Journal of Peace Research - UCLA

http://jpr.sagepub.com

Journal of Peace Research

DOI: 10.1177/0022343304043773 2004; 41; 337 Journal of Peace Research

Michael L. Ross What Do We Know about Natural Resources and Civil War?

http://jpr.sagepub.com/cgi/content/abstract/41/3/337 The online version of this article can be found at:

Published by:

http://www.sagepublications.com

On behalf of:

International Peace Research Institute, Oslo

can be found at:Journal of Peace Research Additional services and information for

http://jpr.sagepub.com/cgi/alerts Email Alerts:

http://jpr.sagepub.com/subscriptions Subscriptions:

http://www.sagepub.com/journalsReprints.navReprints:

http://www.sagepub.com/journalsPermissions.navPermissions:

http://jpr.sagepub.com/cgi/content/refs/41/3/337SAGE Journals Online and HighWire Press platforms):

(this article cites 27 articles hosted on the Citations

© 2004 International Peace Research Institute, Oslo. All rights reserved. Not for commercial use or unauthorized distribution. at UCLA COLLEGE SERIALS/YRL on November 19, 2007 http://jpr.sagepub.comDownloaded from

http://www.prio.no/

http://jpr.sagepub.com/cgi/alerts

http://jpr.sagepub.com/subscriptions

http://www.sagepub.com/journalsReprints.nav

http://www.sagepub.com/journalsPermissions.nav

http://jpr.sagepub.com/cgi/content/refs/41/3/337


337

Introduction

Since the late 1990s, many scholars havestudied the relationship between naturalresource wealth and civil war. Most havebeen motivated by a series of high-profileconflicts – in Angola, Colombia, the Demo-cratic Republic of Congo, Liberia, SierraLeone, and Sudan – that have captured theattention of both international organizationsand the media. Influential studies by Keen(1998), Collier & Hoeffler (1998), Reno

(1995, 1998), Berdal & Malone (2000), andKlare (2001) have touched off new researchby economists, political scientists, andgeographers in both Europe and NorthAmerica.1

What can these studies tell us about therelationship between natural resources –including oil, gas, non-fuel minerals, gem-stones, narcotics, timber, and agriculturalproducts – and violent conflict? At firstglance, the answer appears to be: not much.Table I summarizes the findings of 14 cross-national econometric studies of resourcesand conflict. There appears to be little agree-ment on the validity of the resource–civil war

© 2004 Journal of Peace Research,vol. 41, no. 3, 2004, pp. 337–356Sage Publications (London, Thousand Oaks, CA and New Delhi) www.sagepublications.comDOI 10.1177/0022343304043773 ISSN 0022-3433

What Do We Know About Natural Resources andCivil War?*

MICHAEL L. ROSS

Department of Political Science, University of California, Los Angeles

Since the late 1990s, there has been a flood of research on natural resources and civil war. This articlereviews 14 recent cross-national econometric studies, and many qualitative studies, that cast light onthe relationship between natural resources and civil war. It suggests that collectively they imply fourunderlying regularities: first, oil increases the likelihood of conflict, particularly separatist conflict;second, ‘lootable’ commodities like gemstones and drugs do not make conflict more likely to begin,but they tend to lengthen existing conflicts; third, there is no apparent link between legal agriculturalcommodities and civil war; and finally, the association between primary commodities – a broad categorythat includes both oil and agricultural goods – and the onset of civil war is not robust. The first sectiondiscusses the evidence for these four regularities and examines some theoretical arguments that couldexplain them. The second section suggests that some of the remaining inconsistencies among the econo-metric studies may be caused by differences in the ways they code civil wars and cope with missingdata. The third section highlights some further aspects of the resource–civil war relationship that remainpoorly understood.

* I am grateful to Ian Bannon, Paul Collier, Nils PetterGleditsch, Håvard Hegre, Macartan Humphreys, and fouranonymous reviewers for their encouragement andcomments on earlier drafts of this article. I am alsoindebted to the World Bank’s Conflict Prevention andReconstruction Unit, and Development EconomicsResearch Group, for providing initial funding for thisreview. The views expressed in this article, and all remain-ing errors, are mine alone. Correspondence may be sent [email protected].

1 Other scholars have pursued separate lines of researcharguing that natural resources can produce internationalconflict (e.g. Galtung, 1982; Westing, 1986), and thatresource scarcity can lead to violent conflict (e.g. Homer-Dixon, 1998). These other bodies of research are surveyedin Gleditsch (1998).

06 ross (ds) 15/4/04 4:11 pm Page 337


www.sagepublications.com


correlation. These and other, more quali-tative studies reach varying conclusions on atleast five aspects of the resource–conflictrelationship:

• whether or not natural resources influencethe onset of conflict;

• whether or not resources influence theduration of conflict;

• whether resources influence all types ofcivil wars or only a subtype, e.g. ethnic orseparatist conflicts;

• whether all types of resources, or only asubset (e.g. oil, diamonds) are linked toconflict; and

• what causal mechanisms link resources toconflict.

Quantitative studies of natural resourcesand civil war have also been shadowed byconcerns about misspecification and spuri-ousness. Most scholars measure a country’s‘resource wealth’ by using the ratio of itsresource exports to its GDP. This opens thedoor to two problems. First, there is a dangerthat the causal arrow between naturalresource exports as a fraction of GDP andthe onset of civil war might run the otherway: civil wars might cause resource depen-dence by forcing a country’s manufacturingsector to flee while leaving its resource sector– which is location-specific and cannot easilymove – the major force in the economy bydefault.2 Using lagged independent variablesdoes not eliminate the danger of reversecausality: civil wars can be preceded by yearsof low-level violence that drives off manu-facturing firms, producing a higher level of

resource dependence before the conflictofficially commences.

Second, the natural resource–-civil warcorrelation could be spurious: both civil warand resource dependence might be indepen-dently caused by some unmeasured thirdvariable, such as the weak rule of law. A statewhere the rule of law is weak might beunable to attract investment in its manu-facturing sector, and hence would dependmore heavily on resource exports; it mightalso face a heightened risk of civil warthrough a different process. The result couldbe a statistically significant correlationbetween resource dependence and civil war,even though neither factor would cause theother.

Because the rule of law – and other poten-tially significant missing variables, like thesecurity of property rights – are so difficultto measure across countries and over time, itis hard to test this possibility statistically.

Despite these problems, a close look atboth the quantitative and qualitative studiessuggests four regularities – which could becharacterized as two patterns and two con-spicuous ‘non-patterns’. The first pattern isthat oil exports are linked to the onset ofconflict; the second is that ‘lootable’ com-modities like gemstones and drugs are corre-lated with the duration of conflict. The firstnon-pattern is that agricultural commoditiesseem to be uncorrelated with civil wars, andthe second is that primary commodities – acategory that includes oil, non-fuel minerals,and agricultural goods – is not robustlyassociated with the onset of civil war. Notevery cross-national study fits these fourregularities; still, they are the strongestfindings to emerge so far from this rapidlygrowing field.

What causal mechanisms produce theseoutcomes? Several studies have emphasizedthat we still know little about the processesthat tie natural resources to conflict (Ross,2004a; Humphreys, 2003). Observers often

j ournal o f PE AC E RE S E A RC H volume 41 / number 3 / may 2004338

2 The case of Angola illustrates this pattern. Until its civilwar began, Angola’s economy was relatively diversified andenjoyed high growth: from 1960 to 1974 the economygrew at almost 8% a year. When the war began in 1975,the country’s economic structure collapsed – owing to boththe conflict and the new government’s socialist policies.Between 1973 and 1985, industrial output dropped byalmost half – turning Angola into one of the most resource-dependent countries in the world (Minter, 1994).

06 ross (ds) 15/4/04 4:11 pm 38



Mich

ael L. R

ossN

AT

UR

AL

RE

SO

UR

CE

SA

ND

CIV

ILW

AR

339Table I. Quantitative Research on Resources and Civil War

Coverage Resource measure Dependent variable Finding

Collier & Hoeffler (1998) 27 wars, 1960–92 Primary exports/GDP War onset Increases likelihood of war(curvilinear)

Collier & Hoeffler (2002a) 52 wars, 1960–99 Primary exports/GDP War onset Increases likelihood of war(curvilinear)

Collier & Hoeffler (2002b) 48 wars, 1960–99 Primary exports/GDP War onset Increases likelihood of separatistwars only (curvilinear)

Elbadawi & Sambanis (2002) 108 wars, 1960–99 Primary exports/GDP War onset Weak or no effectFearon & Laitin (2003) 97 wars, 1960–99 Primary exports/GDP War onset No significant effectFearon & Laitin (2003) 122 wars, 1945–99 Oil exporter (dummy) War onset Increases likelihood of war Hegre (2002) 50 wars, 1960–97 Mineral exports/ War onset No significant effect

total exportsHegre (2002) 50 wars, 1960–97 Primary exports/GDP War onset Increases likelihood of war

(curvilinear)Humphreys (2003) 122 wars, 1945–99 Oil production War onset Increases likelihood of warHumphreys (2003) 122 wars, 1945–99 Oil reserves War onset No significant effectHumphreys (2003) 122 wars, 1945–99 Diamond production War onset No significant effectHumphreys (2003) 122 wars, 1945–99 Diamond production War duration Reduces war durationReynal-Querol (2002) 91 wars, 1960–95 Primary exports/GDP War onset and Increases likelihood of non-

prevalence ethnic wars only Elbadawi & Sambanis (2002) 108 wars, 1960–99 Primary exports/GDP War prevalence Weak or no effectCollier, Hoeffler & Söderbom (2004) 52 wars, 1960–99 Primary exports/GDP War duration No significant effectFearon (2004) 122 wars, 1945–99 Contraband (drugs, gems) War duration Increases duration of warDoyle & Sambanis (2000) 124 wars, 1945–97 Primary exports/GDP Peacebuilding Harms successful peacebuilding

successde Soysa (2002b) 77 states, 1989–99 Natural resource Conflict onset No significant effect

stocks/ capita (>25 deaths)de Soysa (2002b) 77 states, 1989–99 Mineral stocks/ capita Conflict onset Decreases likelihood of conflict

(>25 deaths) (curvilinear)de Soysa (2002b) 138 states, 1989–99 Oil exporter (dummy) Conflict onset Increases likelihood of conflict

(>25 deaths)Buhaug & Gates (2002) 262 conflicts, 1946–2000 Mineral resources in Size of conflict Increases size of conflict zone

conflict zone (dummy) zone

06 ross (ds) 15/4/04 4:11 pm Page 339

©

2004 Intern

ation

al Peace R

esearch In

stitute, O

slo. A

ll righ

ts reserved. N

ot fo

r com

mercial u

se or u

nau

tho

rized d

istribu

tion

. at U

CLA

CO

LLEG

E S

ER

IALS

/YR

L on Novem

ber 19, 2007 http://jpr.sagepub.com

Dow

nloaded from


claim that resources have ‘fueled’ a givenconflict but are vague about how thisoccurred. General, cross-national studiesoften suggest causal mechanisms but providelittle evidence to back them up. Case studiestend to provide a wealth of data on causallinks in specific conflicts, but are difficult togeneralize. There are plausible theoriesbehind each of the patterns, though effortsto test them are still in their infancy.

The first section of this article describesthe initial results reported by Collier &Hoeffler (1998, 2002a) that inspired manyof the studies reviewed here: that a country’sdependence on primary commodities isstrongly correlated with the likelihood ofcivil war. It points out that efforts by otherscholars to replicate the primary commodi-ties–civil war correlation have often failed,and suggests that one reason may be that the‘primary commodities’ variable is overlybroad, as it includes a wide range of rawmaterials, some of which may be uncorre-lated with conflict.

It then shows that there is nonethelessstrong evidence, from a wide range of quan-titative and qualitative studies, to supporttwo narrower claims: that oil is linked to theonset of civil war; and that lootable goods areassociated with the duration of civil war. Itdiscusses theories that could explain each ofthese patterns. It also notes the apparentabsence of a link between legal agriculturalcommodities and conflict.

While most quantitative studies are con-sistent with these patterns, studies thatemploy different civil war databases some-times report conflicting results on theresource–civil war relationship. The secondsection suggests that some of these differ-ences may be caused by variations in theways the databases code civil wars, and thedifferent ways that scholars cope withmissing data.

The third section highlights furtheraspects of the resource–conflict issue that

remain poorly understood and are ripe forfurther research. A brief conclusion summa-rizes the central themes.

Primary Commodities and Conflict

Collier & Hoeffler (1998, 2002a) report thatprimary commodity exports tend to increasethe danger of civil war. Their most recentwork covers up to 52 civil wars between1960 and 1999, and suggests that a state’sdependence on natural resources – measuredas the ratio of primary commodity exports toGDP – has a significant influence on thelikelihood that a civil war will begin over thenext five years. Their data suggest thatresource dependence has a non-linear effect:it increases the likelihood of conflict until theresource exports-to-GDP ratio is 32%;beyond this point it diminishes the likeli-hood of conflict. They also find this effect issubstantial: when the other variables are heldat their mean, a rise in resource dependencefrom zero to 32% increases a state’s risk ofcivil war from 1% to 22%. Although theirmost recent work on this topic has not yetbeen published, it has been highly influentialin both the academic and policy communi-ties.

Fearon & Laitin (2003), Elbadawi &Sambanis (2002), and Hegre (2002) have alltried to replicate the Collier & Hoefflerfindings on natural resources and civil war;their results vary. Each of these researchprojects uses a different database of civilwars.

The Fearon & Laitin study provides thesharpest challenge to the Collier & Hoefflerfindings: the authors report that ‘neither theshare of primary commodity exports in GDPnor its square is remotely significant’ in theirmodel (Fearon & Laitin, 2003: 87). Whenthey adjust their database to cover the sameyears as Collier & Hoeffler (1960–99), andgroup the data in five-year periods likeCollier & Hoeffler, the coefficients on their


06 ross (ds) 15/4/04 4:11 pm Page 340



primary commodity variables are evenfarther from statistical significance.

The Elbadawi & Sambanis findings aremore ambiguous: they report that a primarycommodity exports variable is significant insome of their models but not others, depend-ing on the procedure they choose to imputemissing values, and how they operationalizecivil war onsets. They conclude that theCollier & Hoeffler results appear ‘fragile’ and‘cautiously agree with Fearon and Laitin’ thatthere is no significant relationship betweencivil war and natural resource dependence(Elbadawi & Sambanis, 2002: 324–325).

Research by Hegre (2002) has alsoproduced mixed results, although hisfindings lend more support to the Collier &Hoeffler claim. Hegre finds that when heuses a very low threshold for defining conflict(25 deaths in a single year), primary com-modity exports have no influence on thelikelihood of civil conflict. When using the1,000-death threshold employed by Collier& Hoeffler, and a civil war database similarto the Collier & Hoeffler database, his resultsmatch the Collier & Hoeffler findings – eventhough he employs a different set of controlvariables, finer time divisions, and a differentstatistical process. Hegre reports that thecoefficients on his primary commodity vari-ables turn out to be smaller than in theCollier & Hoeffler study, implying thatprimary commodities are somewhat lessimportant than suggested by Collier &Hoeffler.

Two quantitative studies have also lookedat the influence of primary commodityexports on the length of civil wars; their con-clusions differ.3 Collier, Hoeffler &Söderbom (2004) report that primary com-modity exports are not significantly corre-lated with civil war duration, based on theirstudy of 55 civil wars between 1960 and

1999.4 Doyle & Sambanis (2000), however,find that primary commodity exports arenegatively associated with the success ofpeacebuilding efforts in 124 wars between1945 and 1997. Since wars are generallylengthened when peacebuilding efforts fail,their findings imply that primary com-modity exports tend to increase warduration.

Several studies suggest that primary com-modities may only be correlated with asubset of conflicts. Reynal-Querol (2002)divides civil wars into ‘ethnic’ and ‘non-ethnic’ – using the coding of the State FailureTask Force – and finds that primary exportsare correlated only with the onset of non-ethnic wars. Yet Sambanis (2001) alsoexamines the causes of ethnic and non-ethniccivil wars, using a different set of codingrules; he finds that primary commoditieshave no consistent effect.5

In a separate paper, Collier & Hoeffler(2002b) examine the causes of separatistconflicts and find that a state’s dependenceon primary commodity exports is stronglycorrelated with the likelihood it will have aseparatist civil war. The impact of primarycommodity exports is three times as large inseparatist conflicts as in non-separatist con-flicts. For non-separatist civil wars, the influ-ence of primary commodity dependence isnot significant at the 5% level (p = .182). Asin their earlier studies, the relationshipbetween primary commodity exports and theonset of civil wars is curvilinear, decreasingat exceptionally high levels.

What accounts for the conflicting resultsof these studies? One source of variation maybe the differences in the civil war databasesthey employ; this issue is discussed in aseparate section below. Another reason may

Michae l L . Ros s NAT U R A L RE S O U RC E S A N D CI V I L WA R 341

3 Also see the discussion of Fearon (2004) below, on theinfluence of resource contraband on conflict duration.

4 They do find evidence, however, that a drop in primarycommodity prices tends to shorten conflicts; they suggestthis occurs because it weakens the financing of commodity-dependent rebel groups. 5 Sambanis, personal communication, September 2002.

06 ross (ds) 15/4/04 4:11 pm Page 341



be that the ‘primary commodities’ variableand the ‘civil war’ variable are overly broad.The primary commodities variable includescommodities that, according to other studiesdescribed below, appear to be strongly linkedto conflict (e.g. oil) and those that do not(e.g. agricultural goods). If only a subset ofcommodities is associated with conflict, thenthe broad correlation between the primarycommodity variable and conflict may beweak or unstable. Similarly, the civil warvariable measures all types of internal con-flicts that cause a given number of fatalities.But if natural resources are linked to a subsetof internal wars – such as ethnic or separatistwars – again the broader civil war variablemay show only a weak or unstable relation-ship with natural resources.

How should we assess these disparatefindings? The Collier & Hoeffler (2002a)study linking primary commodities to civilwar appears to be robust only when scholarsuse Collier & Hoeffler’s own list of civil wars,or something close to it. Both Hegre (2002)and Humphreys (2003) have successfullyreplicated the Collier & Hoeffler estimationsusing versions of the Collier & Hoeffler civilwar database, and the latter notes that thecorrelation is robust to the exclusion ofoutliers. Yet scholars have found no signifi-cant correlation between primary commodi-ties and violent conflict when usingalternative lists of civil wars, including thosecompiled by Fearon & Laitin, Sambanis, andGleditsch et al. (2002). Moreover, there arereasons to suspect that the Collier & Hoefflerdatabase may be biased in a way that over-states the impact of primary commodities(see below). As a result, the claim thatprimary commodity exports are linked tocivil war appears fragile and should betreated with caution.

Oil and the Onset of ConflictBoth quantitative and qualitative studiessuggest that the production of oil is associ-

ated with the onset of conflict, particularlyseparatist conflict. Although quantitativestudies have generally not explored the roleof non-fuel minerals, qualitative studiesimply that they, too, might be correlatedwith conflict.

Fearon & Laitin (2003) report that adummy variable for oil exporters is corre-lated with the initiation of civil war.6 Theyreport that the oil dummy is not acting as aproxy for either the Middle East–NorthAfrica region or the presence of a largeMuslim population.7 Similarly, de Soysa(2002b) finds that an oil exports dummy issignificantly linked with the likelihood ofconflict – even though his study covers fewerstates and a briefer span of time.

In several of their papers, Collier &Hoeffler also highlight the role of oil exports.In Collier & Hoeffler (2002b), they disag-gregate their primary commodities variableinto five categories of goods and find thatonly oil is significantly linked to civil war intheir baseline regression. Collier & Hoeffler(2002b) suggests that oil is particularlysalient in secessionist civil wars, showing thatan oil dummy can help predict the type ofwar (secessionist vs. non-secessionist), onceprimary commodity exports have been con-trolled for.8 Overall, there is good quantita-tive evidence that oil exports are significantlyassociated with either the full set of civil waronsets (Fearon & Laitin, de Soysa) or a


6 The dummy variable denotes states that derive at leastone-third of their export revenues from oil. Since theauthors lacked data on oil exports for the years between1945 and 1960, they set the values for these years equal tothe values for 1960.7 Humphreys (2003) replicates the Fearon & Laitin esti-mation, using the same dataset, and reports that the corre-lation is not robust: the oil dummy loses significance ifcountries such as Russia and Indonesia are dropped, or ifthe pre-1965 cases are dropped. Yet Humphreys also findsthat when he employs a lagged measure of crude oil pro-duction in place of the oil dummy, oil once again becomessignificantly associated with the onset of civil war. 8 The results are sensitive to the way that secessionist warsare defined: the coefficient on the oil dummy is significantat the 5% level in just one of four alternative codings ofwar types.

06 ross (ds) 15/4/04 4:11 pm Page 342



subset of civil war onsets (Collier &Hoeffler).9

There is also a wealth of case studyevidence linking oil and other minerals toseparatist conflicts: Table II lists nine sepa-ratist conflicts in regions with significantmineral wealth. Important ground-levelresearch on the links between mineral wealthand secessionist movements has been carriedout in the conflict over the Western Saharaissue in Morocco (Hodges, 1983), thestruggle over Cabinda in Angola (Hodges,2001), the many independence movementsin Burma (Lintner, 1999; Sherman, 2003),the Katanga rebellion in the DemocraticRepublic of Congo (Gibbs, 1991), the Acehconflagration in Indonesia (Robinson, 1998;Ross, 2003b), the West Papua independencemovement, also in Indonesia (ICG, 2002),the Bougainville rebellion in Papua NewGuinea (Wesley-Smith, 1992; Regan, 1998),and the rebellion in the south of Sudan(O’Balance, 2000; Anderson, 1999).

Why should oil exports lead to civil war?Many scholars have sketched out incentive-based theories of oil and war, arguing that

the rents available from oil (and sometimes,other minerals) give actors a financial incen-tive to initiate conflict. Some focus on sepa-ratist conflicts, while others look atnon-separatist conflicts.

Collier & Hoeffler (2002b), Le Billon(2001b), and Ross (2003a) have developedoverlapping theories to explain the linksbetween oil and other minerals, and the like-lihood of separatist conflict. Collier &Hoeffler (2002b), drawing on Buchanan &Faith (1987), suggest that the ‘allure ofclaiming ownership of a natural resource dis-covery’ gives populations in peripheralregions an incentive to establish sovereignstates. They suggest this incentive is oftenless rational than it appears, since peopletend to exaggerate the value of the resourcerents available for capture. The likelihood ofresource-based secessions is therefore higherwhen the economic value of the resource ismore difficult to estimate – as is the case withoil and gas – and when the population is lesseducated.

Le Billon (2001b) gives this argumentfurther nuance, suggesting that secessionistmovements are more likely when a resourceis physically concentrated, appropriated bylocals, and requires foreign investment –since locals can attract this investment onlyif their territory achieves recognition as a sovereign state. If the resource is not


Table II. Mineral Resources and Secessionist Movements

Country Region Duration Resources

Angola Cabinda 1975– Oil Burma Hill tribes 1949– Tin, gemsDemocratic Republic of Congo Katanga/Shaba 1960–65 CopperIndonesia West Papua 1969– Copper, gold Indonesia Aceh 1975– Natural gasMorocco West Sahara 1975–88 Phosphates, oilNigeria Biafra 1967–70 OilPapua New Guinea Bougainville 1988– Copper, goldSudan South 1983– Oil

9 There is little evidence that oil has a consistent influenceon the duration of conflicts; the case of Colombia may bean important exception (Dunning & Wirpsa, 2002). Butthere may be an indirect connection: oil may increase thehazard of a secessionist war, and Fearon (2004) finds thatsecessionist wars tend to last much longer than other typesof internal conflicts.

06 ross (ds) 15/4/04 4:11 pm Page 343



geographically concentrated and requireslittle or no foreign investment – such asalluvial diamonds – locals may have anincentive to back a local warlord instead offighting for outright independence.

Ross (2003a) adds a further claim, sug-gesting that resources are more likely toprovoke separatist rebellions if they areextracted through a capital-intensive process,which offers fewer benefits to local, unskilledworkers and more benefits to the state andlarge extraction firms. Resources that can beextracted through labor-intensive methods –such as alluvial diamonds, timber, and agri-cultural goods – produce greater benefits tothe local economy in poor regions and aretherefore less likely to induce secession.

There is less agreement over why oilmight lead to non-secessionist civil wars – ifindeed it does. Several scholars again arguethat the economic allure of oil rents is atfault. Le Billon (2001b), Fearon & Laitin(2003), and others suggest that oil wealthmay increase the value of controlling thestate, hence producing civil wars. Ross(2004a) contends that oil and other mineralwealth may encourage foreign parties to startor support a ‘civil war’ – because mineralwealth either increases the benefits of inter-vention (e.g. the Liberian leader CharlesTaylor’s 1991 support for the RevolutionaryUnited Front in Sierra Leone) or reduces itscost (e.g. the Rwandan government’s ‘self-financing’ incursion into the DemocraticRepublic of Congo in 1998).

Others suggest that civil war is due to thehardships caused by oil wealth, not thebenefits. Switzer (2001), Klare (2001), andothers points out that the extraction processitself can lead to community-level griev-ances, and contend that these may lead tolarger violent conflicts. Fearon & Laitin(2003) argue that oil wealth leads to stateweakness, which in turn causes civil war.Humphreys (2003) notes that resource-dependent countries are more susceptible to

trade shocks, and suggests that trade shocksmay lead to civil war. He also hypothesizesthat mineral dependence might reduce acountry’s level of internal trade, which inturn could diminish the conflict-alleviatingproperties of commercial interaction.

Many of these arguments are, at aminimum, underspecified. The process ofresource extraction sometimes leads to low-level violence (e.g. in Papua New Guinea,Indonesia, and Ecuador); but Ross (2004a)suggests that these low-level conflicts rarelylead to larger civil wars. Resource rents mayincrease the desire of nascent rebels tocapture the state, but not their ability to doso – or even to initiate a civil war. Rebelgroups often face credit constraints: justbecause the value of victory is high does notmean rebels can raise more money for armsand men.10 A resource-rich government mayalso be better able to suppress rebellions;Ross (2001b) shows that both oil depen-dence and non-fuel mineral dependencetend to increase government spending on themilitary.

Gemstones, Drugs, Timber, and ConflictIn the case-study literature, and the popularmedia, civil wars are commonly linked togemstones, drugs, and timber – what LeBillon (2001b) and others call ‘lootable’commodities. Between 1990 and 2000, civilwars occurred in five diamond-producingstates, three opium-exporting states, threemajor cannabis-exporting states, and twoleading coca-exporting states (Table III). Thepartial influence of these commodities oncivil war is difficult to test, however, owingto a shortage of reliable data. Most evidencethus far suggests that gemstones and nar-cotics are linked to the duration of conflict,but surprisingly not to the initiation of


10 Only in exceptional circumstances should this bepossible: when rebels can raise funds by selling future rightsto war booty, or ‘booty futures’. On the small number ofcases where this has occurred, see Ross (2002).

06 ross (ds) 15/4/04 4:11 pm Page 344



conflict. Little research has been done on therole of timber.

Just one quantitative study of diamondsand the onset of conflict has been carried outto date. Humphreys (2003) develops ameasure of diamond production and placesit in a civil war model based on the work ofFearon & Laitin; he finds no significantrelationship.

This finding matches closely the con-clusion of the qualitative analysis in Ross(2004a). This study looks at most of the civilwars in the 1990s that occurred in gemstone-producing states – including Afghanistan,Angola, Burma, Cambodia, DemocraticRepublic of Congo, Liberia, and SierraLeone – and finds that in either five or six ofthese seven cases, the gemstone trade appearsto be causally unrelated to the initiation ofconflict.11 It argues that the lure of gemstonewealth may have motivated rebels in SierraLeone, and possibly in the DemocraticRepublic of Congo, but in the other casesgemstones only became salient years after theconflicts began. In most cases, causality ran

in the opposite direction: the conflict helpedcause the rebel groups to become dependenton gemstone sales, since in countries devas-tated by war (Afghanistan, Angola, Burma,Cambodia, and Liberia) these were amongthe few remaining products that rebels coulduse to raise money.

It also suggests that a similar pattern holdsfor the production of opium and coca: afterexamining the four civil wars in the 1990sthat occurred in major opium- and coca-exporting states (Afghanistan, Burma,Colombia, and Peru), it concludes that nar-cotics had little to do with the initiation ofthese conflicts. In each of these countries, therebel movements had no role in the drugtrade when the war began. It again contendsthat in several cases (Burma and Colombia)the causal arrow ran in the opposite direc-tion: the civil war led to drug production bycreating a region that fell outside the govern-ment’s control for a series of consecutivegrowing seasons. In the two other two cases(Afghanistan and Peru), both drug produc-tion and civil war seemed to be indepen-dently associated with the weakness of thestate’s jurisdiction in remote rural areas.

Even if gems and drugs are not linked tothe initiation of conflict, there is goodevidence that they tend to lengthen pre-existing conflicts. Fearon (2004) finds thatthe availability of contraband resources(generally drugs or gemstones) in a conflictzone tends to lengthen civil wars. He notesthat these resources are often found in aspecial class of separatist conflicts (which heand Laitin dub ‘sons of the soil’ conflicts)that tend to last much longer than othertypes of civil wars.

This finding is consistent with a widearray of case studies. Ross (2004a) covers tencivil wars between 1990 and 2000 in statesthat produced either gemstones, opium, orcoca; it finds that these resources lengthenedseven of the ten conflicts, principally byallowing the weaker party to raise money and


Table III. Gemstone and Drug Producers thathad Civil Wars in the 1990s

Country Gemstones/Drugs

Angola DiamondsDemocratic Republic

of Congo DiamondsLiberia DiamondsRussia DiamondsSierra Leone DiamondsAfghanistan Opium, cannabis, gemsBurma Opium, gemsCambodia Cannabis, gemsColombia Opium, cocaPeru Coca

11 Sierra Leone was a clear exception; diamonds may havealso played a tangential role in the initiation of war in theDemocratic Republic of Congo. Russia was the only otherdiamond-producing country in the 1990s to have a civilwar. In this case, too, the civil war (in Chechnya) had noobvious link to the diamond trade.

06 ross (ds) 15/4/04 4:11 pm Page 345



hence continue fighting. In a study of 16efforts to implement peace accords between1987 and 2000, Stedman (2001: 2) foundthat ‘no peace agreement has been success-fully implemented where there are valuable,easily marketable commodities such as gemsor timber’ – implying that lootable goodstend to prolong conflicts.12 In general, thesestudies show how the sale of gems or drugsallowed the weaker party to a conflict –usually but not always the rebel group – tofinance itself, and hence continue fightinginstead of being crushed or forced to thenegotiating table. In several cases, lootableresources have also made conflict so profit-able that one or more combatants lose theirincentive to reach a peace settlement.13

Still, the claim that gems and drugs tendto lengthen conflicts cannot be consideredsettled. Humphreys (2003), employing aunique database on diamond exports, findsthat diamonds tend to shorten civil wars –although by facilitating military victories,not negotiated settlements. Given the diffi-culty of assembling accurate data on thediamond trade – and, more importantly, thegemstone trade – this issue would profit fromfurther scrutiny.14

Although the role of timber has not yetbeen explored statistically, case studiessuggest that timber exports may have influ-enced the duration (though not theinitiation) of three recent civil wars:Cambodia, Liberia, and the DemocraticRepublic of Congo (Global Witness, 2002).Because this issue has not been exploredstatistically, we cannot know if these areisolated cases or part of a larger pattern.

Timber had the clearest effect inCambodia. Between 1989 and 1995, therebel Khmer Rouge was able to maintain itsviability as a military force owing to its salesof both timber and gemstones; when thisrevenue dropped off after 1995, the KhmerRouge gradually fell apart, and by 1998 ithad collapsed. The sale of timber may havelengthened the conflict by several years (LeBillon, 2000; Brown & Zasloff, 1998).

In the other two civil wars, timber playedat most a small role in perpetuating theconflict. In Liberia, rebel forces led byCharles Taylor raised money from the sale oftimber, but this was part of Taylor’s well-diversified portfolio of booty, whichincluded virtually everything that could besold.15 His most egregious misuse of forestresources came after the civil war ended,when he used timber sales to consolidate hisregime. In the Democratic Republic ofCongo, the Kabila government offeredtimber concessions to Zimbabwe inexchange for Zimbabwe’s military assistance,but it is unclear whether any timber wasactually harvested (Global Witness, 2002).

Agricultural Commodities and ConflictCollier & Hoeffler (2002b) find that a state’sdependence on agricultural commoditiesdoes not influence its civil war risk. None ofthe other major studies report tests of theagriculture–civil war relationship; still, the


12 Other studies that describe how the export of gems andnarcotics has lengthened civil wars include Reno’s (1998)analysis of conflicts in Sierra Leone, Liberia, and theDemocratic Republic of Congo; reports by United NationsPanels of Experts (2000, 2001, 2002) on Sierra Leone andthe Democratic Republic of Congo; studies of the conflictsin Cambodia and Angola by Le Billon (2000, 2001a);research on the Peruvian conflict by Palmer (1994); and astudy of the Colombian conflict by Reyes (1994).13 Several scholars have argued that alluvial diamonds aremore hazardous than kimberlite diamonds. Alluvialdiamonds are found near the earth’s surface and can beextracted by artisanal miners. Kimberlite diamonds, bycontrast, are buried deep in volcanic rock and can only beextracted by large-scale industrial mining. Case studies ofcountries with valuable alluvial gemstones sometimes linkthese deposits to state weakness, noting that theycommonly lead to corruption, the fragmentation of stateauthority, and the rise of private predation. See, forexample, Reno (1995); Richards (1996); MacGaffey &Bazenguissa-Ganga (2000).14 On the paucity of accurate data on the diamond trade,see Smillie (2002).

15 As noted above, what matters most is whether a rebelgroup relied disproportionately on any given economicsector.

06 ross (ds) 15/4/04 4:11 pm Page 346



Collier & Hoeffler study is consistent withmany civil war case studies.

Even though most rebellions occur inpoor countries, and most poor countries areoverwhelmingly agricultural, it seems to berare for rebel forces to receive a dispropor-tionate fraction of their income from theagricultural sector.16 In several civil wars – inLiberia, the Democratic Republic of Congo,and Indonesia (Aceh) – rebels have receivedpart of their income from the agriculturalsector; but in each case, this was apparentlydue to rebel fundraising efforts that cutacross all sectors of the rural economy (Ellis,1999; UN Panel of Experts, 2001, 2002;Ross, 2003b). Since most rebel groupsoperate in rural – and hence, primarily agri-cultural – regions, we should expect them toraise at least part of their income from taxeson agricultural goods. But there is littleevidence so far that agricultural commodi-ties, other than opium and coca, are signifi-cantly associated with either the onset orduration of civil wars.

How Different Civil War DatabasesMay Produce Different Results

Why do some of the quantitative studies ofcivil war disagree over the role of naturalresources? One reason may be that they usedifferent datasets. Recent studies have drawnupon four civil war datasets: Collier &Hoeffler (2002a), Fearon & Laitin (2003),Elbadawi & Sambanis (2002), and Gleditschet al. (2002, also known as the Uppsaladataset).17 These datasets, in turn, draw onthe pioneering ‘Correlates of War’ civil warsdataset compiled by Singer & Small (1994).Since there is no consensus definition of acivil war, and information on civil wars is

generally poor, it is hard to differentiateamong these datasets by quality.18

There are at least three consequentialdifferences among these datasets. First, theyuse somewhat different criteria for determin-ing when a civil war has occurred. Between1960 and 1999, Gleditsch et al. observe 111civil wars, Collier & Hoeffler 78 civil wars,Fearon & Laitin 97 wars, and Elbadawi &Sambanis 108 wars. If these scholars eachemploy their own coding rules consistently,and if natural resources affect all types of civilwars equally, this should make no difference.But this might not be the case. Imagine, forexample, that natural resources tend to havea strong effect on large conflicts but no influ-ence on small ones; and further, that onedataset uses narrow coding rules that classifyonly large conflicts as ‘civil wars’, whileanother uses broader rules that also classifyboth large and small conflagrations as civilwars. In this case, natural resources might besignificantly associated with civil wars in thefirst dataset but not the second, even thoughboth datasets apply their coding rules faith-fully and consistently.

Second, different scholars use differentstrategies for coping with missing data. Ofthe four studies, only Collier & Hoeffler uselistwise deletion, dropping civil wars fromtheir estimations when data is missing on oneor more of the variables in their models. Theother three use either imputation or interpo-lation to fill in missing values for the neededvariables, so that data-scarce civil wars canstill be used in their estimations. BecauseCollier & Hoeffler use listwise deletion, theirregressions can employ no more than two-thirds (52 of 78) of the civil wars in theirdataset. Listwise deletion can produce biasedinferences if the missing observations arenon-random. A glance at the Collier &Hoeffler list of civil wars suggests that the


16 The Casamance rebellion in Senegal, which receivedmuch of its funding from the region’s cashew crop, may bean exception. See Humphreys & Mohamed (2002).17 Hegre (2002) employs the Gleditsch et al. (2002)database.

18 For helpful discussions of the differences among thedatasets, see Sambanis (2001) and Gleditsch et al. (2002).

06 ross (ds) 15/4/04 4:11 pm Page 347



deleted conflicts tend to be less dependent onprimary commodity exports than theincluded conflicts: for their 46 civil wars withcomplete data on the major variables, themean primary exports-to-GDP ratio is .167;for the civil wars they drop, it is .118.19 Thedeletion of these cases should raise the meanprimary exports-to-GDP ratio for theremaining civil wars. If we assume that anynon-wars that are dropped are randomly dis-tributed with respect to primary commodi-ties, the result could be a spurious correlationbetween primary commodity exports and thelikelihood of conflict.

Finally, the datasets differ in how theydetermine when a war has ended. This mayproduce some slippage between the correlatesof war onset and the correlates of war duration.In some resource-dependent countries, such asAfghanistan, Angola, Burma, and the Congo,conflict has fluctuated in intensity over severaldecades. Some observers might code theseconflicts as many civil wars of brief duration;others interpret them as a small number of civilwars of long duration. Those who code theseas many brief wars may subsequently find thata variable that is common to these countries ispositively correlated with the onset of war(since this handful of countries has producedmany war onsets) and negatively correlatedwith war duration (since all these wars arebrief). Yet those who code them as wars of longduration may find the same variable is notsignificantly correlated with the onset of war(since these countries would be coded as pro-ducing only a small number of wars), but posi-tively correlated with the duration of war(since these wars are now long). These codingdifferences could explain why the databaseshave widely varying estimates of the numberof civil wars for the same periods.

This problem may also help explain whytwo major research projects have come todivergent conclusions: Collier & Hoefflerfind that resources are associated with theonset of civil war but not its duration, whileFearon finds that some resources are linkedto the duration of conflict, but Fearon &Laitin find that resources (except for oil) areuncorrelated with civil war onsets.

Some Poorly Understood Issues

Many dimensions of the resource–conflictrelationship are still unclear. This article hasalready highlighted four nagging sources ofambiguity: the problems of endogeneity andmisspecification in the resource–conflictcorrelation; the question of whether oilwealth only triggers separatist conflicts, or alltypes of conflicts; the question of causallinkages; and possible slippage betweenconflict onset and conflict duration. Beloware some additional questions that, ifthoughtfully pursued, could significantlyadvance our understanding of theresource–conflict relationship.

Why Have Quantitative StudiesProduced Varying Results? The major quantitative studies of civil warhave produced different findings about therole of natural resources; these differences werepresumably caused by differences in the under-lying databases, models of civil war, estimationprocedures, and techniques for handlingmissing data. Reconciling these findings – andidentifying the factors that have produced thedifferences in the findings – would help usidentify which of the resource–conflictrelationships are truly robust.

Has the Role of Natural ResourcesChanged Over Time? It should not be surprising that theresource–conflict question gained broadattention in the late 1990s: many of the


19 This mean is derived from the 26 dropped observationsfor which data on primary exports is available. Six otherobservations were also dropped: Azerbaijan 1990, Bosnia1990, Georgia 1990, Tajikistan 1990, Yemen PR 1985,and Yugoslavia 1995.

06 ross (ds) 15/4/04 4:11 pm Page 348



decade’s civil wars occurred in countries thatdepend heavily on oil and mineral exports.But is the resource–conflict association arecent phenomenon?

There are several hints in the qualitativeliterature that it is. Keen (1998) and othershave suggested that the end of the Cold Warturned natural resources into an importantsource of rebel funding; indeed, there isstrong evidence that longstanding rebelgroups in Angola and Cambodia began torely on resource-looting after the Cold War’send caused them to lose their superpowerfunding (Le Billon, 2003b). Several long-running civil wars in mineral-dependentstates – like those in Sudan and Aceh, Indo-nesia – seemed to grow more serious in the1990s.

If this is true, it should be reflected in thequantitative studies: a dummy variable forthe post-Cold War era, or even the decade ofthe 1990s, should be highly significant wheninteracted with measures of naturalresources. None of the major studies have yetreported any temporal changes in the signifi-cance of their resource measures. In fact,Humphreys (2003) reports that in theFearon & Laitin model, the ‘oil exporter’dummy loses significance when observationsprior to 1965 are dropped from the estima-tions – which is the opposite of what wewould expect if natural resource exports havegrown more salient over time.

What Is the Role of Non-Fuel Minerals?The link between non-fuel minerals and civilwar is ambiguous, in part because mineralwealth has received less scrutiny than oilwealth. While the distinction between oiland non-fuel minerals may seem intellectu-ally trivial, it could have huge policy conse-quences for mineral firms andmineral-exporting states.

Both Collier & Hoeffler (2002b) andFearon & Laitin (2003) look at how oilinfluences the likelihood of civil war, but

they offer no test of non-fuel minerals. Hegre(2002) and de Soysa (2002b) both placenon-fuel minerals and oil in a single category,but do not test them separately. As a conse-quence, we do not know if non-fuel mineralspose the same problems as oil and gas.

Which Dimensions Are Dangerous? It is not yet clear what specific qualities makenatural resources likely to ignite or prolongconflicts. Many have made plausible argu-ments; none has been carefully tested. It isnot obvious, for example, whether resourcesmust be exported to be hazardous. Almost allquantitative research on resources andconflict measures natural resource exports,not natural resource production.20 Isresource production as dangerous as resourceexports?

Some scholars suggest that resources areharmful because they produce rents (Torvik,2002). If true, this should be possible toobserve. Different types of resources producedifferent levels of rents; even the samequantity of the same resource can producedifferent quantities of rents in differentcountries, owing to differences in productioncosts and quality. Testing the impact of rentsseparately from resource production orexports could illuminate many untestedhypotheses about rent-seeking and open upnew avenues of research.

Snyder (2003) argues that the key variableis the ruler’s ability to control the extractionprocess: lootable resources only breeddisorder, he suggests, when a ruler lacks thiscontrol. Other studies have implied thatlocal politics around the mine site – includ-ing relations between the extraction firm andlocal communities – determines whether ornot a country’s mineral wealth will lead toviolence (Switzer, 2001; Swanson, 2002).


20 De Soysa (2002a) and Humphreys (2003) are import-ant exceptions.

06 ross (ds) 15/4/04 4:11 pm Page 349



These hypotheses have powerful policyimplications and should be tested.

Several scholars have suggested that thephysical qualities of a resource, and its geo-graphic location, influence the likelihood itwill produce conflict; these have not beencarefully tested, however. Le Billon (2001b),drawing on Auty (2001), makes two key dis-tinctions among resources: between thosethat are proximate to a national capital (andhence easier for governments to capture) andthose that are distant (and hence easier forrebels to hold); and between ‘point source’resources, which are concentrated in a smallarea (and therefore more easily controlled bya single group) and diffuse resources, whichare scattered over a larger area (and henceharder for any single group to capture).These two categories, Le Billon suggests,yield a fourfold typology of conflict: point-source resources near the capital createviolent incentives to control the state, andhence produce coup d’etats; point-sourceresources that are far from the capitalproduce secession movements; diffuseresources near the capital lead to rebellionsand rioting; and diffuse resources far fromthe capital lead to ‘warlordism’, areas of defacto political autonomy. Auty (2004) largelyconcurs with Le Billon, and adds that thedangerousness of a resource is correlated withits value-to-weight ratio. Those with higherratios, like diamonds and drugs, are likely topose greater hazards than those with lowratios, like agricultural commodities.

Ross (2003a) develops an alternativetheory based on whether or not resources are‘lootable’ and ‘obstructable’. Resources arelootable if they can be easily appropriated byindividuals or small groups of unskilledworkers. It suggests that lootable resources –such as diamonds and drugs – pose littledanger of igniting separatist conflicts, sincethey provide many benefits to the localeconomy; but they make non-separatist con-flicts harder to resolve, since they offer

greater marginal benefits to weaker combat-ants, who might otherwise be forced to sur-render, and because they tend to creatediscipline problems within the armies them-selves. It also argues that unlootableresources – like oil, natural gas, and deep-shaft minerals – tend to produce separatistconflicts, since they provide relatively fewbenefits for the local economy, but that theyseldom influence non-separatist conflicts.Finally, it contends that when a resource iseasily ‘obstructable’ – that is, its transporta-tion can be easily blocked by a small numberof individuals with relatively inexpensiveweapons – it will tend to increase theduration and intensity of conflicts, again byoffering the weaker party an advantage.

Each of these typologies is plausible, butspeculative. All can and should be tested,either with quantitative or qualitative data.

How Is the Resource–Civil War IssueLinked to the Resource Curse?The effect of natural resource dependence onconflict may be part of a larger phenomenon:the resource curse. States that depend heavilyon the export of oil, gemstones, and mineralstend to suffer from a variety of problems,including:

• slow economic growth (Sachs & Warner,1995, 1999, 2001; Gylfason, 2001; Leite& Weidmann, 1999; Doppelhofer, Miller& Sala-i-Martin, 2000);

• high poverty rates (Ross, 2003c); • high corruption levels (Sachs & Warner,

1999; Leite & Weidmann, 1999; Collier& Gunning, 1999; Marshall, 2001;Schloss, 2002); and

• authoritarian governance (Ross, 2001a;Wantchekon & Lam, 1999; and de Soysa,2002a).

It would be remarkable indeed if thesepatterns were unrelated to civil conflict. Buthow? Many studies tie at least one of these


06 ross (ds) 15/4/04 4:11 pm Page 350



phenomena to the incidence of civil war:slow economic growth (Collier & Hoeffler,2002a; Elbadawi & Sambanis, 2002), highpoverty rates (Collier & Hoeffler, 2002a),regime type (DeNardo, 1985; Muller &Weede, 1990; Hegre et al., 2001), orsomewhat obliquely, corruption (Reno,1995; Fearon & Laitin, 2003; Le Billon,2003b). There may also be indirect links.Countries with large resource exports tend totax their population less heavily; Fearon &Laitin (2003) argue that low taxes lead tostate weakness, which in turn increases thelikelihood of conflict, and Ross (2004b)suggests that low taxes tend to impededemocracy. Both Gylfason (2001) and Ross(2001a) find that states with large resourcesectors tend to have lower education rates,and Collier & Hoeffler (2002a) contend thatlow education levels boost the risk of civilwar.21

All of these studies suggest ways that civilwars in resource-exporting states may belinked to other elements in the resourcecurse. A comprehensive model that incor-porates these disparate findings – and morecarefully specifies their relationships – wouldbe an important advance.

What Are the Appropriate PolicyInterventions?There has been much scientific research onhow resources influence conflict, but little onwhat can be done about it. This schismreflects the larger disjuncture between theacademic and policy communities on theproblem of civil wars (Mack, 2002).

In many ways, the policy world has beenahead of the academic world. The UNSecurity Council has taken measures againstthe sale of natural resources by rebel forces inSierra Leone, Liberia, the DRC, and Angola.A major effort to restrict the trade in ‘conflict

diamonds’ was launched in May 2000, at aconference in Kimberley, South Africa; the‘Kimberley Process’ entails an agreement bythe diamond industry to trade only diamondsthat can be certified as originating from legiti-mate sources (Le Billon, 2003a). Scholarshave also produced papers that discuss policyoptions on resources and conflict, including astudy by Switzer (2001), reports commis-sioned by Fafo’s ‘Economies of Conflict’project, a book authored by Le Billon (2003b)for the International Institute of StrategicStudies, and a book produced by the WorldBank (Collier et al., 2003).22 While many ofthese papers have well-reasoned argumentsabout what can and should be done to severthe link between commodities and conflict, ingeneral their arguments can and should bestrengthened by systematic testing.

Some of these papers have ambiguous orcontradictory arguments about the mechan-isms that link resources to conflict, which canleading to conflicting policy recommen-dations. For example, Switzer (2001) suggeststhat mining causes conflict because itproduces grievances over environmentaldegradation and access to jobs; in this case,the solution might be greater communityinvolvement by mining firms. But if conflictsoccur because mining provides extortionopportunities for rebel groups and warlords –as Collier & Hoeffler (2002a) suggest – thenthe solution might be stricter mine sitesecurity and less community involvement.Progress in understanding which dimensionsof resources are harmful, and what the causalmechanisms are, could therefore give scholarsa sounder footing for their recommendations.

Conclusion

Over the last five years there has been awealth of new scholarship on the relationship


21 Stijns (2001), however, finds that resource rents are posi-tively correlated with education rates, perhaps because theygenerate higher levels of government revenue.

22 The Fafo papers are available at www.fafo.no/piccr/ecocon.htm.

06 ross (ds) 15/4/04 4:11 pm Page 351



between natural resources and civil war. Theweight of the evidence available so farsuggests four regularities:

(1) oil dependence appears to be linked tothe initiation of conflict, but not conflictduration. There is some evidence that oildependence (and possibly mineraldependence) is more strongly associatedwith separatist conflicts than other typesof conflicts.

(2) Gemstones, opium, coca, and cannabisdo not seem to be linked to theinitiation of conflict, but they do seemto lengthen pre-existing wars. Timber’srole remains untested.

(3) There is no statistical evidence – andvery little case study evidence – that linksagricultural commodities to either theinitiation or the duration of civil war.

(4) The claim that primary commodities areassociated with the onset of civil wardoes not appear to be robust.

Research on resources and conflict hasdeveloped remarkably fast; almost all of thestudies described here have been carried outsince the 1998 publication of Collier &Hoeffler’s seminal article. Yet events in thepolicy world have developed even morequickly. If scholars wish to produce researchthat contributes to better policies, they mustthink carefully about testing their ideas;sharing data; speaking a common conceptuallanguage; and drawing together the findingsof disparate research projects – both quali-tative and quantitative – that cast light onthe determinants of civil war.

References

Anderson, G. Norman, 1999. Sudan in Crisis.Gainesville, FL: University Press of Florida.

Auty, Richard M., ed., 2001. Resource Abundanceand Economic Development. New York: OxfordUniversity Press.

Auty, Richard M., 2004. ‘Natural Resources andCivil Strife: A Two-Stage Process’, Geopolitics9(1): 29–49.

Berdal, Mats & David M. Malone, eds, 2000.Greed and Grievance: Economic Agendas inCivil Wars. Boulder, CO: Lynne Rienner.

Brown, MacAlister & Joseph J. Zasloff, 1998.Cambodia Confounds the Peacemakers. Ithaca,NY: Cornell University Press.

Buchanan, James M. & Roger L. Faith, 1987.‘Secession and the Limits of Taxation: Towarda Theory of Internal Exit’, American EconomicReview 77(5): 1023–1031.

Buhaug, Halvard & Scott Gates, 2002. ‘TheGeography of Civil War’, Journal of PeaceResearch 39(4): 417–433.

Collier, Paul & Jan Willem Gunning, eds, 1999.Trade Shocks in Developing Countries. NewYork: Oxford University Press.

Collier, Paul & Anke Hoeffler, 1998. ‘OnEconomic Causes of Civil War’, OxfordEconomic Papers 50(4): 563–573.

Collier, Paul & Anke Hoeffler, 2002a. ‘Greed andGrievance in Civil War’, Oxford University,Centre for the Study of African Economies,Working Paper 2002–01 (http://www.csae.ox.ac.uk/workingpapers/pdfs/2002–01text.pdf ).

Collier, Paul & Anke Hoeffler, 2002b. ‘ThePolitical Economy of Secession’, manuscript,30 June (http://www.worldbank.org/ogmc/pdfs/Paul-Collier.pdf ).

Collier, Paul; Anke Hoeffler & Måns Söderbom,2004. ‘On the Duration of Civil War’, Journalof Peace Research 41(3): 253–273.

Collier, Paul; V. L. Elliot, Håvard Hegre, AnkeHoeffler, Marta Reynal-Querol & NicholasSambanis, 2003. Breaking the Conflict Trap:Civil War and Development Policy. Washing-ton, DC: World Bank & Oxford UniversityPress.

DeNardo, James, 1985. Power in Numbers.Princeton, NJ: Princeton University Press.

de Soysa, Indra, 2002a. ‘Ecoviolence: ShrinkingPie or Honey Pot?’, Global EnvironmentalPolitics 2(4): 1–34.

de Soysa, Indra, 2002b. ‘Paradise Is a Bazaar?Greed, Creed, and Governance in Civil War,1989–99’, Journal of Peace Research 39(4):395–416.

Doppelhofer, Gernot; Ronald I. Miller & Xavier


06 ross (ds) 15/4/04 4:11 pm Page 352



Sala-i-Martin, 2000. ‘Determinants of Long-Term Growth: A Bayesian Averaging ofClassical Estimates (BACE) Approach’,National Bureau of Economic Research,Working Paper 7750, June (http://www.nber.org/papers/w7750.pdf ).

Doyle, Michael & Nicholas Sambanis, 2000.‘International Peacebuilding: A Theoreticaland Quantitative Analysis’, American PoliticalScience Review 94(4): 779–801.

Dunning, Thad & Leslie Wirpsa, 2002. ‘Oil andthe Political Economy of Conflict inColombia and Beyond: A Linkages Approach’(http://www.santafe.edu/files/gems/obstaclestopeace/wirpsadunning.pdf ). Forthcoming inGeopolitics.

Elbadawi, Ibrahim & Nicholas Sambanis, 2002.‘How Much War Will We See? Explaining thePrevalence of Civil War’, Journal of ConflictResolution 46(3): 307–334.

Ellis, Stephen, 1999. The Mask of Anarchy: TheDestruction of Liberia and the Religious Dimen-sion of an African Civil War. New York: NewYork University Press.

Fearon, James D., 2004. ‘Why Do Some CivilWars Last So Much Longer Than Others?’,Journal of Peace Research 41(3): 275–301.

Fearon, James D. & David D. Laitin, 2003. ‘Eth-nicity, Insurgency, and Civil War’, AmericanPolitical Science Review 97(1): 75–90.

Galtung, Johan, 1982. Environment, Develop-ment, and Military Activity: Towards Alterna-tive Security Doctrines. Oslo: NorwegianUniversity Press.

Gibbs, David N., 1991. The Political Economy ofThird World Intervention: Mines, Money, andU.S. Policy in the Congo Crisis. Chicago, IL:University of Chicago Press.

Gleditsch, Nils Petter, 1998. ‘Armed Conflict andthe Environment: A Critique of the Litera-ture’, Journal of Peace Research 35(3):381–400.

Gleditsch, Nils Petter; Peter Wallensteen, MikaelEriksson, Margareta Sollenberg & HåvardStrand, 2002. ‘Armed Conflict 1946–2001: ANew Dataset’, Journal of Peace Research 39(5):615–637.

Global Witness, 2002. ‘The Logs of War: TheTimber Trade and Armed Conflict’, Pro-gramme for International Co-operation and

Conflict Resolution, Fafo-report 379, March(http://www.fafo.no/pub/rapp/379/379.pdf ).

Gylfason, Thorvaldur, 2001. ‘Natural Resources,Education, and Economic Development’,European Economic Review 45 (May):847–859.

Hegre, Håvard, 2002. ‘Some Social Requisites ofa Democratic Civil Peace: Democracy,Development, and Armed Conflict’, paperpresented to the Annual Meeting of theAmerican Political Science Association,Boston, MA, 29 August–1 September(http://econ.worldbank.org/files/18170_DDCWAPSA2002.PDF).

Hegre, Håvard; Tanja Ellingsen, Scott Gates &Nils Peter Gleditsch, 2001. ‘Toward a Demo-cratic Civil Peace? Democracy, PoliticalChange, and Civil War, 1816–1992’,American Political Science Review 95(1):33–48.

Hodges, Tony, 1983. Western Sahara: The Roots ofa Desert War. Westport, CT: Lawrence Hill.

Hodges, Tony, 2001. Angola: From Afro-Stalinismto Petro-Diamond Capitalism. Bloomington,IN: Indiana University Press.

Homer-Dixon, Thomas, 1998. Environment,Scarcity, and Conflict. Princeton, NJ: Prince-ton University Press.

Humphreys, Macartan, 2003. ‘Natural Resource,Conflict, and Conflict Resolution’, paperprepared for the Santa Fe Institute/JaverianaUniversity ‘Obstacles to Robust NegotiatedSettlements’ workshop, Bogota, 29–31 May(http://www.santafe.edu/files/gems/obstaclestopeace/hymphreys.pdf ).

Humphreys, Macartan & Habaye ag Mohamed,2002. ‘Senegal and Mali’, paper presented atthe World Bank/Center for United NationsStudies Conference, Yale University, 12–15April (http://www.columbia.edu/~mh2245/papers1/sen_mali.pdf ).

International Crisis Group, 2002. ‘Indonesia:Resources and Conflict in Papua’, ICG AsiaReport 39, 13 September (http://www.crisisweb.org//library/documents/report_archive/A400774_13092002.pdf ).

Keen, David, 1998. ‘The Economic Functions ofViolence in Civil Wars’, Adelphi Paper 320.London: International Institute of StrategicStudies.


06 ross (ds) 15/4/04 4:11 pm Page 353



Klare, Michael T., 2001. Resource Wars. NewYork: Metropolitan.

Le Billon, Philippe, 2000. ‘The Political Ecologyof Transition in Cambodia 1989–1999: War,Peace and Forest Exploitation’, Developmentand Change 31(4): 785–805.

Le Billon, Philippe, 2001a. ‘Angola’s PoliticalEconomy of War: The Role of Oil andDiamonds 1975–2000’, African Affairs100(398): 55–80.

Le Billon, Philippe, 2001b. ‘The Political Ecologyof War: Natural Resources and Armed Con-flicts’, Political Geography 20(5): 561–584.

Le Billon, Philippe, 2003a. ‘Getting It Done:Instruments of Enforcement’, in Ian Bannon& Paul Collier, eds, Natural Resources andViolent Conflict: Options and Actions.Washington, DC: World Bank (215–286).

Le Billon, Philippe, 2003b. Fuelling War: NaturalResources and Armed Conflicts. Oxford: OxfordUniversity Press for the International Instituteof Strategic Studies.

Leite, Carlos & Jens Weidmann, 1999. ‘DoesMother Nature Corrupt? Natural Resources,Corruption, and Economic Growth’, IMFWorking Paper WP/99/85 (http://www.imf.org/external/pubs/ft/wp/1999/wp9985.pdf ).

Lintner, Bertil, 1999. Burma in Revolt: Opiumand Insurgency Since 1948, 2nd edn. ChiangMai: Silkworm.

MacGaffey, Janey & Remy Bazenguissa-Ganga,2000. Congo–Paris: Transnational Traders onthe Margins of Law. Bloomington, IN: IndianaUniversity Press.

Mack, Andrew, 2002. ‘Civil War: AcademicResearch and the Policy Community’, Journalof Peace Research 39(5): 515–525.

Marshall, Ian E., 2001. ‘A Survey of CorruptionIssues in the Mining and Mineral Sector’,Report No. 15, Minerals, Mining, and Sus-tainable Development, International Institutefor Environment and Development (http://www.iied.org/mmsd/mmsd_pdfs/marshall_corruption.pdf ).

Minter, William, 1994. Apartheid’s Contras: AnInquiry into the Roots of War in Angola andMozambique. London: Zed.

Muller, Edward N. & Erich Weede, 1990. ‘Cross-National Variation in Political Violence’,Journal of Conflict Resolution 34(4): 624–651.

O’Ballance, Edgar, 2000. Sudan, Civil War, andTerrorism 1956–1999. New York: St. Martin’s.

Palmer, David Scott, 1994. ‘Peru, Drugs, and theShining Path’, in Bruce M. Bagley & WilliamO. Walker III, eds, Drug Trafficking in theAmericas. Miami, FL: North-South CenterPress (179–191).

Regan, Anthony J., 1998. ‘Current Develop-ments in the Pacific: Causes and Course of theBougainville Conflict’, Journal of PacificHistory 33(3): 269–285.

Reno, William, 1995. Corruption and StatePolitics in Sierra Leone. New York: CambridgeUniversity Press.

Reno, William, 1998. Warlord Politics and AfricanStates. Boulder, CO: Lynne Rienner.

Reyes, Alejandro, 1994. ‘Drug Trafficking andthe Guerrilla Movement in Colombia’, inBruce M. Bagley & William O. Walker III,eds, Drug Trafficking in the Americas. Miami,FL: North-South Center Press (121–130).

Reynal-Querol, Marta, 2002. ‘Ethnicity, PoliticalSystems, and Civil Wars’, Journal of ConflictResolution 46(1): 29–54.

Richards, Paul, 1996. Fighting for the Rain Forest:War, Youth, and Resources in Sierra Leone.Oxford: James Currey.

Robinson, Geoffrey, 1998. ‘Rawan is as RawanDoes: The Origins of Disorder in New OrderAceh’, Indonesia 66 (October): 127–156.

Ross, Michael L., 2001a. ‘Does Oil HinderDemocracy?’, World Politics 53(3):325–361.

Ross, Michael L., 2001b. Extractive Sectors and thePoor. Washington, DC: Oxfam America.

Ross, Michael L., 2002. ‘Booty Futures: Africa’sCivil Wars and the Futures Market for NaturalResources’, manuscript (http://www.polisci.ucla.edu/faculty/ross/bootyfutures.pdf ).

Ross, Michael L., 2003a. ‘Oil, Drugs andDiamonds: The Varying Role of NaturalResources in Civil War’, in Karen Ballentine& Jake Sherman, eds, The Political Economy ofArmed Conflict: Beyond Greed and Grievance.Boulder, CO: Lynne Rienner (47–70).

Ross, Michael L., 2003b. ‘Resources and Rebel-lion in Indonesia’, manuscript (http://www.indopubs.com/in3.html).

Ross, Michael L., 2003c. ‘How Does MineralWealth Affect the Poor?’, manuscript


06 ross (ds) 15/4/04 4:11 pm Page 354



(http://www.polisci.ucla.edu/faculty/ross/minpoor.pdf ).

Ross, Michael L., 2004a. ‘How Does NaturalResource Wealth Influence Civil War?Evidence from 13 Cases’, InternationalOrganization 58(1): 35–67.

Ross, Michael L., 2004b. ‘Does Taxation Lead toRepresentation?’, British Journal of PoliticalScience 34, forthcoming.

Sachs, Jeffrey D. & Andrew M. Warner, 1995.‘Natural Resource Abundance and EconomicGrowth’, Development Discussion Paper No.517a, Harvard Institute for InternationalDevelopment, Cambridge, MA (http://www.cid.harvard.edu/hiid/517.pdf ).

Sachs, Jeffrey D. & Andrew M. Warner, 1999.‘Nature Resource Intensity and EconomicGrowth’, in Jorg Mayer, Brian Chambers &Ayisha Farooq, eds, Development Policies inNatural Resource Economies. Cheltenham:Edward Elgar (13–38).

Sachs, Jeffrey D. & Andrew M. Warner, 2001.‘The Curse of Natural Resources’, EuropeanEconomic Review 45 (May): 827–838.

Sambanis, Nicholas, 2001. ‘Do Ethnic andNon-Ethnic Civil Wars Have the SameCauses? A Theoretical and Empirical Inquiry(part 1)’, Journal of Conflict Resolution 45(3):259–282.

Schloss, Miguel, 2002. ‘Transparency andGovernance in the Management of MineralWealth’, working paper, World Mines Min-istries Forum and Transparency International(http://www.transparency.org/working_papers/mining_mineral/trans_gov_mineralwealth.html).

Sherman, Jake, 2003. ‘Burma: Lessons from theCeasefires’, in Karen Ballentine & JakeSherman, eds, The Political Economy of ArmedConflict: Beyond Greed and Grievance.Boulder, CO: Lynne Rienner (225–255).

Singer, J. David & Melvin Small, 1994. ‘Corre-lates of War Project: International and CivilWar Data, 1816–1992’, Inter-UniversityConsortium for Political and Social Research,Ann Arbor, MI (http://www.icpsr.umich.edu:8080/DDI/SAMPLES/09905.xml?part = 2).

Smillie, Ian, 2002. ‘Dirty Diamonds: ArmedConflict and the Trade in Rough Diamonds’,Programme for International Co-operation

and Conflict Resolution, Fafo Report 377,March (http://www.fafo.no/pub/rapp/377/377.pdf ).

Snyder, Richard, 2003. ‘Does Lootable WealthBreed Disorder? A Political Economy ofExtraction Framework?’, paper prepared forthe Santa Fe Institute / Javeriana University‘Obstacles to Robust Negotiated Settlements’workshop, Bogota, Colombia, 29–31 May(http://www.santafe.edu/files/gems/obstaclestopeace/snyder.pdf ).

Stedman, Stephen John, 2001. ‘ImplementingPeace Agreements in Civil Wars: Lessons andRecommendations for Policymakers’, Inter-national Peace Academy and Center for Inter-national Security and Cooperation, New York, May (http://www.ipacademy.org/PDF_Reports/Pdf_Report_Implementing.pdf ).

Stijns, Jean-Philippe, 2001. ‘Natural ResourceAbundance and Human Capital Accumu-lation’, manuscript (http://econwpa.wustl.edu/eps/dev/papers/0112/0112001.pdf ).

Swanson, Philip, 2002. ‘Fuelling Conflict: TheOil Industry and Armed Conflict’, Programfor International Co-operation and ConflictResolution, Fafo Report 378, March(http://www.fafo.no/pub/rapp/378/378.pdf ).

Switzer, Jason, 2001. ‘Armed Conflict andNatural Resources: The Case of the MineralsSector’, Report No. 12, Minerals, Mining, andSustainable Development, International Insti-tute for Environment and Development, July(http://www.iied.org/mmsd/mmsd_pdfs/jason_switzer.pdf ).

Torvik, Ragnar, 2002. ‘Natural Resources, RentSeeking and Welfare’, Journal of DevelopmentEconomics 67(2): 455–470.

UN Panel of Experts, 2000. ‘Report of the Panelof Experts Appointed Pursuant to SecurityCouncil Resolution 1306, 2000. Paragraph19, in Relation to Sierra Leone’, UnitedNations Security Council (S/2000/1195), 20December (http://www.un.org/Depts/dpko/dpko/1195.pdf ).

UN Panel of Experts, 2001. ‘Report of the Panelof Experts on the Illegal Exploitation ofNatural Resources and Other Forms of Wealthof the Democratic Republic of Congo’,United Nations Security Council(S/2001/357), 12 April (http://www.unhchr.


06 ross (ds) 15/4/04 4:11 pm Page 355



ch/Huridocda/Huridoca.nsf/0/5e423385c10ae294c1256b1100505218/$FILE/N0132354.pdf ).

UN Panel of Experts, 2002. ‘Final Report of thePanel of Experts on the Illegal Exploitation ofNatural Resources and Other Forms of Wealthof the Democratic Republic of Congo’,United Nations Security Council(S/2002/1146), 16 October (http://www.un.dk/doc/S20021146.pdf ).

Wantchekon, Leonard & Ricky Lam, 1999. ‘Dic-tatorships as a Political Dutch Disease’,Economic Growth Center, Yale University,Center Discussion Paper No. 795, January(http://www.library.yale.edu/socsci/egcdp795.pdf ).

Wesley-Smith, Terence, 1992. ‘Development and

Crisis in Bougainville: A Bibliographic Essay’,Contemporary Pacific 4(2): 407–433.

Westing, Arthur H., ed., 1986. Global Resourcesand International Conflict: EnvironmentalFactors in Strategic Policy and Action. Oxford:Oxford University Press.

MICHAEL L. ROSS, b. 1961, PhD inPolitics (Princeton University, 1996);Assistant Professor, University of Michigan(1996–2001); Visiting Scholar (World Bank,2000); Assistant Professor, Department ofPolitical Science, University of California, LosAngeles (2001– ). Author of Timber Boomsand Institutional Breakdown in Southeast Asia(Cambridge University Press, 2001).


06 ross (ds) 15/4/04 4:11 pm Page 356



Journal of Peace Research - UCLA

Documents

Transcript of Journal of Peace Research - UCLA