. 2005. “Natural Resources,Conict, and Conict Resolution: Uncovering theMechanisms.” Journal of Conict Resolution, 49(4):508–37.

Humphreys, Macartan, and Jeremy M. Weinstein.2004. “What the Fighters Say: A Survey of Ex-com-batants in Sierra Leone June–August 2003.” Cen-

ter on Globalization and Sustainable Development Working Paper 20.Humphreys, Macartan, and Jeremy M. Weinstein.

2006. “Handling and Manhandling Civilians in Civil War: Determinants of the Strategies of Warring Fac-tions.” American Political Science Review, 100(3):429–47.

Humphreys, Macartan, and Jeremy M. Weinstein.2007. “Demobilization and Reintegration.” Journalof Conict Resolution, 51(4): 531–67.

Humphreys, Macartan, and Jeremy M. Weinstein.2008. “Who Fights? The Determinants of Partici-pation in Civil War.” American Journal of Political

Science, 52(2): 436–55.Imbens, Guido, and Wilbert van der Klaauw. 1995.

“Evaluating the Cost of Conscription in the Nether-lands.” Journal of Business and Economic Statistics,13(2): 207–15.

Iyengar, Radha, and Jonathan Monten. 2008. “Is There

an ‘Emboldenment’ Effect? Evidence from theInsurgency in Iraq.” National Bureau of EconomicResearch Working Paper 13839.

Jackson, Matthew O., and Massimo Morelli. 2007.“Political Bias and War.” American EconomicReview, 97(4): 1353–73.

Jervis, Robert. 1976. Perception and Misperception inInternational Politics. Princeton and Oxford: Princ-eton University Press.

Jones, Benjamin F., and Benjamin A. Olken. 2009. “Hitor Miss? The Effect of Assassinations on Institutionsand War.” American Economic Journal: Macroeco- nomics, 1(2): 55–87.

Justino, Patricia. 2007. “On the Links between Violent

Conict and Household Poverty: How Much Do WeReally Know?” Micro Level Analysis of Violent Con-ict Working Paper 1.

Justino, Patricia. 2009. “Poverty and Violent Conict:A Micro-level Perspective on the Causes and Dura-tion of Warfare.” Journal of Peace Research, 46(3):315–33.

Justino, Patricia, and Philip Verwimp. 2006. “Poverty Dynamics, Violent Conict and Convergence inRwanda.” Households in Conict Network WorkingPaper 16.

Kahl, Colin H. 2006. States, Scarcity, and Civil Strife in the Developing World. Princeton and Oxford:Princeton Universtiy Press.

Kaldor, Mary. 1999. New and Old Wars: OrganizedViolence in a Global Era. Cambridge: Polity Press.

Kalyvas, Stathis N. 2000. “Commitment Problems inEmerging Democracies: The Case of Religious Par-ties.” Comparative Politics, 32(4): 379–98.

Kalyvas, Stathis N. 2005. “Warfare in Civil Wars.”In Rethinking the Nature of War , ed. IsabelleDuyvesteyn and Jan Angstrom, 88–108. London andNew York: Taylor and Francis, Frank Cass.

Kalyvas, Stathis N. 2006. The Logic of Violence inCivil War . Cambridge and New York: CambridgeUniversity Press.

Kalyvas, Stathis N. 2007. “Civil Wars.” In The OxfordHandbook of Comparative Politics, ed. Carles Boix

and Susan C. Stokes, 416–34. Oxford and New York:Oxford University Press.Kalyvas, Stathis N., and Matthew Kocher. 2007. “How

Free is ‘Free Riding’ in Civil Wars? Violence, Insur-gency, and the Collective Action Problem.” WorldPolitics, 59(2): 177–216.

Kaplan, Robert D. 1994. “The Coming Anarchy.” Atlantic Monthly: 44–76.

Keen, David. 2005. Conict and Collusion in Sierra Leone. New York: Palgrave.

Kennan, John, and Robert Wilson. 1993. “Bargaining with Private Information.” Journal of Economic Lit-erature, 31(1): 45–104.




Kenny, Paul D. 2008. “Discipline, Identity and Cohe-sion in Armed Organizations.” Unpublished.

Kevane, Michael, and Leslie Gray. 2008. “Darfur:Rainfall and Conict.” Unpublished.

King, Gary, and Langche Zeng. 2001. “LogisticRegression in Rare Events Data.” Political Analysis,

9(2): 137–63.Knack, Stephen, and Philip Keefer. 1997. “Does SocialCapital Have an Economic Payoff? A Cross-Country Investigation.” Quarterly Journal of Economics,112(4): 1251–88.

Kondylis, Florence. 2008a. “Agricultural Outputsand Conict Displacement: Evidence from a Policy Intervention in Rwanda.” Economic Developmentand Cultural Change, 57(1): 31–66.

Kondylis, Florence. 2008b. “Conict Displacementand Labor Market Outcomes in Post-war Bosniaand Herzegovina.” Households in Conict Network Working Paper 45.

Krueger, Alan B., and Jitka Maleckova. 2003. “Educa-

tion, Poverty and Terrorism: Is There a Causal Con-nection?” Journal of Economic Perspectives, 17(4):119–44.

La Ferrara, Eliana, and Robert H. Bates. 2001. “Politi-cal Competition in Weak States.” Economics andPolitics, 13(2): 159–84.

Laitin, David D. 2007. Nations, States, and Violence.Oxford and New York: Oxford University Press.

Le Billon, Philippe. 2001. “The Political Ecology of War: Natural Resources and Armed Conicts.”Political Geography, 20(5): 561–84.

Le Billon, Philippe. 2005. Fuelling War: Natu-ral Resources and Armed Conict. New York:Routledge.

Leonard, David K., and Scott Straus. 2003. Africa’sStalled Development: International Causes andCures. London: Lynne Rienner Publishers.

Levine, Ross, and David Renelt. 1992. “A Sensitivity Analysis of Cross-Country Growth Regressions.” American Economic Review, 82(4): 942–63.

Levy, Jack S. 1997. “Prospect Theory, Rational Choice,and International Relations.” International StudiesQuarterly, 41(1): 87–112.

Lichbach, Mark Irving. 1994. “What Makes RationalPeasants Revolutionary? Dilemma, Paradox, andIrony in Peasant Collective Action.” World Politics,46(3): 383–418.

Lichbach, Mark Irving. 1995. The Rebel’s Dilemma.

Ann Arbor: University of Michigan Press.Licklider, Roy, ed. 1993. Stopping the Killing: HowCivil Wars End. New York and London: New YorkUniversity Press.

Loayza, Norman, Malcolm Knight, and Delano Villan-ueva. 1999. “The Peace Dividend: Military Spend-ing Cuts and Economic Growth.” World Bank Policy Research Working Paper 1577.

Lobell, David B., Marshall B. Burke, Claudia Tebaldi,Michael D. Mastrandrea, Walter P. Falcon, andRosamond L. Naylor. 2008. “Prioritizing ClimateChange Adaption Needs for Food Security in 2030.”Science, 319(5863): 607–10.

Lucas, Robert E., Jr. 1988. “On the Mechanics of Eco-nomic Development.” Journal of Monetary Econom- ics, 22(1): 3–42.

Luttwak, Edward N. 1999. “Give War a Chance.” For-eign Affairs, 78(4): 36–44.

Lyall, Jason. 2009a. “Does Indiscriminate Violence

Incite Insurgent Attacks?: Evidence from Chech-nya.” Journal of Conict Resolution, 53(3): 331–62.Lyall, Jason. 2009b. “How Ethnicity Shapes Insurgent

Violence: A Matched Analysis of ‘Sweep’ Operationsin Chechnya.” Unpublished.

Macours, Karen. 2008. “Increasing Inequality andCivil Conict in Nepal.” Unpublished.

Mankiw, N. Gregory, David Romer, and David N. Weil. 1992. “A Contribution to the Empirics of Eco-nomic Growth.” Quarterly Journal of Economics,107(2): 407–37.

Marshall, Monty G., and Keith Jaggers. 2006. “Pol-ity IV Project: Political Regime Characteristics andTransitions, 1800–2004.” Severn, Md.: Center for

Systemic Peace.Mason. T. David, and Dale A. Krane. 1989. “The Polit-ical Economy of Death Squads: Toward a Theory of the Impact of State-Sanctioned Terror.” Interna- tional Studies Quarterly, 33(2): 175–98.

Masten, Ann S. 2001. “Ordinary Magic: ResilienceProcesses in Development.” American Psychologist,56(3): 227–38.

McBride, Michael, and Stergios Skaperdas. 2007.“Explaining Conict in Low-Income Countries:Incomplete Contracting in the Shadow of theFuture.” In Institutions and Norms in EconomicDevelopment, ed. Mark Gradstein and Kai A. Kon-rad, 141–62. Cambridge and London: MIT Press.

McMaster, H. R. 2008. “On War: Lessons to BeLearned.” Survival, 50(1): 19–30.Miguel, Edward. 2007. “Poverty and Violence: An

Overview of Recent Research and Implications forForeign Aid.” In Too Poor for Peace? Global Poverty,Conict, and Security in the 21st Century, ed. LaelBrainard and Derek Chollet, 50–59. Washington,D.C.: Brookings Institution Press.

Miguel, Edward, and Mary Kay Gugerty. 2005. “Eth-nic Diversity, Social Sanctions, and Public Goodsin Kenya.” Journal of Public Economics, 89(11–12):2325–68.

Miguel, Edward, and Gerald Roland. 2006. “The LongRun Impact of Bombing Vietnam.” National Bureau

of Economic Research Working Paper 11954.Miguel, Edward, Sebastian M. Saiegh, and ShankerSatyanath. 2008. “National Cultures and Soccer Violence.” National Bureau of Economic Research Working Paper 13968.

Miguel, Edward, Shanker Satyanath, and Ernest Ser-genti. 2004. “Economic Shocks and Civil Conict:An Instrumental Variables Approach.” Journal of Political Economy, 112(4): 725– 53.

Mkandawire, Thandika. 2002. “The Terrible Toll of Post-Colonial ‘Rebel Movements’ in Africa: Towardsan Explanation of the Violence against the Peas-antry.” Journal of Modern African Studies, 40(2):




181–215.Montalvo, Jose G., and Marta Reynal-Querol. 2005.

“Ethnic Polarization, Potential Conict, and Civil Wars.” American Economic Review, 95(3): 796–816.

Montalvo, Jose G., and Marta Reynal-Querol. 2007.“Fighting against Malaria: Prevent Wars While

Waiting for the ‘Miraculous’ Vaccine.” Review of Economics and Statistics, 89(1): 165–77.Moore, Barrington. 1993. Social Origins of Dictator-

ship and Democracy: Lord and Peasant in the Mak- ing of the Modern World. Boston: Beacon Press.

Moore, Mick, and Vishal Jadhav. 2006. “The Politicsand Bureaucratics of Rural Public Works: Maharash-tra’s Employment Guaranteed Scheme.” Journal of Development Studies, 42(8): 1271–1300.

Morris, Stephen, and Hyun Song Shin. 1998. “UniqueEquilibrium in a Model of Self-Fullling Cur-rency Attacks.” American Economic Review, 88(3):587–97.

Mueller, John. 2000. “The Banality of ‘Ethnic Con-

ict’.” International Security, 25(1): 42–70.Murdoch, James C., and Todd Sandler. 2004. “Civil Wars and Economic Growth: Spatial Dispersion.” American Journal of Political Science, 48(1): 138– 51.

Murshed, S. Mansoob, and Scott Gates. 2005. “Spatial-Horizontal Inequality and the Maoist Insurgency inNepal.” Review of Development Economics, 9(1):121–34.

Mvukiyehe, Eric, Cyrus Samii, and Gwendolyn Tay-lor. 2008. “Wartime and Postconict Experiences inBurundi.” Unpublished.

Nagl, John A. 2002. Counterinsurgency Lessons fromMalaya and Vietnam: Learning to Eat Soup with aKnife. Westport, Conn. and London: Praeger.

National Research Council. 1995. “Research Doctor-ate Programs in the United States: Continuity andChange.” Washington, D.C.: National AcademiesPress.

Olson, Mancur. 1971. The Logic of Collective Action:Public Groups and the Theory of Groups. Cam-bridge and London: Harvard University Press.

Organski, A. F. K., and Jacek Kugler. 1977. “The Costsof Major Wars: The Phoenix Factor.” American Polit- ical Science Review, 71(4): 1347–66.

Organski, A. F. K., and Jacek Kugler. 1980. The War Ledger . Chicago and London: University of ChicagoPress.

Østby, Gudrun. 2005. “Horizontal Inequalities and

Civil Conict.” Paper presented at the AnnualMeeting of the International Studies Association,Honolulu.

Ostrom, Elinor. 1990. Governing the Commons: TheEvolution of Institutions for Collective Action.Cambridge; New York and Melbourne: CambridgeUniversity Press.

Paige, Jeffery M. 1975. Agrarian Revolution: SocialMovements and Export Agriculture in the Under-developed World. New York: Macmillan, Free Press.

Petersen, Roger D. 2001. Resistance and Rebellion: Lessons from Eastern Europe. Cambridge and New York: Cambridge University Press.

Petraeus, David H. 2006. “Learning Counterinsur-gency: Observations from Soldiering in Iraq.” Mili- tary Review, Jan–Feb: 45–55.

Popkin, Samuel L. 1979. The Rational Peasant: ThePolitical Economy of Rural Society in Vietnam.Berkeley and London: University of California Press.

Popkin, Samuel L. 1988. “Political Entrepreneurs andPeasant Movements in Vietnam.” In Rationality andRevolution, ed. Michael Taylor, 9–62. Cambridgeand New York: Cambridge University Press.

Posen, Barry R. 1993. “The Security Dilemma andEthnic Conict.” Survival, 35(1): 27–47.

Posner, Daniel N. 2004a. “Measuring Ethnic Fraction-alization in Africa.” American Journal of PoliticalScience, 48(4): 849–63.

Posner, Daniel N. 2004b. “The Political Salience of Cultural Difference: Why Chewas and TumbukasAre Allies in Zambia and Adversaries in Malawi.” American Political Science Review, 98(4): 529–45.

Powell, Robert. 2002. “Bargaining Theory and Inter-

national Conict.” Annual Review of Political Sci-ence, 5: 1–30.Powell, Robert. 2006. “War as a Commitment

Problem.” International Organization, 60(1):169–203.

Powell, Robert. 2007. “Allocating Defensive Resources with Private Information About Vulnerability.” American Political Science Review, 101(4): 799–809.

Princeton University. 2009. “Shapiro Awarded $8.4Million to Examine Development Aid’s Impact onTerrorism.” http://wws-edit.princeton.edu/news/ shapiro_DoDgrant/index.xml.

Przeworski, Adam, Michael E. Alvarez, Jose AntonioCheibub, and Fernando Limongi. 2000. Democracy

and Development: Political Institutions and Well-Being in the World, 1950–1990. Cambridge andNew York: Cambridge University Press.

Pugel, James. 2007. “What the Fighters Say: A Survey of Ex-combatants in Liberia.” Monrovia: UnitedNations Development Programme Liberia.

Punamäki, Raija-Leena, Samir Qouta, and Eyad ElSarraj. 1997. “Relationships between TraumaticEvents, Children’s Gender, and Political Activity,and Perceptions of Parenting Styles.” International Journal of Behavioral Development, 21(1): 91–109.

Ray, Debraj. 1998. Development Economics. Princ-eton and Oxford: Princeton University Press.

Ray, Debraj. 2007. A Game-Theoretic Perspective on

Coalition Formation. Oxford and New York: OxfordUniversity Press.Ray, Debraj. 2009. “Remarks on the Initiation of

Costly Conict.” Presented at the Yale University Workshop on Rationality and Conict, New Haven.

Reno, William. 1999. Warlord Politics and AfricanStates. Boulder: Lynne Rienner Publishers.

Reynal-Querol, Marta, and Jose G. Montalvo. 2007.“Ethnic Polarization and the Duration of Civil Wars.” World Bank Policy Research Working Paper 4192.

Rodrik, Dani. 1999. “Where Did All the Growth Go?External Shocks, Social Conict, and Growth Col-lapses.” Journal of Economic Growth, 4(4): 385– 412.




Roemer, John E. 1985. “Rationalizing Revolutionary Ideology.” Econometrica, 53(1): 85–108.

Ross, Michael L. 2004a. “How Do Natural ResourcesInuence Civil War? Evidence from ThirteenCases.” International Organization, 58(1): 35–67.

Ross, Michael L. 2004b. “What Do We Know about

Natural Resources and Civil War?” Journal of PeaceResearch, 41(3): 337–56.Ross, Michael L. 2006. “A Closer Look at Oil, Dia-

monds, and Civil War.” Annual Review of PoliticalScience, 9: 265–300.

Rush, Robert Sterling. 1999. “A Different Perspective:Cohesion, Morale, and Operational Effectiveness inthe German Army, Fall 1944.” Armed Forces andSociety, 25(3): 477–508.

Rush, Robert Sterling. 2001. Hell in Hürtgen Forest:The Ordeal and Triumph of an American InfantryRegiment. Lawrence: University Press of Kansas.

Sachs, Jeffrey D. 2005. The End of Poverty: EconomicPossibilities for Our Time. New York and Toronto:

Penguin Press.Sala-i-Martin, Xavier. 1997. “I Just Ran Two MillionRegressions.” American Economic Review, 87(2):178–83.

Salehyan, Idean, and Kristian Skrede Gleditsch. 2006.“Refugees and the Spread of Civil War.” Interna- tional Organization, 60(2): 335–66.

Sambanis, Nicholas. 2001. “Do Ethnic and NonethnicCivil Wars Have the Same Causes?: A Theoreticaland Empirical Inquiry (Part 1).” Journal of ConictResolution, 45(3): 259–82.

Sambanis, Nicholas. 2002. “A Review of RecentAdvances and Future Directions in the QuantitativeLiterature on Civil War.” Defence and Peace Eco-

nomics, 13(3): 215–43.Sambanis, Nicholas. 2004. “What Is Civil War? Con-ceptual and Empirical Complexities of an Opera-tional Denition.” Journal of Conict Resolution,48(6): 814–58.

Sambanis, Nicholas. 2005. “Using Case Studies toRene and Expand the Theory of Civil War.” InUnderstanding Civil War: Evidence and Analysis,Volume 1: Africa, ed. Paul Collier and Nicholas Sam-banis. Washington, D.C.: World Bank.

Sambanis, Nicholas. 2007. “Short-Term and Long-Term Effects of United Nations Peace Operations.” World Bank Policy Research Working Paper 4207.

Sambanis, Nicholas, and Ibrahim A. Elbadawi. 2000.

“External Interventions and the Duration of Civil Wars.” World Bank Policy Research WorkingPaper 2433.

Sandler, Todd, and Walter Enders. 2004. “An EconomicPerspective on Transnational Terrorism.” European Journal of Political Economy, 20(2): 301–16.

Sawyer, Amos. 2004. “Violent Conicts and Gover-nance Challenges in West Africa: The Case of theMano River Basin Area.” Journal of Modern AfricanStudies, 42(3): 437–63.

Sawyer, Amos. 2005. Beyond Plunder: Toward Dem-ocratic Governance in Liberia. Boulder: LynneRienner Publishers.

Schneider, Gerald, and Nina Wiesehomeier. 2007.“Ethnic Polarization, Potential Conict, and Civil Wars: Comment.” Unpublished.

Scott, James C. 1976. Moral Economy of the Peasant:Rebellion and Subsistence in Southeast Asia. New Haven and London: Yale University Press.

Sepp, Kalev I. 2005. “Best Practices in Counterinsur-gency.” Military Review, May–June: 8–12.Shapiro, Jacob N. 2008. “Bureaucracy and Control In

Terrorist Organizations.” Unpublished.Shemyakina, Olga. 2006. “The Effect of Armed Con-

ict on Accumulation of Schooling: Results fromTajikistan.” Households in Conict Network Work-ing Paper 12.

Shils, Edward A., and Morris Janowitz. 1948. “Cohe-sion and Disintegration in the Wehrmacht in World War II.” Public Opinion Quarterly, 12(2): 280–315.

Singer, J. David, and Melvin Small. 1994. “Correlatesof War Project: International and Civil War Data,1816–1992.” Inter-university Consortium for Politi-

cal and Social Research Study 9905.Skaperdas, Stergios. 1992. “Cooperation, Conict, andPower in the Absence of Property Rights.” AmericanEconomic Review, 82(4): 720–39.

Skaperdas, Stergios. 2008. “An Economic Approach toAnalyzing Civil Wars.” Economics of Governance,9(1): 25–44.

Skaperdas, Stergios, and Constantinos Syropoulos.1996. “Can the Shadow of the Future Harm Coop-eration?” Journal of Economic Behavior and Orga- nization, 29(3): 355–72.

Slater, Dan. 2005. “Ordering Power: Contentious Poli-tics, State-Building, and Authoritarian Durability inSoutheast Asia.” Unpublished.

Smith, L. Alison, Catherine Gambette, and ThomasLongley. 2004. “Conict Mapping in Sierra Leone: Violations of International Humanitarian Law from1991 to 2002.” New York: No Peace Without Justice.

Stedman, Stephen John. 1997. “Spoiler Problems inPeace Processes.” International Security, 22(2):5–53.

Steele, Abbey. 2007. “Massive Civilian Displacementin Civil War: Assessing Variation in Colombia.”Households in Conict Network Working Paper 29.

Stewart, Frances. 2001. Horizontal Inequalities: A Neglected Dimension of Development. Helsinki:United Nations University World Institute for Devel-opment Economics Research.

Straus, Scott. 2006. The Order of Genocide: Race,Power, and War in Rwanda. Ithaca and London:Cornell University Press.

Straus, Scott. 2007. “Perpetrators in the Study of Civil War and Violence.” Unpublished.

Svensson, Jakob. 1998. “Investment, Property Rightsand Political Instability: Theory and Evidence.”European Economic Review, 42(7): 1317–41.

Tilly, Charles, ed. 1975. The Formation of NationalStates in Western Europe. Princeton and Oxford:Princeton University Press.

Tilly, Charles. 1982. “Warmaking and Statemaking asOrganized Crime.” University of Michigan Center




for Research on Social Organizations Working Paper256.

Tilly, Charles. 1992. Coercion, Capital, and EuropeanStates, AD 990–1992. Malden, Mass. and Oxford:Blackwell.

Tirole, Jean. 1988. The Theory of Industrial Organiza-

tion. Cambridge, Mass. and London: MIT Press.Tirole, Jean. 2006. The Theory of Corporate Finance.Princeton and Oxford: Princeton University Press.

Todaro, Michael. 1999. Economic Development, Sev-enth edition. Reading, Mass.: Addison Wesley.

Toft, Monica Duffy. 2008. “Peace through Victory.”Unpublished.

UCDP. 2008. “Uppsala Conict Data Program.”Uppsala University.

U.S. News and World Report. 2007a. “America’s BestColleges.”

U.S. News and World Report. 2007b. “America’s BestGraduate Schools: Economics Ph.D.”

Valentine, Theodore R. 1993. “Drought, Transfer Enti-

tlements, and Income Distribution: The BotswanaExperience.” World Development, 21(1): 109–26. Van Creveld. Martin. 1982. Fighting Power: German

and U.S. Army Performance, 1939–1945. London:Greenwood Press.

Verwimp, Philip. 2003. “Testing the Double-GenocideThesis for Central and Southern Rwanda.” Journalof Conict Resolution, 47(4): 423–42.

Verwimp, Philip. 2005. “An Economic Prole of Peas-ant Perpetrators of Genocide: Micro-level Evidencefrom Rwanda.” Journal of Development Economics,77(2): 297–323.

Vreeland, James Raymond. 2008. “The Effect of Politi-cal Regime on Civil War: Unpacking Anocracy.”

Journal of Conict Resolution, 52(3): 401–25. Walter, Barbara F. 1997. “The Critical Barrier to Civ il War Settlement.” International Organization, 51(3):335–64.

Walter, Barbara F. 2004. “Does Conict Beget Con-ict? Explaining Recurring Civil War.” Journal of Peace Research, 41(3): 371–88.

Walter, Barbara F. 2006. “Information, Uncertainty,and the Decision to Secede.” International Organi- zation, 60(1): 105–35.

Walter, Barbara F., and Jack Snyder, eds. 1999. CivilWars, Insecurity, and Intervention. New York:

Columbia University Press. Weber, Max. 1965. Politics as a Vocation. Minneapolis:

Fortress Press. Weingast, Barry R. 1997. “The Political Foundations

on Democracy and the Rule of Law.” AmericanPolitical Science Review, 91(2): 245–63.

Weinstein, Jeremy M. 2005a. “Autonomous Recovery and International Intervention in Comparative Per-spective.” Center for Global Development WorkingPaper 57.

Weinstein, Jeremy M. 2005b. “Resources and theInformation Problem in Rebel Recruitment.” Jour- nal of Conict Resolution, 49(4): 598–624.

Weinstein, Jeremy M. 2007. Inside Rebellion: ThePolitics of Insurgent Violence. Cambridge and New York: Cambridge University Press.

Wimmer, Andreas, and Brian Min. 2006. “FromEmpire to Nation-State: Explaining Wars in theModern World, 1816–2001.” American SociologicalReview, 71(6): 867–97.

Wood, Elisabeth Jean. 2003a. “Civil Wars: What WeDon’t Know.” Global Governance, 9(2): 247–61. Wood, Elisabeth Jean. 2003b. Insurgent Collective

Action and Civil War in El Salvador . Cambridgeand New York: Cambridge University Press.

Wood, Elisabeth Jean. 2006. “Variation in Sexual Violence during War.” Politics and Society, 34(3):307–42.

Wood, Elisabeth Jean. 2008. “The Social Processes of Civil War: The Wartime Transformation of SocialNetworks.” Annual Review of Political Science, 11:539–61.

Wood, Elizabeth Jean. 2009. “Armed Groups and Sex-ual Violence: When Is Wartime Rape Rare?” Politics

and Society, 37(1): 131–61. World Bank. 2008. World Development Indicators 2008. Washington, D.C.: World Bank.

Young, Alwyn. 2005. “The Gift of the Dying: The Trag-edy of AIDS and the Welfare of Future African Gen-erations.” Quarterly Journal of Economics, 120(2):423–66.

Zinn, Annalisa. 2005. “Theory versus Reality: Civil War Onset and Avoidance in Nigeria since 1960.” InUnderstanding Civil War: Evidence and Analysis:Volume 1: Africa, ed. Paul Collier and Nicholas Sam-banis, 89–121. Washington, D.C.: World Bank.



This article has been cited by:

1. M. Shayo, A. Zussman. 2011. Judicial Ingroup Bias in the Shadow of Terrorism. The Quarterly

Journal of Economics . [CrossRef ]

2. LARS-ERIK CEDERMAN, NILS B. WEIDMANN, KRISTIAN SKREDE GLEDITSCH. 2011.

Horizontal Inequalities and Ethnonationalist Civil War: A Global Comparison. American PoliticalScience Review 1-18. [CrossRef ]

3. Timothy Besley, Torsten Persson. 2011. FRAGILE STATES AND DEVELOPMENT POLICY.

Journal of the European Economic Association no-no. [CrossRef ]

4. Stergios Skaperdas. 2011. The costs of organized violence: a review of the evidence. Economics

of Governance 12:1, 1-23. [CrossRef ]

5. Richard A. Nielsen, Michael G. Findley, Zachary S. Davis, Tara Candland, Daniel L. Nielson.

2011. Foreign Aid Shocks as a Cause of Violent Armed Conflict. American Journal of Political

Science no-no. [CrossRef ]

6. K. Macours. 2011. Increasing inequality and civil conflict in Nepal. Oxford Economic Papers 63:1,1-26. [CrossRef ]

7. Christopher Blattman, Jeannie Annan. 2010. The Consequences of Child Soldiering. Review of

Economics and Statistics 92:4, 882-898. [CrossRef ]

8. Adria Lawrence. 2010. Triggering Nationalist Violence: Competition and Conflict in Uprisings

against Colonial Rule. International Security 35:2, 88-122. [CrossRef ]

9. Debraj Ray, . 2010. Uneven Growth: A Framework for Research in Development

EconomicsUneven Growth: A Framework for Research in Development Economics. Journal of

Economic Perspectives 24:3, 45-60. [Abstract] [View PDF article] [PDF with links]

10. Edward Miguel, Gérard Roland. 2010. The long-run impact of bombing Vietnam. Journal of

Development Economics . [CrossRef ]

11. Daron Acemoglu, Andrea Vindigni, Davide Ticchi. 2010. PERSISTENCE OF CIVIL WARS.

Journal of the European Economic Association 8:2-3, 664-676. [CrossRef ]

12. O. Olsson. 2010. After Janjaweed? Socioeconomic Impacts of the Conflict in Darfur. The World

Bank Economic Review 24:3, 386-411. [CrossRef ]

http://dx.doi.org/10.1016/j.jdeveco.2010.07.004

http://dx.doi.org/10.1257/jep.24.3.45

http://pubs.aeaweb.org/doi/pdf/10.1257/jep.24.3.45

http://pubs.aeaweb.org/doi/pdfplus/10.1257/jep.24.3.45


http://dx.doi.org/10.1162/ISEC_a_00019

http://dx.doi.org/10.1093/wber/lhq019

http://dx.doi.org/10.1111/j.1542-4774.2010.tb00536.x

http://dx.doi.org/10.1016/j.jdeveco.2010.07.004


http://pubs.aeaweb.org/doi/pdf/10.1257/jep.24.3.45

http://dx.doi.org/10.1257/jep.24.3.45

http://dx.doi.org/10.1162/ISEC_a_00019

http://dx.doi.org/10.1162/REST_a_00036

http://dx.doi.org/10.1093/oep/gpq013

http://dx.doi.org/10.1111/j.1540-5907.2010.00492.x

http://dx.doi.org/10.1007/s10101-010-0085-8

http://dx.doi.org/10.1111/j.1542-4774.2011.01022.x

http://dx.doi.org/10.1017/S0003055411000207

http://dx.doi.org/10.1093/qje/qjr022

blattman and miguel 2010_civil war

Documents