Ethnic Minority Rule and Civil War Onset

7/28/2019 Ethnic Minority Rule and Civil War Onset

1/22

http://cmp.sagepub.com/Peace Science

Conflict Management and

http://cmp.sagepub.com/content/28/5/438The online version of this article can be found at:

DOI: 10.1177/0738894211418426

2011 28: 438Conflict Management and Peace ScienceDan Miodownik and Ravi Bhavnani

Natural Resource Profiles Moderate OutcomesEthnic Minority Rule and Civil War Onset How Identity Salience, Fiscal Policy, and

Published by:

http://www.sagepublications.com

On behalf of:

Peace Science Society (International)

can be found at:Conflict Management and Peace ScienceAdditional services and information for

http://cmp.sagepub.com/cgi/alertsEmail Alerts:

http://cmp.sagepub.com/subscriptionsSubscriptions:

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

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

http://cmp.sagepub.com/content/28/5/438.refs.htmlCitations:

What is This?

- Nov 14, 2011Version of Record>>

at UNIV OF WISCONSIN on January 10, 2012cmp.sagepub.comDownloaded from
http://cmp.sagepub.com/http://cmp.sagepub.com/http://cmp.sagepub.com/http://cmp.sagepub.com/content/28/5/438http://cmp.sagepub.com/content/28/5/438http://www.sagepublications.com/http://www.sagepublications.com/http://pss.la.psu.edu/http://cmp.sagepub.com/cgi/alertshttp://cmp.sagepub.com/cgi/alertshttp://cmp.sagepub.com/subscriptionshttp://cmp.sagepub.com/subscriptionshttp://cmp.sagepub.com/subscriptionshttp://www.sagepub.com/journalsReprints.navhttp://www.sagepub.com/journalsReprints.navhttp://www.sagepub.com/journalsPermissions.navhttp://www.sagepub.com/journalsPermissions.navhttp://www.sagepub.com/journalsPermissions.navhttp://cmp.sagepub.com/content/28/5/438.refs.htmlhttp://online.sagepub.com/site/sphelp/vorhelp.xhtmlhttp://online.sagepub.com/site/sphelp/vorhelp.xhtmlhttp://cmp.sagepub.com/content/28/5/438.full.pdfhttp://cmp.sagepub.com/content/28/5/438.full.pdfhttp://cmp.sagepub.com/http://cmp.sagepub.com/http://cmp.sagepub.com/http://online.sagepub.com/site/sphelp/vorhelp.xhtmlhttp://cmp.sagepub.com/content/28/5/438.full.pdfhttp://cmp.sagepub.com/content/28/5/438.refs.htmlhttp://www.sagepub.com/journalsPermissions.navhttp://www.sagepub.com/journalsReprints.navhttp://cmp.sagepub.com/subscriptionshttp://cmp.sagepub.com/cgi/alertshttp://pss.la.psu.edu/http://www.sagepublications.com/http://cmp.sagepub.com/content/28/5/438http://cmp.sagepub.com/


2/22

Conlict Management and Peace Science The Author(s). 2011. Reprints and permissions:http://www.sagepub.co.uk/journalsPermissions.nav

[DOI: 10.1177/0738894211418426]Vol 28(5): 438458

438

Ethnic Minority Rule and Civil War OnsetHow Identity Salience, Fiscal Policy, and

Natural Resource Proiles Moderate Outcomes

DAN MIODOWNIKDepartments o Political Science and International Relations, Hebrew University o Jerusalem

RAVI BHAVNANI*Graduate Institute o International and Development Studies

Using an agent-based computational ramework designed to explore the incidence oconict between two nominally rival ethnic groups, we demonstrate that the impacto ethnic minority rule on civil war onset couldbe more nuanced than posited in theliterature. By testing the eects o three key moderating variables on ethnic minorityrule, our analysis demonstrates that: (i) when ethnicity is assumed to be salient or allindividuals, conict onset increases with size o the minority in power, although whensalience is permitted to vary, onset decreases as minority and majority approachparity; (ii) fscal policythe spending and investment decisions o the minorityEGIPmoderates conict; conict decreases when leaders make sound decisions,increases under corrupt regimes, and peaks under ethno-nationalist regimes thatplace a premium on territorial conquest; and lastly (iii) natural resourcestheir typeand distributionaect the level o conict which is lowest in agrarian economies,higher in the presence o lootable resources, and still higher when lootable resourceare diuse. Our analysis generates a set o propositions to be tested empirically,subject to data availability.

KEYWORDS: agent-based modeling; civil war; ethnic minority rule;political exclusion

1. Introduction

Can ethnic minority rule eectively increase the risk o civil war onset? Commentingon the nature o majority and minority rule in pre-1994 Rwanda and Burundirespectively, Uvin (1999: 253, 264) notes that the dynamics that led to violence inRwanda and Burundi are textbook cases o entirely dierent processes and serve

418426CMPXXX10.1177/0738894211418426Miodownik & Bhavnani:EthnicMinority Rule and Civil WarOnsetConfict Management and Peace Science

* Author names are listed in reverse alphabetical order. Support or this work was providedby the Center or the Advanced Study o International Development at Michigan State

University, the Hebrew University Intramural Research Fund, and the Israeli ScienceFoundation (Grant #330/07).

http://cmp.sagepub.com/http://cmp.sagepub.com/http://cmp.sagepub.com/http://cmp.sagepub.com/


3/22

Miodownik & Bhavnani: Ethnic Minority Rule and Civil War Onset

439

as archetypical examples o very dierent categories o violence. In Rwanda,discontent ran high among members o the Hutu majority itsel, whereas inneighboring Burundi discontent was directed at the minority-led Tutsi regime mostvehemently by the disenranchised Hutu majority. This, according to Uvin, resulted

inundamentally dierent dynamics o violence in both countries, with the Hutu elitein Rwanda playing the ethnic card to legitimize state control among ellow Hutu,and with Burundis Tutsi elite systematically repressing the countrys Hutu majorityto maintain their stranglehold on power.

Rule by Burundis Tutsi minorityarguably more uncompromising thanmajority rule in Rwandaincreased the restiveness o the dominated majorityand the insecurity o the dominant minority, resulting in requent rebellion andbrutal repression. In marked contrast, the majority-led Hutu regime in Rwanda,preoccupied more notably by regional cleavages that served to weaken the regime,did not completely exclude Tutsi rom the social, economic, and political realm.1

Besides, Prunier (1995: 90) notes that in 1994 the Rwandan political system was onthe verge o collapse and any strong push rom outside would complete the process,whereas Burundis minority-led regime was not averse to power-sharing experiments,given its ability to intervene militarily and reassert Tutsi rule.

An examination o the ethnic composition o 48 Arican regimes in 2005 shows thatthe heads o state in 26 countries belonged to a secondary or minority ethnic group,defined respectively as 10% or more (short o a majority) and 10% or less o a countryspopulation.2 Leaders in six other countries were members o the plurality ethnicgroup, the largest group short o a 50% population share, whereas leaders in only16 countries hailed rom the ethnic majority, defined as 50% or more o the population.

Only seven armed conficts unolded in these 48 states in 2005; five in countries led byan ethnic minority (Chad, Sudan, Uganda, and Ethiopia with two conficts) and two instates led by an ethnic majority (Algeria and Burundi). With over 32% o all civil warsin the period between 1946 and 2006 occurring in Arica, and given sizeable variationin the ethno-political composition o Arican ruling regimes, it seems reasonable toquestion the connection between the ethnic identity o leadersthose hailing romethnic minorities, in particularand the incidence o civil violence.3

Clearly, examples o ethnic minority rule are not limited to Arica alone. Take thecase o the Pakistani civil war in 1971, which led to the secession o East Pakistan andthe oundation o Bangladesh. The insurgents, the Mukti Bahini or the Liberation

Army, claimed to represent the discriminated Bengali populationthe largestethnic group in Pakistan at the time with, some 44% o the populationruled bya coalition o Punjabis (33% o the population), Sindhis (7%), and Pashtuns (5%).Armed confict between the Jordanian army and various Palestinian actions in

1 We reer to the nature o majority rule in Rwanda or the duration o the Second Republic,and prior to the 1994 genocide.

2 For a description o the data and sources used see Table A1, Appendix A: http://plutomail.huji.ac.il/~miodownik/MioBhavCMPS_AppendixA.

3 This igure represents the total number o internal armed conlicts and internationalizedinternal armed conlicts as coded by the UCDP/PRIO Armed Confict dataset v.4-2007,19462006 (Gleditsch et al., 2002; Harbom and Wallensteen, 2007).



4/22

Conlict Management and Peace Science 28(5)

440

September 1970 provides yet another example. The kingdom o Jordan was ruledby members o the Hashemite amily in a coalition with loyal Bedouin tribes(constituting about 40% o the population), while Palestinians in Jordan amountedto over 58% o the population.

Addressing the link between a leaders ethnic identity and civil war onset or allcountries rom 1945 to 1999, Fearon, Kasara, and Laitin (2007) (henceorth FKL)suggest that minority leaders held power in 25.12% o all country years, albeit withsignificant cross-regional variation: lows o 7.82% and 8.14% or all country years inAsia and the West; a high o 57.19% in Latin America and the Caribbean;4 and the stillhigher figure o 58.51% in Sub-Saharan Arica. As such, FKL argue that civil wars brokeout slightly more requently under ethnic minority leadership: in 2.05% o all ethnicminority leader country years versus 1.5% o all country years with a leader rom theplurality group. The numbers are somewhat higher or Sub-Saharan and North Arica(which includes the Middle East) with 2.58% and 2.26% o all minority leader years

exhibiting the outbreak o a civil war versus 1.63% and 1.75% o all plurality leaderyears with a civil war onset, leading the authors to conclude that countries with headso state rom a minority ethnic group have been marginally more likely to have civilwars begin, although the association is weak and uncertain (p. 192).

Two recent studies by Wimmer, Cederman, and Min (2009 and henceorth WCM)and Cederman, Wimmer, and Min (2010 and henceorth CWM) question thisconclusion, given that FKLs use o a dummy variable to indicate whether the heado state is a member o a minority group is arguably a less than perect measure oinstitutionalized exclusion rom power (WCM: 325 n9).5The act that the head o statehails rom a minority need not necessarily imply that members o other ethnic groups,

including the ethnic majority, are wholly excluded rom state power. Addressing thisconcern, the authors develop a set o indicators to better capture how access to statepower (or lack thereo) may lead to the outbreak o ethno-nationalist civil war.

Using the Ethnic Power Relations (EPR) dataset, CWM suggest that ethnicconficts involving excluded groups are more requent than conficts involving groupsthat share political power.6 Specifically, they observe confict onset in 27 o the 11,622group years (0.23%) or groups holding a monopoly o power, dominant groups,

4 I Mestizo and White are coded as separate identities. Only 18.68% o country yearsare coded as controlled by a minority leader i Mestizo and White are coded as two

separate identities.

5 Speciically, WCM (p. 325) suggest that Most researchers deine exclusion narrowly, ocusingon a small number o minority rights rather than explicitly measuring access to state power.[D]ata sets that use a broader deinition o exclusion are limited in geographic scope and purelycross-sectional and thereore do not record changes in ethnic power relations over time.

6 The EPR data (http://dvn.iq.harvard.edu/dvn/dv/epr) used in WCM and CWM measuresaccess to executive-level state power or all politically relevant ethnic groups in the world rom1946 to 2005, matching group inormation with conlict episodes that reach an annual battle-death threshold o 25 people rom the Uppsala/PRIO Armed Conlicts Data Set (Gleditsch et al.,2002). The data used in these papers includes 215 incidents o armed conlicts, o which 110

were ethnic conlicts (57 with secessionist aims), 20 o which were initiated by the ethnic groupin power and the remaining 90 initiated by a marginalized ethnic group (WCM: 327, Table 1).



5/22


441

and senior or junior coalition partners (i.e. included groups), whereas confictinvolving excluded groupsgroups enjoying a modicum o regional autonomy,powerless, discriminated and separatist autonomy groupsbroke out in 119 o the17,897 group years (0.66%). Within the category o excluded groups, powerless and

discriminated groups initiated wars in 87 out o the 11,988 group years (0.73%)in comparison to groups that enjoyed some level o regional autonomy, the latteraccounting or 13 onsets in 5,433 group years (0.24%). And when groups soughtseparatist autonomy, the likelihood o confict rose urther still, accounting or 19out o 476 group years (3.99%).

Examining the WCM data, we find that ethnic minorities held power in 1023 (o atotal 7155) country years (14.3%) whereas the majority held power in 6123 o the 7155years (85.7%). Civil war broke out in 42 o the 1023 (4.1%) the minority-dominatedcountry years, and in only 173 o the 6132 (2.8%) majority-dominated country years.And ethnic civil war broke out in 25 o the 1023 (2.4%) minority-dominated country

years, and in only 85 o the 6132 (1.4%) years with the majority rule.The finding that the incidence o civil confict is marginally more likely under

ethnic minority rule in the sample as a whole also obscures the act that the eecto minority rule tends to vary. Clearly, not allethnic minority regimes exclude othergroups rom power, and even when they do, exclusion does not always lead to theoutbreak o war. As the most violent regions, Asia and Sub-Saharan Arica dierwith respect to the onset o confict under ethnic minority rule. In Asia only 6.6% oall confict and 8.6% o all ethnic confict took place during periods o minority rule,the preponderance o wars breaking out when a leader rom the plurality group wasin power. In marked contrast, 46.3% o all confict onsets and 47.5% o all ethnic

conficts in Sub-Saharan Arica occurred under ethnic minority rule. Even usingthe much higher threshold o 1000 battle-deaths, the FKL data attest to a similarpatternwith 12.9% o all confict onsets and a third o all ethnic conficts in Asia,and 70.6% o all confict onsets and 100% (10/10) o ethnic confict onsets in Sub-Saharan Arica occurring under minority rule. Table 1 summarizes the distributiono confict in each region using the WCM and FKL data.

The regional dierences outlined above lead us to question the conditions underwhich the marginalization o the ethnic majority rom state power triggers conict.For as FKL note, To believe that the coeficients on N* or minority leader providedecent estimates o causal eect. one has to believe that these variables are

uncorrelated with other, unmeasured determinants o civil war risk (p. 192).7

Itollows that our concern in this article lies less with re-testing the significance oethnic minority rule per se. Rather, using a theoretical ramework designed to explorethe incidence o confict between two nominally rival ethnic groups, we explorehow the propensity o the marginalized ethnic group (MEG)excluded majoritiesin this caseto rebel against the ethnic group in power (EGIP)minority-ledregimescould be shaped by the ollowing actors: (i) the salience members o the

7 The N* index (Cederman and Girardin, 2007) is comprised o a star-like coniguration withthe Ethnic Group in Power (EGIP) at the center surrounded by peripheral groups (E*), where

the likelihood o conlict is determined by the ability o the latter to challenge the ormer asa unction o resources (M*). Thus, N*=E*M*.



6/22


442

Table 1. Percent and Frequency o Civil War Onset under Minority and Majority Rule [% (N)]

Fearon et al. (2007) Wimmer et al. (2009)

War Onsets Ethnic Onsets War Onset Ethnic Onset

Minority Majority Minority Majority Minority Majority Minority Majority

West 0 (0) 100 (2) 0 (0) 0 (0) 0 (0) 100 (3) 0 (0) 100 (2)

Eastern

Europe

0 (0) 100 (9) 0 (0) 100 (1) 0 (0) 100 (22) 0 (0) 100 (14)

Asia 12.9 (4) 87.1 (27) 33.3 (2) 66.7 (4) 6.6 (4) 93.4 (57) 8.6 (3) 91.4 (32)

N. Arica &

Middle East

29.4 (5) 70.6 (17) 50 (1) 50 (1) 10.7 (3) 89.3 (25) 14.3 (2) 85.7 (12)

Sub-

Saharan

Arica

70.6 (24) 29.4 (10) 100 (10) 0 (0) 46.3 (31) 53.7 (36) 47.5 (19) 52.5 (21)

Latin

America &

Caribbean

60 (9) 40 (6) 0 (0) 100 (1) 11.7 (4) 88.3 (30) 20 (1) 80 (4)

Total 38.9 (42) 61.1 (66) 65 (13) 35 (7) 19.5 (42) 80.5 (173) 22.7 (25) 77.3 (85)

EGIP and MEG attach to ethnicity as a core or defining identity; (ii) the EGIPsfiscal policy; and (iii) the countrys resource base and distribution.

The article proceeds as ollows. The next section accounts or dierences in ethnicminority rule as a unction o two aggregate-level actorsfiscal policy and natural

resourcesand a third, individual-level actorethnic salience. Section 3 introducesour theoretical ramework, and section 4 presents our experiments and results.A final section discusses the implications o our findings.

2. Accounting for Differences in Minority Rule

CWM and WCM argue that the state and its role as a key actor have been largelyoverlooked by students o civil war. Most states, they note, tend to avor co-ethnicsand exclude ethnic outsiders. Rebellions break out when leaders o excluded ormarginalized groups, in an eort to increase their share o the spoils, seek to capture

the state and redistribute revenue along ethnic lines. In short, ethnicity mattersbecause the state relies on ethno-nationalist principles o political legitimacy,avoring co-ethnics to gain political supportan argument that goes a long way inclariying the institutional and configurational mechanisms likely to breed confictwhen groups are excluded rom power (WCM: 321). In spite o our agreement withthe basic premise, we express three reservations with this line o reasoning.

First, the authors do not acknowledge the possibility that leaders may, ater all, domore than promote the interests o the EGIP in the absence o a vibrant civil societythat articulates policy requirements and moderates ethnic cleavages. Absent a vibrantcivil society, it seems unreasonable to assume that the state will naturally denigrate

into an arena or ethnically preerential policies.8 Surely the degree to which politics

http://cmp.sagepub.com/http://cmp.sagepub.com/http://cmp.sagepub.com/


7/22


443

have been ramed in ethnic terms and the weight individuals accord to ethnicity as acore or defining identity matter. And even regimes that are purportedly ethnic intheir composition have been known to distribute largesse more exclusively, withinmore narrowly defined clusters, as in Rwanda and, to a lesser extent, in Burundi

where regionalism eatured more prominently than ethnicity in the distribution opower and spoils (Lemarchand, 1996).9

Second, even i we agree with the general WCM/CWM proposition that exclusionbreeds confict and the EGIP may exclusively and evenly avor its members, theability o the EGIP to distribute revenue depends critically upon its capability to

generate the same. Spending decisions, how leaders spend revenuewhether theyconsume it rivolously or spend it prudently by strengthening the military, providingsocial welare, or investing in the economysignificantly shape uture revenuestreams. Absent the revenue required to sustain the support o ones own ethnickinthe ability to avor co-ethnics and maintain legitimacythe likelihood o

challenges rom both within and across the ethnic divide is likely to rise.Third, and in a related vein, the resource endowment places a hard constraint on

all economic activity. The low economic barriers to entry that typically characterizelootable resources, like alluvial diamonds, make it easy or small-scale, artisanal minersto profit rom extraction, thereby decreasing the EGIPs ability to establish monopolycontrol over these resources. In addition, the spatial distribution o resources shapesthe EGIPs ability to exercise control over resources and generate revenue. As such,

point-source recourse distributionsdense clusters o lucrative resourcesmaymore easily be subject to exclusive state control than diuse resource distributionslocated ar rom the center(s) o state power (Le Billon, 2001, 2005). Taken together,

the resource endowment and distributionwhat we reer to as a countrys resourceprofilecomprise a budget constraint in states that rely on revenue rom primarycommodities.

The three actors we examineidentity, fiscal policy, and resource profileworkin tandem to shape the propensity o an excluded majority to rebel against theminority EGIP. The resource profile constitutes a budget constraint within whichleaders crat fiscal policy to achieve any number o goals, including but not limited togaining public support and legitimacy, maintaining security and territorial control, or

8 Two caveats are in order: irst, one needs to distinguish the bridging and bonding

unctions o civil society or social capital, given that the two may have largely diametric eectson social order; second, apart rom tests in the South Asian context which lack suicientempirical evidence, the link between social capital and social order in other contexts suchas Sub-Saharan Arica has been understudied. A recent study by Backer et al. (2010) evensuggests that particular orms o organizational membership, contrary to expectation, increaseindividual dispositions to engage in violence.

9 The predominantly Hima-Tutsi political elite that ruled Burundi rom independencein 1962 instituted predatory and divisive policies unmistakably avoring residents o thesouthern province o Bururi, rom where they hailed. Likewise, the 20-year reign o PresidentHabyarimana in neighboring Rwanda avored a small group o ethnic Hutus known as

Akazu, composed o relatives and other members rom the presidents home district oGisenyi in the northwest o the country.



8/22


444

maximizing personal revenue. Revenue may be distributed evenly across economicsectors and groups, or in a manner that avors particular sectors or groups over others,generating economic grievances that increase the likelihood o ethnic mobilization.

In the paragraphs that ollow, we urther discuss how variation in ethnic salience,

fiscal policy, and natural resource profiles may shape the relationship between ethnicminority rule and civil war onset.

Ethnic SalienceCWM and WCM explicitly account or the power dierentials among politicallyrelevant ethnic groups. Their new indexes, nevertheless, implicitly assume that thedistribution o individuals across ethnic groups determines the salience o ethnicity,or the weight that individuals place on ethnicity as a defining or core identity.One consequence o this assumption is that absent a change in relative group sizes,ethnic salience eectively remains constant across groups and, more notably, across

individuals within ethnic groups i the power configuration remains unchanged.Clearly, leaders o excluded groups may find it easier to instigate violence iethnic salience is high across group members (Brass, 1997; Tambiah, 1986, 1996;Kaperer, 1988). Yet, simply assuming this is the case significantly oversimplifiesmatters. The treatment o ethnicity as an aggregate-level measure captured by theEthno-Linguistic Fractionalization (ELF) index (Alesina and La Ferrara, 2005;Collier and Hoefer, 2004; Hegre and Sambanis, 2006), Politically Relevant EthnicGroups (PREG) index (Posner, 2004), Reynal-Querol (RQ)-index o polarization(Reynal-Querol, 2002; Esteban and Ray, 2008), N* index (Cederman and Girardin,2007), and even the Ethnic Power Relations (EPR) index (Wimmer, Cederman,

and Min, 2009; Cederman, Wimmer, and Min, 2010) thereore ails to elucidate themechanisms underpinning government or opposition support along ethnic lines.In an eort to address this limitation, we explicitly relax the assumption o fixed

salience permitting ethnic salience to vary across individualsmembers o boththe EGIP and MEGas a unction o relative revenue. As such, we subscribe to adistinctly materialist conception o ethnicity (Azam, 2001; Bates, 1974; Gurr, 1970;Hechter and Okamoto, 2001; Laitin, 1998) in keeping with our broader ocus on therevenue imperativethe undamental requirement that all leaders ace getting theincome with which to govern.10 We thereore allow or the possibility that variationin the importance placed on ethnicity as a core or defining identity aects a leaders

10 Hechter and Okamoto (2001) identiy two dierent market mechanisms or intergroupstratiication. The irst suggests that social identities emerge as a by-product o collectiveexperience, that markets produce a cultural division o labor in which individuals with distinctcultural markers cluster together. This clustering may be hierarchical, the Indian caste systembeing a prima acie example, or segmental, as with early Portuguese immigrants to the UnitedStates who concentrated in New England and dominated the local ishing industry. The secondmechanism, which resonates with our approach, suggests that contact and competition overscarce resources triggers collective action along ethnic or nationalist boundaries. Bates (1974)argues that the ormation and persistence o ethnic groups in Arica is, by and large, a result

o competition or modern beneits. And Azam (2001) suggests that ethnic capital in Arica,which ensures the provision o public services, constitutes a system o redistribution whose

http://cmp.sagepub.com/http://cmp.sagepub.com/http://cmp.sagepub.com/


9/22


445

ability to gain popular support and maintain territorial control, with predictableconsequences or revenue accumulation and confict.

Fiscal Policy

The revenue imperative lies at the core o a vast literature on the fiscal sociology ostate-building in Western Europe (Ardant, 1975; Levi, 1988; Moore, 2004; Schumpeter,1991; Skocpol, 1979; Tilly, 1990; Weber, 1978) and the rentier state in resource-richsocieties (Anderson, 1987; Beblawi and Luciani, 1987; Karl, 1997; Lowi, 2004; Mahdavy,1970; Ross, 2001; Yates, 1996). Our theoretical ramework rests upon the assumptionthat leaders require revenue to govern, and that a lack o revenue increases the risko state collapse, which, in turn, increases the risk o civil war. Given the revenueimperative, how leaders obtain and spend revenuetheir fiscal policiesaects theprobability o civil war onset. For instance, spending on administrative and coerciveinrastructure can lengthen the reach o the state, making it harder or challengers to

organize; spending on social welare may help mitigate societal grievances in the shortterm, albeit with serious long-term consequences in the absence o revenue-enhancinginvestment; and wasteul consumptioni leaders squander their income instead oinvesting in public services and the militaryis all too likely to oment grievances.

The states fiscal policy may also have more direct implications or the behavior ochallengers, given that revenue is undamental to launching and sustaining rebellionthe ability to purchase equipment, organize military activity, and pay recruits (Collierand Hoefer, 2004; Collier et al., 2004; Gates, 2002). Policies that sharply decrease theability o challengers to generate a steady fow o income or increase the opportunitycosts o rebellion are thereore likely to reduce civil war onset. Indeed, a small but

growing subset o scholarship on civil war more explicitly analyzes the role o rebelfinance (Fearon, 2004; Lujala, 2010; Thies, 2010; Weinstein, 2005).Based on the concerns elucidated above, we examine how variation in spending

and investment could infuence the eect o ethnic minority rule on civil war onset.Our approach moves beyond testing the eect oper capita incomeon confict onset, ameasure that reveals precious little about the relative income o individuals rom theEGIP or MEG, and even less about the availability o income to and its distributionbetween leaders o these nominally rival groups.11 By ocusing on fiscal policy, weexplicitly account or the possibility that calls or rebellion by leaders o the MEG

breakdown triggers political violence. Taking the logic o materialist considerations urtherstill, Laitin (1998) notes that in Estonia, ethnic-Russian parents sent their children to Estonianspeaking schools based on the expectation that knowledge o the vernacular was necessaryto complete eectively in social and economic realms. Note that while we subscribe to amaterialist conception o ethnicity, our ramework makes it possible to change the mannerin which ethnic salience is calculated.

11 The per capita income measure is used in dissimilar ways. Collier and Hoeler (2004),or instance, use mean per capita income as a proxy or orgone rebel income (i.e. incomeorgone by joining a rebellion), hypothesizing that war onset will be less likely the largerthe opportunity cost. Others, like Fearon and Laitin (2003: 7576), utilize per capita income

as a proxy or state capacity, arguing that state weakness renders insurgency more easibleand attractive due to weak local policing or inept and corrupt counterinsurgency practices.



10/22


446

may meet with apathy, even resistance, i the EGIP successully distributes goodsacross all sectors o society with little regard or ethnicity.

Resource Profle

The story does not end with fiscal policy. Leaders require income to govern, and theirchallengers require revenue to finance rebellion. Thus, the source o revenuewhatwe reer to as a states resource profileis key. Despite the progression o research onthe resourceconfict link, scholars continue to reduce natural resourcestheir types,characteristics, and distributionsto single categorical variables (e.g. oil exporterdummy). As a result, they typically ail to speciy finer-grained resource categoriessuch as alluvial or kimberlite diamondsor distinguish between dierent resourcedistributions (point source or diuse), to name just a ew o the increasingly standarddistinctions made in the literature (Auty, 2001; Le Billon, 2001, 2005; Lujala et al.,2005; Ross, 2006; Snyder and Bhavnani, 2005; Snyder, 2006).

Alluvial diamonds, or instance, are typically considered lootable. Their high valuesand low economic barriers to entry, together with their extraction by dificult-to-taxartisanal miners, impedes the ability o the state to establish monopoly control (Lujalaet al., 2005; Snyder and Bhavnani, 2005; Ross, 2006). In marked contrast, non-lootableresources such as kimberlite or deep-shat diamonds have high economic barriers toentry. In most cases, only the state is capable o urnishing the capital and technologyrequired to exploit these resources, thereby enabling monopoly control. Along similarlines, point source resource distributions, typified by extractive industries such asmining, are spatially concentrated and thereore more susceptible to monopoly controland exclusion, whereas diuse resource distributions are dispersed over a larger

area, making them less amenable to monopoly control and consequently more opento capture by revenue-seeking challengers (Le Billon, 2001, 2005).It ollows that a countrys resource endowment and distributiontaken together

to denote a resource profileare key actors to consider. Challengers may aceewer barriers in capturing and controlling resources dispersed across geographicallyremote regions o a state (Buhaug et al., 2009; Buhaug and Rd, 2006; Buhaug andLujala, 2005; Cederman et al., 2009). In contrast, the leaders ability to control andextract revenue, essential or combating challenges and maintaining public support,may be commensurately greater i resources are clustered close to centers o statepower. Likewise, the motivation or rebellion may be significantly reduced i the

resources in question are not lootable or i leaders use resource revenue to benefitwider sections o society, ethnic afiliations aside, and secure its control o the state.In sum, the relationship between ethnic exclusion and civil war onset may be

shaped, in no small measure, by the salience individuals (both members o the EGIPand MEG) aord to ethnicity as a core or defining identity, fiscal policy and the statesnatural resource profile. Specifically, we argue that the instrumentalization o theseactors through the ethnic minority in power aects the likelihood o distributionalissues being ramed in ethnic terms, the ability to nurture political support bothwithin and across the ethnic divide, and the ability to combat rebellion violentlysubject to the availability and use o revenue.



11/22


447

3. Methodology

Given a limited set o natural experiments to test the eect o minority rule oncivil war onset, together with a paucity o fine-grained data on the preerences andbehavior o individuals rom nominally rival ethnic groups, agent-based modeling(ABM) constitutes a plausible complement to, rather than a substitute or,empirically-based research that lends itsel well to counteractual analysis.12 ABMs(Bankes, 2002; Bonabeau, 2002; Conte et al., 1997) are ormal, unambiguous, thusreplicable and testable (Canessa and Riolo, 2006: 274) and provide powerul waysto develop, evaluate, and test theories by undertaking complex thought experimentsthat would be dificult to conduct in the real world, or using traditional analytictechniques (Parunak et al., 1998).

The ABM rameworkdescribed ormally in the appendix13 to this article andreerred to by the acronym REsCape (see Bhavnani et al., 2008)permits us tospeciy dierent patterns o ethnic domination given that the model is intrinsicallya game between leaders o two nominally rival ethnic groups, one o which holdspower and consequently serves as a target or takeover by the rival leadership. 14The ramework also permits us to vary ethnic salience across individuals as aunction o relative income. And in contrast to a standard optimization problemin which a leader would maximize some objective unction subject to constraints,we can vary parameter values to define a leaders strategy set. We model, in otherwords, boundedly-rational agents in a purposive eort to illustrate how fiscal policy

12

Indeed, ABM and empirical observation are well-suited to iteration: existing empiricalresearch may guide model speciication; model results may either conirm or disconirmempirical observation; when conirmatory, the model results help uncover the mechanismsand dynamics by which the observed outcomes occur; when not, they oer new directionsor empirical research, which, when completed, may be used to guide the next generation oABM. Like all modes o analysis, ABM requires researchers to select actors careully andshow restraint in the number they choose.

13 See Appendix B available at http://plutomail.huji.ac.il/~miodownik/MioBhavCMPS_AppendixB.

14 It ollows that two leaders make choices about taxation and investment only when both

exercise territorial control over one or more cells on the landscape. Given a landscape that allsentirely under the control o single leader, we enter a domestic political context in which allcitizens are represented by a single national government, in which only the EGIP leader hasthe ability to dictate national budgets, enorce taxation, and accrue revenue. Consistent withFearon and Laitin (2003), our model leads to conlict when the state (EGIP) is weak, unableto deter potential challenges by the leadership o the MEG, and unable to retain control oall its territory. Thus, while localized rebellions need not trigger a ull-blown insurgency thatspreads to all parts o the state, control o even a single cell permits a leader o the MEG tomake localized (read cell-speciic) spending and investment decisions in an eort to accruerevenue. This logic may be extended to three ethnic groups, arrayed in dierent powerconigurations (with the EGIP orming a ruling coalition with one nominally rival group, or

potentially excluding both rival groups with the latter either cooperating or competing witheach other).



12/22


448

(both optimal and sub-optimal spending and investment) aects levels o confict.15And lastly, the ramework permits us to vary the natural resource endowment anddistributionthe resource profile.

Table 2 summarizes key model parameters and value ranges. Leaders o nominallyrival ethnic groups (A and B) may adopt one o our stylizedfscal policies: benevolentrule, robbery, social welare, and territorial control; the resource base may vary romagriculture to one based on harder to extract kimberlite or artisanally extractedalluvial diamonds; peasants may belong to one o two nominally rival ethnic groups;ethnic salience may be fixedollowing the primordial notion that ethnicity is always

salient or all individualsor vary across individuals based on grievances thatresult rom disparities in income; income disparities may, in turn, be calculated overdierentper capita ranges, beginning locally with an agents own cell and increasingto cover the entire landscape; and lastly, one may seed the model to capture dierentpatterns oethnic polarization and ethnic domination.

Conceptually, and at a high level o abstraction, our model works as ollows: wedefine the model landscapeas a discrete cellular grid with fixed borders and a capitalcity located in the center. Each cell on the landscape may contain members romtwo rival ethnic groupsthe EGIP and MEG. The landscape is characterized byproduction, which alls into one o our economic sectors. Fiscal policy, defined by the

sectoral and spatial spending decisions taken by a leader, determines the amount orevenue available to garner popular support. Where such support is weak, individualsmay relocate or migrate to cells populated and/or controlled by members o theirown ethnic group. Revenue rom production is consequently used by the leader oan ethnic group to control territory and we underscore the importance o territorialcontrol in this ramework, given that control is a necessary condition or spending,investment, revenue generation, and popular support. All control is cell-specific,as is the breakdown o economic sectors, spending decisions, and peasant support.

Table 2. Summary o Model Parameters and Settings

Parameter Name Parameter Settings

Fiscal PolicyA; B benevolent rule robbery social welare territorial control

Resource Base agriculture kimberlite alluvial industrial alluvial artisanalResource

Distribution

diuse point source

Ethnic Group A B

Ethnic Salience fxed variable

Per Capita Range single cell entire landscape

Ethnic Polarization high lowEthnic Dominance majority rule contest power minority rule

A, B represent the leaders o ethnic groupsA and B respectively.

15 Other ABMs used to study civil conlict include MASON, developed by Cioi-Revilla

and Rouleau (2010) and a model o insurgency developed by Bennett (2008). Earlier eortsinclude Epsteins (2002) model o civil violence.



13/22


449

Conict, also cell-specific, emerges when the leaders o rival groups seek to controlthe same territory or cell.

At the level o social mechanisms, our model includes the ollowing: robbery leadsto a decline in economic growth, undermining popular support and weakening the

state, making it more vulnerable to capture over time; spending on coercive poweralters popular support and is essential or territorial control; and investment in theeconomy, subsumed under the notion obenevolent rule, increases the fow o revenueover time with a positive eect on popular support. Key eedback loops include theollowing: changes in revenue (relative to the revenue o nominal rivals) increase (ordecrease) the salience o ethnicity; ethnic salience aects popular support; high levelso popular support decrease the cost o control, and control has a non-monotoniceect on support (excessive control lowers support, as does weak or insecure control);when support or the leader in control o a cell is weak, individuals may exercise theoption to migrate to ethnic enclaves in an eort to find saety in numbers; migration

changes the calculus o control, and thus aects spending, investment, and supportor leaders;16 confict, which arises when leaders seek to control the same territory,alters the control o individual cells and may ultimately alter control o the state.

In sum, a leader(s) makes spending and investment decisions at each timestep;these decisions in turn generate varying amounts o revenue, aect territorial controland popular support, and determine the emergence or progression o confict whichoccurs when rival leaders vie or control o the same cell(s); confict, in turn, leadsto new patterns o control over time. We consequently underscore the notion thatpatterns o spending and investment directly infuence the instigation and spread ocivil unrest by altering popular support, the leaderships capacity to control territory,

and by implication, the ability to wage confict.Our computational model is best conceived o as an exploratory device, one

that may be used to understand key causal drivers and mechanisms underpinningthe incidence o civil war in artificial landscapes or specific real-world cases. Usingthis model, it is possible to conduct complicated thought experiments or whichempirical data would, in most cases, be dificult i not impossible to collect. Forexample, those interested in the behavior o key agents could explore how a leadersability to generate revenue through the taxation or looting o natural resourcesaects the onset and duration o civil war; those interested in understanding themicro-oundations o recruitment could ocus on the conditions under which the

opportunity costs aced by peasants or joining a rebellion increase or decrease;while those interested in studying the conditions under which minority rule alters theincidence o civil war, as we are in this article, could ormally study the consequenceso dierent distributions o salience and patterns o domination.

The next section describes our experiments and results; it is ollowed by a discussiono our findings.

16 In contrast to Lim et al.s (2007) assumption that both highly mixed and well-segregated regionsdo not engage in violence, we analyze the conditions under which nominal rivals orm ethnicenclaves and whether any resulting group segregation prevents violence. Migration thereore

serves as a non-trivial model mechanism, one that underpins the ormation o majority/minorityenclaves on the landscape, and one that we devote more attention to in other research.



14/22


450

4. Experiments and Results

Using the ABM described above, we run three experiments to illustrate how the eecto minority rule could be shaped by varying levels o ethnic salience (experiment 1),dierent fiscal policies (experiment 2), and finer-grained specifications o naturalresource profiles (experiment 3). In all three experiments (Tables 35), we recordlocalized conict onsetas the percentage o cells on the landscape that experience atleast one confict event during a model run and assume that the minority ethnic group

A holds power. Rather than coding civil war onset as 1 or the entire landscape whenany cell experiences confict, we utilize a disaggregated measure precisely becausestandard coding practices obscure the notion that confict need not aect a giventerritory uniormly; the notion that civil war is a sub-national event (Hegre et al.,2009; Lubkemann, 2005; Raleigh et al., 2009; Raleigh and Hegre, 2009; Weidmannand Kuse, 2009; Weidmann and Ward, 2008).17

Experiment 1: Group Size, Ethnic Salience, and Conflict Onset

Our results, reported in Table 3, indicate that the eect o minority rule on confictonset is sensitive to the specification o ethnic salience. When ethnicity is assumedto be fixed and highly salient or all individuals across all groups, confict onsetincreases with the size o the minority in power (nA), peaking at 38% when thegroups are close but not yet equal in size (nA= 0.45), and declining thereater. Whenethnic salience is permitted to vary as a unction o relative income, however, the

Table 3. Group Size, Ethnic Salience, and Confict Onset

Size o Group A

Confict Onset (s.d.)

Fixed Salience Variable Salience

nA= 0.15 0.25 (0.01) 0.34 (0.00)

nA= 0.25 0.26 (0.02) 0.27 (0.01)

nA= 0.35 0.28 (0.01) 0.27 (0.02)

nA= 0.45 0.38 (0.03) 0.22 (0.03)nA= 0.50 0.35 (0.05) 0.21 (0.02)

A, B represent the leaders o ethnic groupsA and B respectively. In all model runs our baseline parameter

settings are as ollows: fscal policyA, B=Robbery, Benevolent Rule; resource base =Alluvial Diamonds; resource

distribution =Point Source. We subsequently varyAs fscal policy, the resource base and distribution in

experiments 2 and 3. Results rom an ANOVA indicate that the variance in conict onset across group sizes

(nA) or each experimental condition (Fixedand VariableSalience) is larger than the within-outcome variance, or

the variance obtained when both group size and salience are held constant: FFixed= 133.852 (p< 0.0001); FVariable= 243.043 (p< 0.0001). A pair-wise comparison o means across experimental conditions, now holding group

size constant, yields the ollowing results: FnA= 15 = 125.444 (p< 0.0001); FnA= 25 = 6.871 (p< 0.05); FnA= 35 =

56.726 (p< 0.001); FnA= 45 =n.s.; FnA= 50 = 56.526 (p< 0.001).

17 These papers all rely on ACLED, the Armed Conlict Location and Event Dataset, whichprovides coordinates or the exact location, date, and other inormation pertinent to battle

events occurring during a civil war. The dataset tracks rebel activity and provides details abouttransers in territorial control, the locations o bases, and one-sided violence.



15/22


451

non-monotonic relationship between group size and war onset no longer holds.Rather, confict onset varies inversely with the size o the EGIP, peaking at 34%

when a large ethnic majority (nB= 0.85) is excluded rom power. Thus, explicitassumptions about ethnic salience at the individual levelwhether salience is fixedand significant or all individuals or permitted to vary across individualsmay havenoteworthy implications or levels o confict onset under minority rule.

Experiment 2: Group Size, Fiscal Policy, and Conflict OnsetCould fiscal policy infuence the relationship between minority rule and civil waronset? Our results, reported in Table 4, point to the absence o confict underbenevolent rulewhen the minority EGIP eectively invests in the economy, incoercive power, and engages in no thet. Under this policy, the minority regime enjoys

widespread support, has suficient resources to suppress rebellion rom co-ethnicsand nominal ethnic rivals, and eectively excludes the leadership o the MEG romrevenue accumulation.

This prudent fiscal policy stands in marked contrast to our deault policy orobberycharacterized by ineective economic managementthe absence o investment andmilitary spendinghigh levels o thet, and a paucity o resources to secure peasantsupport and suppress rebellion. Under these conditions, confict onset peaks at 34%when nA= 0.15, driven largely by the act that the kleptocratic minority regime isunable to accrue any revenue until its population share reaches 35%, beyond whichpointand despite greater revenue retention by the minority EGIPlevels o

confict onset all to only 21% as the two groups approach parity.

Table 4. Group Size, Fiscal Policy, and Confict Onset

Size o

Group A


Benevolent Rule Robbery Territorial Control

RevenueConfict

Onset (s.d.)

RevenueConfict

Onset (s.d.)

RevenueConfict

Onset (s.d.)A B A B A B

nA= 0.15 1024.86 0.00 0.00 (0.00) 0.00 914.80 0.34 (0.00) 0.00 906.25 0.47 (0.03)

nA= 0.25 1041.30 0.00 0.00 (0.00) 0.00 921.94 0.27 (0.01) 744.54 1.69 0.56 (0.05)

nA= 0.35 1040.30 0.00 0.00 (0.00) 21.89 885.42 0.27 (0.01) 798.93 0.13 0.48 (0.05)

nA= 0.45 1061.88 0.00 0.00 (0.00) 367.24 628.07 0.22 (0.03) 853.32 0.15 0.38 (0.06)nA= 0.50 1056.47 0.00 0.00 (0.00) 504.28 532.66 0.21 (0.02) 917.48 0.00 0.35 (0.03)

A, B represent the leaders o ethnic groupsA and B respectively, changes in fscal policy apply toA alone,

and Revenue denotes the mean value at t= 2000. Results rom an ANOVA indicate that the variance inconict onset across group sizes (nA) or each experimental condition (Robbery (rb) and Territorial Control (tc))

is larger than the within-outcome variance, the variance obtained when both group size and salience are

held constant: Frb= 243.043 (p< 0.0001); Ftc= 98.592 (p< 0.0001). A pair-wise comparison o means across

experimental conditions, now holding group size constant, yields the ollowing results: Frb,tc,nA=15= 778.7 (p


16/22


452

And as evidenced by the even higher level o confict onset (peaking at 56%when nA= 0.25), a minority regime playing the ethno-nationalist card is likely tobe involved in more confict than a minority EGIP that contends itsel with sel-enrichment (as in the case o robbery). Thus, when the primary objective o the

minority EGIP is territorial controlcharacterized by a high level o spending oncoercive power and a moderate level o investment in the economywe find thatbeyond a population share o 15% the regime in power retains the bulwark o allrevenue. We do, nonetheless, observe a moderate decline in the likelihood o confictto 35%, as the size o the minority increases and the groups approach parity.18

Experiment 3: Group Size, Resource Base, and Conflict OnsetThe results o our final experiment, reported in Table 5, suggest that confict onset underminority rule varies both with the natural resource endowment and its distribution.In an agrarian economy, the size o the minority in power has a non-monotonic

eect on confict onset: onset equals 22% when the minority comprises 25% o thepopulation, rises to 31% when the minority comprises 35% o the population, andalls again to 27% when the minority comprises 45% o the population. This standsin marked contrast to the secular decline in confict onset in our baseline economy(characterized by alluvial diamonds and a point-source distribution) as the size othe minority increases.

Holding our baseline resource type constant and varying the distribution rompoint-source to diuse suggests that the location o resources significantly increasesthe level o confict onset or almost all values onA< 0.45 (rom 27% to 40% whennA= 0.25; rom 27% to 35% when nA= 0.35), beyond which confict levels remain

largely unchanged. And lastly, we find that the resource distribution has importantimplications or revenue: that the minority EGIPs ability to capture revenue romalluvial diamonds is curtailed when the resource distribution is diuse, even whenthe minority is sizeable (nA= 0.45).

Results rom these experiments demonstrate that the impact o ethnic minorityrule on the incidence o civil confict could be ar more nuanced than posited byrecent scholarship: (i) the relationship is sensitive to the specification o ethnicsaliencewhen ethnicity is assumed to be equally salient or all individuals, confict

18 We ran a sensitivity analysis o the results in Table 4, repeating the experiment using

agriculture as the resource base (see Table A2, Appendix A: http://plutomail.huji.ac.il/~miodownik/MioBhavCMPS_AppendixA). Our results are in line with patterns reportedin the article. With respect to revenue, the EGIP retains all revenue under benevolent rule,its total share generally increasing with nA. Robbery results in the majority o all revenuegoing to the MEG, decreasing with nA. And territorial controlresults in the lions share orevenue going to the EGIP, again increasing with nA. With respect to conlict, benevolent ruleis characterized by the absence o conlict onset, with higher levels o conlict under robberyand still higher levels under territorial control. Comparing the results o Tables 4 and A2, weind that in an agricultural, relative to an alluvial diamond economy: total revenue accrued bythe EGIP is lower (with the exception o one cell, robbery, nA= 0.15); under robbery, conlictis less likely when nA< 0.35, and more likely when nA>= 0.35 and the division o revenue is

skewed in avor o the MEG as with nA= 0.50; and under territorial controlwe observe higherlevels o conlict onset.



17/22


453

onset increases with size o the minority EGIP, peaking when the rival groupsare close but not yet equal in size; and when ethnic salience is permitted to varyacross individuals as a unction o relative income, confict onset decreases as thesize o the minority EGIP approaches that o the majority MEG; (ii) variation infiscal policythe regimes spending and investment decisionsshapes the eecto minority rule on civil war onset, eliminating war risk entirely under benevolentrule, increasing it under a corrupt or kleptocratic regime that engages in robbery,and increasing this risk still urther under an ethno-nationalist regime that seeks

to expand territorial control; and (iii) the natural resource endowment and itsdistribution have implications or levels o confict onset, directly aecting the abilityo the minority EGIP to capture resource rents or revenue.

5. Discussion

Can ethnic minority rule eectively increase the risk o civil war onset? We respondto this question with a qualified yes based on our analysis, which demonstrates howthe eect o minority rule on the incidence o civil war could be shaped by dierentspecifications o ethnic salience, variation in fiscal policy, and changes in natural

resource profiles. In short, we expect these actors to aect both the nature o minorityrule and the likelihood o confict onset, to eectively conound the relationship.

Table 5. Group Size, Resources, and Conlict Onset

Size o

Group A


Agriculture

Alluvial Diamonds

Point Source Alluvial DiamondsDiuse

RevenueConfict

Onset (s.d.)

RevenueConfict

Onset (s.d.)

RevenueConfict

Onset (s.d.)A B A B A B

nA= 0.15 0.00 924.09 0.25 (0.00) 0.00 914.80 0.34 (0.00) 0.00 915.29 0.32 (0.04)

nA= 0.25 0.00 893.43 0.22 (0.00) 0.00 921.94 0.27 (0.01) 0.00 884.34 0.40 (0.00)

nA= 0.35 0.00 884.06 0.31 (0.01) 21.89 885.42 0.27 (0.01) 0.00 881.91 0.35 (0.05)

nA= 0.45 0.00 888.60 0.27 (0.00) 367.24 628.07 0.22 (0.03) 0.00 876.54 0.22 (0.00)nA= 0.50 169.65 613.26 0.24 (0.01) 504.28 532.66 0.21 (0.02) 17.32 849.63 0.20 (0.01)

A, B represent the leaders o ethnic groupsA and B respectively, whereas Revenue denotes the mean value at

t= 2000. Results rom an ANOVA indicate that the variance in conict onset across group sizes ( nA) or two o

the three experimental conditions (Agriculture (ag) andAlluvial Diamonds-Diuse (ad)) is larger than the within-

outcome variance, the variance obtained when both group size and salience are held constant: Fag= 1264.48

(p < 0.0001); Fap= 243.04 (p < 0.0001); Fad= 272.55 (p < 0.0001). A pair-wise comparison o means across

experimental conditions (withAlluvial Diamonds-Point Source given by ap), now holding group size constant,

yields the ollowing results: (1) Fag,ap, nA=15= n.s; Fag,ap, nA=25= 527.15 (p< 0.0001); Fag,ap, nA=35= 152.41 (p< 0.0001);

Fag,ap, nA=45= 90.79 (p< 0.0001); Fag, ap nA=50= 76.40 (p< 0.0001); (2) Fad, ap,nA=15= 7.654 (p< 0.01); Fad,ap,nA=25= 3265.3

(p< 0.0001); Fad,ap,nA=35= 83.03 (p< 0.0001);Fad, ap,nA=45= n.s.; Fad,ap,nA=50= 13.16 (p< 0.01); (3) Fag,ad,nA=15= 86.87

(p< 0.0001);Fag, ad,nA=25= n.s; Fag, ad,nA=35= 16.44 (p< 0.0001);Fag, ad,nA=45= n.s; Fag,ad,nA=50= 214.58 (p< 0.0001).



18/22


454

We believe that minority rule is an important and, until recently, understudiedactor in the civil war literature. Measures o ethnic polarizationto take oneprominently used proxy or the catch-all category o ethnicity in recent scholarshipon civil warpay ar too little attention to the structure o power relations between

ethnic groups.19 Stated alternatively, we note that even measures coding minorityrule at the elite or aggregate level show significant dierences in the conditions thatgive rise to and later characterize violence in civil wars.

In an eort to elucidate how the relationship between minority rule and confictmay be conditioned by other, largely unmeasured determinants o civil war risk(FKL: 192) we conduct a set o exploratory agent-based experiments. Our experimentsexamine how variation in ethnic salience, government fiscal policy, and the resourceprofile aect the likelihood o confict when a large share o the population is excludedrom power. Some patterns observed in these experiments coincide with empiricallygrounded theoretical expectations, while others suggest hypotheses that require

empirical confirmation, subject to data availability.To recapitulate, our experiments yield three central findings. The first experiment

demonstrate that the eect o minority rule is highly sensitive to the specification oethnic salience, with confict onset increasing with size o the minority EGIP whensalience is fixed and decreasing when salience is permitted to vary. The ormer findingsuggests that the risk o violence rises with the level o ethnic polarization (Montalvoand Reynal-Querol, 2005), whereas the latter suggests just the opposite: the largerthe minority in power, the greater its ability to buy out the support o co-ethnicsand members o excluded groups in an eort to secure control. A larger minorityalso implies a smaller excluded majority, though this depends more specifically upon

the number and political relevance o ethnic groups within a state.A second experiment addresses the relationship between fiscal policy and confict.

In particular, how revenue is spent clearly aects the well-being o the populationat large, members o both the minority and majority ethnic groups. Even minority-led regimes that spend prudently, examples o what we reer to as benevolent rule,enjoy widespread support across the ethnic divide, whereas regimes that engage inhigh levels o thet tend to experience greater levels o violent opposition, acingchallenges that increase proportionately with the size o the excluded majority. Andwhen the primary objective o the minority EGIP is territorial controlcharacterizedby high levels o military spendingthe level o confict rises higher still, relative to

that in minority-led regimes characterized by uncontrolled sel-enrichment.Our third experiment suggests that the specification o finer-grained resourcecategories clearly infuences the resourceconfict relationship. As such, minority-ledregimes that depend on lootable natural resources are more prone to confict thansimilarly structured regimes that depend primarily on agricultural revenue. In addition,the likelihood o confict is likely to be higher i resource distributions are diuse,bringing into sharper ocus the ability o leaders to capture resource rents or revenue.

A logical next step involves subjecting these findings to empirical scrutiny. Large-Ndatasets ace problems o over-aggregation with respect to the ethnicity variable,

19

See the special issue o the Journal o Peace Research 45(2) (2008) on polarization andconfict, edited by Esteban and Schneider, and Cederman and Girardin (2007).



19/22


455

although surveys like theArobarometernow speciy measures o ethnic salience atthe individual level (also see work by Eiert et al., 2007; Bratton et al., 2010). Withrespect to fiscal policy, developing less aggregated, time-variant measures o spending,investment, and taxation would shed light on the conditions under whichminority-rule

is likely to be challenged by excluded majorities on economic grounds. And finally, theincreasingly widespread use o disaggregated data on the location o resources, theircharacteristics and control over time, together with the use o geographic inormationsystems, moves us even closer towards nuanced micro-level theories.

We duly acknowledge that our contribution to recent research assessing theconfict probability o marginalized ethnic groups is largely theoretical and remainssuggestive at best. Our exploratory analysis nonetheless calls or richer theorizingabout the causes and consequences o civil wara plea, i you will, or scholarsto continue breaking new ground, to explore the complexity o confict, civil warbeing no exception to complex causes and dynamics, and to find novel ways to study

precisely those questions and issues that evade rigorous empirical analysis bothat the macro- and micro-levels.

References

Alesina, Alberto F., and Eliana La Ferrara. 2005. Ethnic diversity and economic perormance.Journal o Economic Literature 43(3): 762800.

Anderson, Lisa. 1987. The state in the Middle East and North Arica. Comparative Politics20(1): 118.

Ardant, Gabriel. 1975. Financial policy and economic inrastructure o modern states andnations. In The Formation o National States in Western Europe, ed. Charles Tilly, pp.164242. Princeton, NJ: Princeton University Press.

Auty, Richard M. 2001. Resource Abundance and Economic Development. Oxord: OxordUniversity Press.

Azam Jean-Paul. 2001. The redistributive state and conficts in Arica. Journal o PeaceResearch 38(4): 429444.

Backer, D., R. Bhavnani, and C. Bodea. 2010. Does associational membership infuencepolitical violence in Sub-Saharan Arica? Working paper.

Bankes, Steven. 2002. Tools and techniques or developing policies or complex and uncertainsystems. Proceedings o the National Academy o Sciences o the United States o America99: 72637266.

Bates, Robert. 1974. Ethnic competition and modernization in contemporary Arica.

Comparative Political Studies 6(4): 457484.Beblawi, Hazem, and Giacomo Luciani, eds. 1987. The Rentier State. London: Croom Helm.Bennett, Scott. 2008. Governments, civilians, and the evolution o insurgency: Modeling the

early dynamics o insurgencies.Journal o Artifcial Societies and Social Simulation 11(4):http://jasss.soc.surrey.ac.uk/11/4/7.html.

Bhavnani, Ravi, Dan Miodownik, and Jonas Nart. 2008. REsCape: An agent-based rameworkor modeling resources, ethnicity, and confict.Journal o Artifcial Societies and SocialSimulation 11(2): http://jasss.soc.surrey.ac.uk/11/2/7.html.

Bonabeau, Eric. 2002. Agent-based modeling: Methods and techniques or simulating humansystems. Proceedings o the National Academy o Sciences o the United States o America99: 72807287.

Brass, Paul. 1997. Thet o an Idol. Princeton, NJ: Princeton University Press.



20/22


456

Bratton, Michael, Ravi Bhavnani, and Tse Hsin Chen. 2011. Voting intentions in Arica: Ethnic,economic or partisan?Arobarometer Working Paperno. 27.

Buhaug, Halvard, and Pivi Lujala. 2005. Accounting or scale: Measuring geography inquantitative studies o civil war. Political Geography 24(4): 399394.

Buhaug, Halvard, and Jan Ketil Rd. 2006. Local determinants o Arican civil wars, 19702001.Political Geography 25(3): 315335.

Buhaug, Halvard, Scott Gates, and Pivi Lujala. 2009. Geography, rebel capability, and theduration o civil confict.Journal o Conict Resolution 53(4): 544569.

Canessa, Enrique, and Rick L. Riolo (2006) An agent-based model o the impact o computer-mediated communication on organizational culture and perormance: An example o theapplication o complex systems analysis tools to the study o CIS.Journal o InormationTechnology 21(4): 272283.

Cederman, Lars-Erik, and Luc Girardin. 2007. Beyond ractionalization: Mapping ethnicityonto nationalist insurgencies.American Political Science Review 101(1): 173185.

Cederman, Lars-Erik, Halvard Buhaug, and Jan Ketil Rd. 2009. Ethno-nationalist dyads andcivil war: A GIS-based analysis.Journal o Conict Resolution 53(4): 496525.

Cederman Lars-Erik, Andreas Wimmer, and Brian Min. 2010. Why do ethnic groups rebel?New data and analysis. World Politics 61(1): 87119.

Ciofi-Revilla, Claudio, and Mark Rouleau. 2010. MASON RebeLand: An agent-based modelo politics, environment, and insurgency.International Studies Review 12(1): 3152.

Collier, Paul, and Anke Hoefer. 2004. Greed and grievance in civil war. Oxord EconomicPapers 56(4): 563595.

Collier, Paul, Anke Hoefer, and Mns Sderbom. 2004. On the duration o civil war.Journalo Peace Research 41(3): 253273.

Conte, Rosia, Rainer Hegselmann, and Pietro Terna, eds. 1997. Simulating Social Phenomena(Lecture Notes in Economics and Mathematical Systems). Berlin: Springer-Verlag.

Eiert, Ben, Miguel Edward, and Daniel N. Posner. 2007. Political sources o ethnicidentification. Paper presented at the conerence on Micro-Foundations and Mass Politicsin Arica, Michigan State University, 1213 May.

Epstein, Joshua. 2002. Modeling civil violence: An agent-based computational approach.Proceedings o the National Academy o Sciences 99(3): 72437250.

Esteban, Joan, and Debraj Ray. 2008. Polarization, ractionalization and confict.Journal oPeace Research 45(2): 163182.

Esteban, Joan, and Gerald Schneider. 2008. Polarization and confict: Theoretical and empiricalissues. Introduction to special issue.Journal o Peace Research 45(2): 131141.

Fearon, James D. 2004. Why do some civil wars last so much longer than others? Journal oPeace Research 41(3): 275302.

Fearon, James D., and David D. Laitin. 2003. Ethnicity, insurgency, and civil war.AmericanPolitical Science Review 97(1): 7590.Fearon, James D., Kimuli Kasar, and David D. Laitin. 2007. Ethnic minority rule and civil war

onset.American Political Science Review 101(1): 187193.Gates. Scott. 2002. Recruitment and allegiance: The microoundations o rebellion.Journal

o Conict Resolution 46(1): 111130.Gleditsch, Nils P., Peter Wallensteen, Mikael Eriksson, Margareta Sollenberg, and Hvard

Strand. 2002. Armed Confict 19462001: A new dataset.Journal o Peace Research 39(5):615637.

Gurr, Robert. 1970. Why Men Rebel. Princeton, NJ: Princeton University Press.Harbom, Lotta, and Peter Wallensteen. 2007. Armed confict, 19892006.Journal o Peace

Research 44(5): 623634.



21/22


457

Hechter, Michael, and Dina Okamoto. 2001. Political consequences o minority groupormation.Annual Review o Political Science 4: 189215.

Hegre, Hvard, and Nicholas Sambanis. 2006. Sensitivity analysis o the empirical literatureon civil war onset.Journal o Conict Resolution 50(4): 508535.

Hegre, Hvard, Gudren Ostby, and Clionadh Raleigh. 2009. Poverty and civil war events:A disaggregated study o Liberia.Journal o Conict Resolution 53(4): 598623.

Kaperer, Bruce. 1988. Legends o People/Myths o State: Violence, Intolerance, and PoliticalCulture in Sri Lanka and Australia. Washington, DC: Smithsonian Institution Press.

Karl, Terry Lynn. 1997. The Paradox o Plenty: Oil Booms and Petro-States. Berkeley, CA:University o Caliornia Press.

Laitin, David. 1998.Identity in Formation: The Russian-Speaking Populations in the NearAbroad. Ithaca, NY: Cornell University Press.

Le Billon, Philippe. 2001. The political ecology o war: Natural resources and armed conficts.Political Geography 20(5): 561584.

Le Billon, Philippe. 2005. Fuelling War: Natural Resources and Armed Conicts. New York:Routledge.

Lemarchand, Rene. 1996. Burundi: Ethnic Conict and Genocide. Cambridge, MA: WoodrowWilson Center Press.

Levi, Margaret. 1988. O Rule and Revenue. Berkeley, CA: University o Caliornia Press.Lim, May, Richard Metzler, and Yaneer Bar Yam. 2007. Global pattern ormation and ethnic/

cultural violence. Science 14(317): 15401544.Lowi, Miriam. 2004. Oil rents and political breakdown in patrimonial states: Algeria in

comparative perspective.Journal o North Arican Studies 9: 83102.Lubkemann, Stephen C. 2005. Migratory coping in wartime Mozambique: An anthropology o

violence and displacement in ragmented wars.Journal o Peace Research 42(4): 493508.Lujala, Pivi. 2010. The spoils o nature: Armed civil confict and rebel access to natural

resources.Journal o Peace Research

47(1): 1548.Lujala, Pivi, Nils Petter Gleditsch, and Elisabeth Gilmore. 2005. A diamond curse? Civil warand a lootable resource.Journal o Conict Resolution 49(4): 538562.

Mahdavy, Hossein. 1970. The pattern and problems o economic development in rentier states:The case o Iran. In Studies in the Economic History o the Middle East, ed. M. A. Cook,pp. 428467. Oxord: Oxord University Press.

Montalvo, Jos G., and Marta Reynal-Querol. 2005. Ethnic polarization, potential confict,and civil wars.American Economic Review 95(3): 796816.

Moore, Mick. 2004. Revenues, state ormation, and the quality o governance in developingcountries.International Political Science Review 25(3): 297319.

Parunak, Van Dyke, Robert Savit, and Rick Riolo. 1998. Agent-based modeling vs. equation-

based modeling: A case study and users guide. In Proceedings o Multi-Agent Systemsand Agent-Based Simulation (MABS 98), eds Jaime S. Sichman, Rosaria Conte,and NigelGilbert, pp. 1025. Berlin: Springer.

Posner, Daniel N. 2004. Measuring ethnic ractionalization in Arica.American Journal oPolitical Science 48(4): 849863.

Prunier, Gerard. 1995. The Rwanda Crisis: History o a Genocide. New York: ColumbiaUniversity Press.

Raleigh, Clionadh, and Hvard Hegre. 2009. Population size, concentration and civil war:A geographically disaggregated analysis. Political Geography 28(4): 224238.

Raleigh, Clionadh, Andrew Linke, and Hvard Hegre. 2010. Introducing ACLED: An ArmedConfict Location and Event Dataset.Journal o Peace Research 47(5): 651660.

Reynal-Querol, Marta. 2002. Ethnicity, political systems, and civil wars.Journal o ConictResolution 46(1): 2954.



22/22


458

Ross, Michael. 2001. Does oil hinder democracy? World Politics 53(3): 325361.Ross, Michael. 2006. A closer look at oil, diamonds, and civil war.Annual Review o Political

Science 9: 265300.Schumpeter, Joseph. 1991. The crisis o the tax state. InJoseph A. Schumpeter: The Economics

and Sociology o Capitalism, ed. Richard Swedberg, pp. 99140. Princeton, NJ: PrincetonUniversity Press.

Skocpol, Theda. 1979. States and Social Revolutions: A Comparative Analysis o France, Russiaand China. New York: Cambridge University Press.

Snyder, Richard. 2006. Does lootable wealth breed disorder? A political economy o extractionramework. Comparative Political Studies 39(8): 943968.

Snyder, Richard, and Ravi Bhavnani. 2005. Blood, diamonds, and taxes: Lootable wealthand political order in Sub-Saharan Arica.Journal o Conict Resolution 49(4): 563597.

Tambiah, Stanley. 1986. Sri Lanka: Ethnic Fratricide and the Dismantling o Democracy.Chicago, IL: University o Chicago Press.

Tambiah, Stanley. 1996. Leveling Crowds: Ethnonationalist Conicts and Collective Violencein South Asia. Berkeley, CA: University o Caliornia Press.

Thies, Cameron G. 2010. O rulers, rebels, and revenue: State capacity, civil war onset, andprimary commodities.Journal o Peace Research 47(3): 321332.

Tilly, Charles. 1990. Coercion, Capital, and European States, AD 9901992.Cambridge:Blackwell.

Uvin, Peter. 1999. Ethnicity and power in Burundi and Rwanda: Dierent paths to massviolence. Comparative Politics 31(3): 253271.

Weber, Max. 1978. Economy and Society, ed. Guenther Roth and Claus Wittich. Berkeley,CA: University o Caliornia Press.

Weidmann, Nils B., and Doreen Kuse. 2009. WarViews: Visualizing and animating geographicdata on civil war.International Studies Perspectives 10(1): 3648.

Weidmann, Nils, and Michael D. Ward. 2008. Spatialtemporal modeling o civil war: Theexample o Bosnia. Paper prepared or the GROW-Net Conerence, Zurich.Weinstein, Jeremy M. 2005. Resources and the inormation problem in rebel recruitment.

Journal o Conict Resolution 49(4): 598624.Wimmer Andreas, Lars-Erik Cederman, and Brian Min. 2009. Ethnic politics and armed

confict: A configurational analysis o a new global data set.American Sociological Review74(2): 316337.

Yates, Douglas. 1996. The Rentier State in Arica: Oil Rent Dependency and Neocolonialismin the Republic o Gabon. Trenton, NJ: Arica World Press.

DAN MIODOWNIK isAssociate Proessor (Senior Lecturer) o Political Science and International

Relations at the Hebrew University o Jerusalem, Israel. His research examines the emergence, unold-ing and regulation o anti-regime mobilization, protest behavior, ethnic polarization, and civil wars.

He has signiicant interest in the development o computer simulationsagent-based modeling in

particularto assist comparative political analysis o these and other complex social phenomena.

RAVI BHAVNANI is Associate Proessor o Political Science at the Graduate Institute o International

and Development Studies in Geneva, Switzerland. His research explores the micro-level dynamics

o participation in civil violence. Relying on agent-based computational modeling and disaggregated

empirical analysis, his work underscores the endogenous relationships among the diverse charac-

teristics, belies, and interests o relevant actors; social mechanisms and emergent social structures

that shape attitudes, decision-making and behavior; and patterns o violence.

Ethnic Minority Rule and Civil War Onset

Documents

Transcript of Ethnic Minority Rule and Civil War Onset