5 dec III

23
Democratizing with Ethnic Divisions: A Source of Conflict? Author(s): Demet Yalcin Mousseau Reviewed work(s): Source: Journal of Peace Research, Vol. 38, No. 5 (Sep., 2001), pp. 547-567 Published by: Sage Publications, Ltd. Stable URL: http://www.jstor.org/stable/424775 . Accessed: 02/12/2011 12:13 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. Sage Publications, Ltd. is collaborating with JSTOR to digitize, preserve and extend access to Journal of Peace  Research. http://www.jstor.org

Transcript of 5 dec III

Page 1: 5 dec III

8/3/2019 5 dec III

http://slidepdf.com/reader/full/5-dec-iii 1/22

Democratizing with Ethnic Divisions: A Source of Conflict?Author(s): Demet Yalcin MousseauReviewed work(s):Source: Journal of Peace Research, Vol. 38, No. 5 (Sep., 2001), pp. 547-567Published by: Sage Publications, Ltd.Stable URL: http://www.jstor.org/stable/424775 .

Accessed: 02/12/2011 12:13

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of 

content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms

of scholarship. For more information about JSTOR, please contact [email protected].

Sage Publications, Ltd. is collaborating with JSTOR to digitize, preserve and extend access to Journal of Peace

 Research.

http://www.jstor.org

Page 2: 5 dec III

8/3/2019 5 dec III

http://slidepdf.com/reader/full/5-dec-iii 2/22

? 2001 Journal fPeaceResearch,vol.38, no.5, 2001, pp. 547-567

SagePublicationsLondon,ThousandOaks,Cl andNewDelhi)

[0022-3433(200109)38:5;547-567; 019465]

Democratizing with Ethnic Divisions: A Source of

Conflict?*

DEMET YALCIN MOUSSEAU

Department of International Relations,Kof University

This article investigatesthe conditions that are conducive to extremepolitical violence in ethnically

heterogeneousnations. Theories of resourcemobilization, ethnic competition, and split labormarket

propose that democratizationand economic modernizationencourageethnic competition, increasing

the likelihoodof extremepoliticalviolence within nationsexperiencingpoliticaland economic change.In the light of these theories, the conditions that possibly foster conflict in multi-ethnic nations are

identified with respectto levels of democracy,politicalchange(or democratization),and levelsof econ-

omic development.The effectsof these variableson extremepoliticalviolence areexaminedwith several

logit regressionanalyseson a pooled time-seriessampleof 126 nations from 1948 to 1982. The find-

ingsshow that ethnic heterogeneity s not associatedwith higher evelsof violence within nations,exceptunder certainpolitical conditions. Both democracyand economic developmentrelate to politicalvio-

lence in a curvilinearinverted U-shape form. For ethnically heterogeneoussocieties, however, the

inverted U-curve for democracyis asymmetric,with democracy'spacifying impact relative to semi-

democraciesonly about half aspotent as in ethnicallyhomogeneoussocieties and less than that of strict

autocracy.

Introduction

Do ethnicdivisions n a nationinevitablyead

to extreme evelsof politicalviolence?If not,

what are the factors that fosterpoliticalvio-

lence in multi-ethnic societies?While pri-mordialists attribute the politicization of

ethnic groups for violence to primordialcharacteristics f ethnicity,and consider the

existence of ethnic divisionsin a nation as a

sufficientconditionfor the occurrence f vio-lence, instrumentalistssuggest that ethnic

groupsareusually tablebut mobilizeat times

with the impact of certain environmental

* I am gratefulto StuartBremer,FrankCohen, Scott

Gates, Havard Hegre, Richard Hofferbert, Michael

Mousseau,RichardTucker,ackSnyder,EduardA. Ziegen-

hagen,and thereviewerstJPR or theirhelpfulcommentsandinsights.The datausedin this articlecanbe found at

http://www.prio.noljpr/datasets.asp.

conditions. Whether political violence in

multi-ethnicnations s primordial r environ-

mental is an importantquestionconsideringthatthe intensityof ethnicity-relatedoliticalviolencearound heworldhasbeenincreasingsincethe 1970s (Gurr& Harff,1994).

The purpose of this study is thus to

investigate the possibility of a conditional

(interactive)relationshipbetween ethnichet-

erogeneitynd politicaland economic factors

as they affect levels of political violence in

nations. Consistent with previous research,

both democracyand economic developmentareexpectedto relateto politicalviolence in

an invertedU-shapeform. In corollary, t is

also expected that democratizing nations

have higher levels of extreme political vio-

lence than nationswith stablepoliticalinsti-

tutions. However, as the instrumentalist

547

3^^^^^l

.pd'.?",:iPq

Page 3: 5 dec III

8/3/2019 5 dec III

http://slidepdf.com/reader/full/5-dec-iii 3/22

548 journal of PEACE RESEARCH

ethnic politicsliterature uggests,democracyand economic development might increase

theprobabilityof extremepoliticalviolence if

multipleethnicgroupsexist within a nation.

These hypothesesaretestedusingmaximumlikelihood estimationtechniqueson a pooled

time-seriessample for 126 nations for each

yearduringthe period1948-82.

In the following section, I brieflyreview

previous esearch n politicalviolence.Follow-

ing this,I describe he theoreticalxpectationsand the hypothesesto be tested.The article

then proceeds with the operationalizations,

measures,and data sourcesof the dependent

andtheindependent ariables.n thefinalsec-tions, I reporton the empirical indingsand

drawconclusionson the implicationsof this

studyfordomesticconflictstudies.

Researchon PoliticalViolence

In analyzingdomesticpoliticalviolence,most

empiricalcross-nationalstudies hypothesizeandtestthe existenceof ethnicheterogeneity,autocraticgovernance,and economic under-

development (or peripheral status in the

world system)as contributoryfactorsto the

occurrence of political violence within

nations (Boswell & Dixon, 1990; Brown &

Boswell, 1997; Hibbs, 1973; Muller, 1985;

Muller & Seligson, 1987; Schock, 1996;

Sigelman & Simpson, 1977; Weede, 1981).

These studies usuallyfocus on certainindi-

cators of relativedeprivation and resource

mobilization, such as income inequality,

economic development, and regime repres-siveness, as facilitating conditions for the

occurrenceof group grievances see Boswell

& Dixon, 1990; Muller, 1985). Employing

cross-nationaldesigns, they are particularly

concernedwith the short-term ffectsof these

political and economic variableson political

violence. Boswell & Dixon (1990), for

example,assertthat such structural ariables

as income inequality and regimerepressive-

nesshave a 'relativelynstantaneous ffect'on

politicalviolence, so they can be tested for

short-termperiods.They thereforeuse cross-

nationaldesigns o test thesevariables, egard-less of thevariationacross ime.1

Another stream of research n the sameliterature concentrates on the impact of

democracy and economic development on

variousforms of domestic politicalconflict,

controllingfor ethnicheterogeneityasa con-

tributoryfactorto politicalviolence. Studies

in this streamsuggest that the more demo-

cratic or economicallydeveloped a nation,

the lesslikelyit will experienceviolent forms

of political conflict. Most of these studies

produce supportive empirical findings forthishypothesis Flanigan& Fogelman,1970;

Rummel,1997; Sigelman& Simpson, 1977;

Weede, 1981; Ziegenhagen,1994).

However,what is less well understoodin

the political violence literature s, first, the

long-term mpactof thevariation n different

levelsof democracy,andchangen these levels

(the processof democratization), n extreme

political violence, and, second, how these

macro-structural actors affect the incidence

and intensity of violence in multi-ethnic

societies.Accordingly,n this articleI investi-

gate the relationshipbetween ethnic hetero-

geneity and political violence, with several

conditional (interactive) hypotheses that

specifythis relationshipwith respectto vari-

ation in these structural onditions. Below,I

outline severalof these varying conditions

and consequent hypotheses, starting with

two competing views in the ethnicity litera-

ture:theprimordialist and the instrumentalist(Douglass,1988; Glazer& Moynihan,1975;

McKay,1982; Scott, 1990).

The PrimordialistApproach

The primordialist pproachconceivesethnic

bonds as highly persistent and significant

1 Boswell& Dixon(1990)accept hatthereasonheyusea

cross-nationalesign n theirresearchspartly ueto thefactthatthelong-term atadonotexist or ncome nequality.

volume 8 / number / september001

Page 4: 5 dec III

8/3/2019 5 dec III

http://slidepdf.com/reader/full/5-dec-iii 4/22

Demet Yalcin Mousseau DEMOCRATIZING WITH ETHNIC DIVISIONS

regardless f contextual differences hat have

emerged throughout history. Primordialists

explain strong ethnic attachments with

psychologicalor biologicalfactorsthat, they

think, have primarysignificancein the for-mation of a sense of belonging, in-group

identity,and solidarityamong the members

of an ethnic group (Scott, 1990). They

believe that ethnic ties aredeeply rooted in

history and passdown from one generationto the next. According to primordialists,ethnic differences are stable 'givens' that

often lead to tension andconflictin a society.

Briefly, he primordialperspectiveocuseson

thestrengthof ethnicidentityand claimsthatthe presenceof ethnic divisionsin a nation is

a sufficient condition for the occurrenceof

ethnic violence (Connor, 1994; Greeley,

1974; McKay, 1982; Scott, 1990). Conse-

quently, these views of the primordialistapproachcan be hypothesizedas:

H1: The more ethnically heterogeneous a

society, the higher the likelihood of

extremepoliticalviolence.

politicalviolence= f1 {ethnichetero-

geneity,.. .}

(+)

This conflictview of ethnicity playsaleading

theoretical role in cross-nationalanalysesof

domestic political violence. As mentioned

previously,most empiricalstudieshaveoper-

ationalizedand tested ethnicheterogeneityas

a contributory factor to political violence,

with severalpoliticaland economic factorsasconditions independent from one another

(see Brown& Boswell, 1997; Muller, 1985;

Rummel, 1997). However, the empirical

findingsin this literaturedo not appearwith

consistentresultson therelationshipbetween

ethnic heterogeneityand political violence.

Some studies report either insignificant

results(Muller, 1985) or a negativerelation-

ship (Brown& Boswell, 1997) between the

two.

This inconsistencyamong the empiricalstudiessuggeststhat politicaland economic

mechanisms that may be essential for the

transformation f ethnic divisions nto politi-

cal violence need to be moreclearly pecified.Brown & Boswell (1997) point out that

'ethnic diversity'and 'ethnic mobilization'

havedifferentmeaningsand they should not

be treatedequally n empiricalresearch.This

possibilityis also implied by the instrumen-

talist approachin the ethnic conflict litera-

ture that stresses the forces of political and

socio-economic development as motivatingconditions forviolent ethnic mobilization.

The InstrumentalistApproach

Whereas primordialists emphasize psycho-

logical or biological aspectsof ethnicity as

mobilizingforces, nstrumentalistsunderline

socio-economic and politicalfactorsasmajor

motivatingcausesof ethno-politicalconflict.

The latter perspective s more dynamic than

the formerbecause t stresses hange,contex-

tuality, and competition among ethnic

groups for resources.For instrumentalists,

ethnic conflictin the modernera s not a con-

sequence of the politicizationof primordial

needs, but a result of the mobilization of

ethnic interests to obtain access to social,

political, and economic resources(McKay,

1982).

In the 1970s and 1980s, scholarsof the

instrumentalist ethnic politics literature

reformulatedthe modernization theory to

explain the seeming resurgenceof ethno-politicalconflict around theworld.A new set

of theories in this trend asserted that the

process of modernization nvigoratesethnic

conflict (Bonacich, 1972; Brass, 1985;

Connor, 1994; Glazer& Moynihan, 1975;

Newman, 1991; Olzak & Nagel, 1986;

Rothschild, 1981). Forexample, split labor-

market theory (Bonacich, 1972; Brown &

Boswell, 1995) and ethnic competition

theory (Olzak& Nagel, 1986) point out that

549

Page 5: 5 dec III

8/3/2019 5 dec III

http://slidepdf.com/reader/full/5-dec-iii 5/22

550 journal of PEACE RESEARCH

processes of development reinforce labor-

marketcompetitionalongethnic lines as the

sourceof ethnic mobilization.Thus, workingwithin a 'conflictual modernization'para-

digm, these theories suggest that 'ethnicmobilization' s not a mereexpressionof pri-

mordialsentiments. Ethnic divisions that are

presumablybased on primordialdifferences

do not inevitably lead to ethnic conflict.

Rather, the organization of ethnicity into

ethnic mobilizationis likelyto be 'situation-

allydetermined'.

Sincethe instrumentalistapproachattrib-

utes a fundamental role to political and

economic factors n explainingethno-politi-cal violence, I examine the relationshipbetween ethnic heterogeneityand political

violence with the concomitant effects of

democracy,political change, and economic

development. Below,I define the theoretical

expectationson these variables.The relation-

ship between democracy and political vio-

lence is examined under two different

competing theoreticalframeworks:democ-

racy as a method of conflict management,

and democracyas a sourceof conflict.2

Democracyas a Method of ConflictManagement

Most studies of political violence propose

that the type of political system makes a

difference n a nation's evel of politicalvio-

lence. These studies suggest that the more

democratic a nation, the less likely it will

experiencehigh levels of political violence.

Democratic regimes tend to be legitimateand accepted n people'seyes, sincetheyallow

the political participation of individuals

through nonviolent methods that are not

aimed at destroyingthe regime'sbasicfoun-

dationsor institutions.The pluraland open

nature of democratic regimes toleratesthe

development of a culture and norms that

2 This classificationf thedemocracyiteratures inspiredby Ellingsen& Gleditsch 1997) andEllingsen 2000).

emphasize negotiation and conciliation.

Authoritarianregimes, on the other hand,

preventparticipation n policy and decision-

making and try to control society by force

and coercion.These conditions are likely tocreatean intense sense of injusticeanddepri-vationamongpeople;therefore, he ascentof

violentoppositionand civil war s more ikelyin autocraticthan in democratic regimes.

(Auvinen, 1997; Rummel, 1995).

Some empirical indingsin cross-national

researchsupport these arguments, showingthat nationswith higherlevels of democracyandpoliticaldevelopment endto experience

lower levels of politicalviolence (Flanigan&Fogelman,1970; Gurr, 1993; Ziegenhagen,

1994). While thesestudies seem to suggesta

directlinearrelationshipbetweendemocracyandpoliticalviolence,some empirical tudies

report hisrelationshipn apolynomial,or an

invertedU-shape, form. Muller (1985) and

Muller & Seligson(1987) found that highlyrepressive egimesexperience ower numbers

of deaths from political violence than

medium repressive regimes. Ellingsen &

Gleditsch (1997) and Ellingsen (2000) also

confirmed he invertedU-shape, findingthat

the highest frequencyof conflicts occurs in

'semi-democracies'both in the FirstWorld

and the ThirdWorld.

The polynomial or inverted U-shape

impact of democracy on political violence

shouldbe expectedforseveralreasons.As the

above studies suggest, at high levels of

democracy, political violence is less likely

because democratic systems are capable ofaccommodating the opposing demands

throughlegaland nonviolentformsof politi-

cal mechanisms. However, on the other

extreme, highly undemocraticor autocratic

regimesmightalso be less likelyto experience

politicalviolence, becauseat very low levels

of democracy the political regime has the

means to repress he society so severely hat

little room is left for individualsor groupsto

engagein dissidentbehavior.

volume 8 / number / september001

Page 6: 5 dec III

8/3/2019 5 dec III

http://slidepdf.com/reader/full/5-dec-iii 6/22

Demet Yalcin Mousseau DEMOCRATIZING WITH ETHNIC DIVISIONS

Consequently, the occurrence of highlevels of political violence might be more

likely in nationswith middle-levelor quasi-democraticregimes. In middle-leveldemo-

cratic regimes, individuals and groups areallowedto some extent to protestor oppose

government policies. In these societies,

opposing groups may voice their demands,but the regime is incapable of accommo-

dating them becausedemocraticproceduresand politicalculture are less likely to be well

established.Hence, in middle-leveldemocra-

cies, dissidentbehavioris less often accom-

modated than repressed, increasing the

probabilityofviolence.Accordingly, t canbehypothesizedthat:

H2:Whereashigh and low levels of democ-

racy are associated with lower levels of

political violence, medium levels of

democracyarerelated o higherlevelsof

politicalviolence.

politicalviolence= f2 {democracy,democ-

racy2,. . .}

(+) (-)

Democracyas a Sourceof Grievanceand

Channelfor Conflict n Multi-Ethnic

Societies

Contrary o the aboveviews thatconceptual-

ize democracy as a means for conflict

management, some studies suggest that

democracy might encourage conflict.

Resourcemobilizationtheory proposes that

the more democratica regime,the more theextent of political conflict. Democratic

systems increasethe availabilityof political

resources for the organization of group

demands, facilitating their mobilization

(Ellingsen& Gleditsch, 1997; Hibbs, 1973).

Moreover,some scholars referto democ-

racy as a stimulatingcondition of mobiliz-

ation,particularlyn ethnicallyheterogeneous

societies(Horowitz,1994;Rothschild,1981).

Diamond, Linz & Lipset (1995) claim that

democratic elections in deeply divided

societies are often perceived as zero-sum.

When ethnicgroupsorpartiesose in theelec-

tions, they tend to rejectdemocratic insti-

tutions and appeal to violent means. In asimilar fashion, Melson & Wolpe (1970)contend that in plural societies democratic

procedures and the participation of the

masses in the political process encourage

political eaders o appeal o communal oyal-ties. This processs likelyto reinforcepolitical

competition and mobilization along ethnic

lines. In this way, some studies link demo-

craticandelectoralpoliticswith highlevelsof

domestic political violence, especially inmulti-ethnicnations, suggestingthe follow-

ing testhypothesis:

H3: In ethnically heterogeneous societies,

higher levels of democracy ncreasethe

probabilityof extremepoliticalviolence.

politicalviolence = f3 {(heterogeneity*

democracy2),. . .}

(+)

Nevertheless, an important qualification

relates o thematurityof democraticregimes.

Many scholars n the democratizationitera-

ture place importance on a 'habitation'

process in which the continuous, successful

practiceof democracyover time consolidates

a political culture conducive to resolving

political conflict peacefully(Diamond, Linz

& Lipset,1995; Rustow, 1970). Also, assug-

gestedby split labormarket heory,the labor

market in heterogeneous nations usuallyfragments along ethnic lines (Bonacich,

1972; Brown & Boswell 1995). Fragmen-

tation of labor n this sensemayfacilitate he

emergence of different political cultures

among ethnic groups. Unfortunately,when

such groups compete for resourceswithout

sharing hesame democraticpoliticalculture,

expression of their demands may be more

likely to take violent forms.Accordingly, he

appropriate procedure for examining

551

Page 7: 5 dec III

8/3/2019 5 dec III

http://slidepdf.com/reader/full/5-dec-iii 7/22

552 journal of PEACE RESEARCH

Hypothesis 3 includes control for the matu-

rity of democratic regimes, to distinguish

matured or well-established democracies

fromless maturedemocracies.3

Political Change(Democratization)

It is often argued that political change is

accompaniedby political nstabilityand con-

flict. Politicalchange s a destabilizingprocess

because it increasesuncertaintyabout con-

ditionsof peopleorgroups nsociety,creatinga zero-sum environment. Democratization

does not occurovernight.The establishment

and interpenetrationof democratic culture

and valuesrequirea process hatmight bringintenseissues of conflict (Gurr,1993; Hunt-

ington, 1997; O'Donnell, Schmitter &

Whitehead, 1986).

According to Huntington (1997), the

process of democratizationbolsters ethno-

politicalconflictbecausethe introductionof

democraticprocedures n multi-ethnicstates

tends to produce a political competition

along ethnic lines with communal political

agendas.Similarly,Gurr (1993) arguesthat

democratization brings plentiful oppor-

tunities for ethnic mobilization. Unfortu-

nately,democratizingstates are not capable

of accommodating these movements, since

democratic institutions and norms are not

well established n these states.Accordingly,

violent forms of ethno-politicalconflict are

expectedto occurmorefrequentlyn democ-

ratizing states than in stable and well-

institutionalized democracies.

In the empirical literature,although anumber of studies have investigated the

impact of 'levels'of democracyon domestic

politicalviolence,littlehasbeen done forthe

examination of how 'changes'in levels of

democracymightaffect evelsof politicalvio-

lence.While a 'level'of democracyndicatesa

staticmagnitudeof democracyat one pointof

3 This control variable is referred to as 'democratic ma-

turity' in the operationalization section.

time, i.e. low, medium, and high levels of

democracy,a 'change' in democratic insti-

tutions refers to the differencebetween the

levels of democracyat two or morepoints of

time. Sincea changevariable uchas democ-ratizationcapturesa dynamic process, t has

differentmplications rom the staticvariables

of democracy hatareusedin thisstudy.Gurr

(1993), for instance, tested the impact of

democratizationon variousforms of ethnic

conflict for theyearsbetween1975 and 1986,

and found moderately significant positive

results.Accordingly, his study examinesthe

effect of politicalchangeon extremepolitical

violencewith the following hypotheses:

H4: Democratizingnations thatexperiencea

change in their political structures are

morelikelyto experiencehigher evels of

political violence than nations with

stablepoliticalregimes.

politicalviolence = f4 {democratiza-

tion,. . .}

(+)

H5: In ethnically heterogeneous nations,

democratizations likely to foster higher

levelsof politicalviolence.

politicalviolence= f5 {heterogeneity

democratization,. .}

(+)

EconomicDevelopment

It has long been arguedin the comparative

politics literature that economic develop-

ment createsconditions conducive to politi-cal stability.Lipset's (1959) famous article

describingeconomic development as a pre-

requisitefor political legitimacy and stable

democracy stands out as the leading rep-

resentativeof this trend.Likewise,according

to relativedeprivation heory,people will be

moresatisfiedwith a higher evelof economic

development and prosperityand will there-

forebe lesswilling to resort o rebelliousand

violent actions(Gurr,1994).

volume 8 / number / september001

Page 8: 5 dec III

8/3/2019 5 dec III

http://slidepdf.com/reader/full/5-dec-iii 8/22

Demet Yalcin Mousseau DEMOCRATIZING WITH ETHNIC DIVISIONS

Indeed, the empirical findings on the

relationship between economic develop-ment and political violence generally

suggest that higher levels of economic

development areassociatedwith lowerlevelsof violence across nations (Auvinen, 1997;

Flanigan & Fogelman, 1970; Sigelman &

Simpson, 1977; Weede, 1981; Ziegen-

hagen, 1994).4 Whereassome studies find a

negative linear relationship (Flanigan &

Fogelman, 1970; Weede, 1981), others

report a non-monotonic curvilinear

relationship (Hibbs, 1973), suggesting the

following test hypothesis:

H6: While low and high levels of economic

development are associated with lower

levels of political violence, medium

levels of development are relatedwith

higherlevelsof violence.

politicalviolence = f6 {development,development2,. . .}

(+) (-)

However, the ethnic conflict literaturepro-

poses that this expectation may not hold

4 Weede(1981), Sigelman& Simpson 1977), andotherstest economic development along with income inequality

in order to gauge which one has the more predictive power

in explaining cross-national variation in levels of political

violence. Whereas they confirm the violence-reducing

impact of economic development, they report an insignifi-

cant relationship between income inequality and political

violence. Muller (1985), Muller & Seligson (1987), and

others, however, report a positive significant effect of

income inequality on political violence. As already men-

tioned, these studies focus mostly on the short-term effectof these structuralvariables, using cross-national designs. It

seems that, in support of relative deprivation theory, the

findings of the recent literatureestablish the inequality vari-

able as an important determinant of political violence.

However, this study does not control for income inequality

because long-term data for this variable across a large

number of countries are not available. Moreover, the major

theoretical interest of this study is to assess whether or not

levels of economic development (and democracy) create an

encouraging context for the association of ethnic divisions

with extreme levels of political violence. The relation of

economic development and income inequality to political

violence is a highly complex one that goes beyond the

boundaries of this study.

among ethnically heterogeneous societies.

Resourcemobilization, ethnic competition,andsplit labor-market heoriesproposethat

development provides resources for group

mobilization, generates more issues overwhich to compete, and creates more areas

for group interactions and conflict

(Bonacich, 1972; Connor, 1994; Glazer&

Moynihan, 1975; Horowitz, 1985;

Newman, 1991). Moreover, this compe-tition may be considered to be contentious

rather hanpeaceful in multi-ethnic nations

becausethe struggleover the distributionof

resourcesmay turn into a zero-sum game,

since ethnic demands also include issues ofsymbol and worth that are indivisible and

unmeasurable (Diamond, Linz & Lipset,

1995). Thus, the ethnic conflict literature

suggests the following hypothesis in regardto development:

H7: In ethnically heterogeneous societies,

higher evelsof development ncrease he

probabilityof extremepoliticalviolence.

politicalviolence= f3 {(heterogeneity

development2),. . .}

(+)

Having identified several theoreticalexpec-

tations,I now turn to the operationalization

and measuresof the dependentandindepen-

dent variables.Following this, I discussthe

methods and the research design to be

employedto test the hypotheses.

Research Design,

Operationalizations,and Data

The theoriesand testhypothesesconcern the

impact of ethnic and structural factors on

extreme levels of political violence within

nations. While some hypotheses - such as

the impact of democratization on political

violence - mandate a time-series design,

others identifyvariables hat vary little if at

all acrosstime - such as the degreeof ethnic

553

Page 9: 5 dec III

8/3/2019 5 dec III

http://slidepdf.com/reader/full/5-dec-iii 9/22

554 journal of PEACE RESEARCH

heterogeneitywithin a nation. Accordingly,the appropriate design in this study is a

cross-sectional time-series one, with the

nation-yearas the fitting unit of analysis.As

addressedbelow, data on political violenceconfinethe domain to most nationsover the

years 1948 to 1982.

DependentVariable:PoliticalViolence

Political conflict manifestsitself in different

forms. The main interestof this study is to

look at the most intense forms of political

conflict, violence that results in the loss of

human life. Thus, politicalviolencerefersto

the most aggressivewaysof expressingpoliti-cal discontentwith the statusquo that lead to

human deaths.The intensityand frequency

of politicaldeaths n a nation across imemay

give an approximate understanding of

general levels of political violence in that

nation. Taylor & Jodice (1983) provide

annual data for political deaths caused by

anti-systemevents such as protests,strikes,

riots, armed attacks,and assassinations.All

theseeventsmore orlessreflect hegrievancesof dissidentswith the existingsystem or its

policies.Taylor& Jodice'smeasureof deaths

from political violence has been widely

employed in domestic conflict studies

(Boswell & Dixon, 1990; Muller, 1985,

1988; Muller & Seligson, 1987; Weede,

1981). Weede (1981) suggeststhat this indi-

cator is the best alternative,because deaths

are less underreportedcross-nationally han

alternativesuch as armedattacks.

However,Brockett(1992) considers hesedataproblematicbecause,he contends, they

underreport the political deaths for some

nations, particularly in Central America.

Still,Dixon, Muller& Seligson(1993) argue

that the underreporting of the political

deaths usually happens during periods of

large-scalensurgencyor civilwar.Therefore,

Dixon and his colleagues uggesttreating his

problem by establishinga censoring point

that representsmoderately high levels of

violence. Muller & Seligson (1987) and

Boswell & Dixon (1990), forexample,estab-

lish their ceiling point at 50 deaths permillion. They propose that since the under-

reported cases in Central America

(Guatemala, El Salvador, and Nicaragua)exceed this threshold,these casesshould not

causea seriousproblem.

Asreportedn theanalysisection,the most

frequentpoliticaldeathsin our datasetoccur

at the lower evel.The occurrence f 50 deaths

or above n anysinglenation-yearsrare 9%).

Thus, a caseof 50 deathsor morerepresents

comparatively igh level of politicalviolence.

Since our aim is to explainhigheror extremelevels of politicalviolence acrossnations, in

this studyI used a censorwith a cutoff of 50

deathspernation-year.Moreimportantly,t is

likelythat 50 deathswill be reportedeven in

developing nations. For reasons that will

become clear,I leave fuirther perationaliza-tion of the dependentvariable or the analysis

section.Fornow,dataon the dependentvari-

ablewereobtained romTaylor&Jodice's ari-

able of 'number of deaths from domestic

politicalviolence' rom theirWorldHandbook

ofPoliticalandSocial ndicators1983).5

IndependentVariables

Ethnic heterogeneity In determiningthe

level of ethnic heterogeneityin a nation,

scholarsmake use of a varietyof theoretical

criteriabased on cultural,racial,linguistic,

religious, or historical distinctions among

ethnic groups.However,althoughthese cat-egorieshave often been used to distinguish

ethnicgroups, t hasbeendifficult o establish

5 Sincethe measure f deaths rom politicalviolencealso

includesstate-sponsorediolence,one reviewer uggested

that I might also controlfor some measureof non-rebel-

lious violence.Upon this suggestion, performedall the

analyseswith controladdedforTaylor& Jodice's 1983)

'attacks y state'variable.The variable emainednsignifi-

cantin all the regressions,nd its inclusiondid not affect

the relationship f the othervariableswith politicalvio-

lence.I therefore ptedto report he analyseswithout nclu-

sion of this variable.

volume 8 / number / september001

Page 10: 5 dec III

8/3/2019 5 dec III

http://slidepdf.com/reader/full/5-dec-iii 10/22

Demet Yalcin Mousseau DEMOCRATIZINGWITH ETHNIC DIVISIONS

common groundfor definingandmeasuring

ethnicity.Since avarietyof factorsmaydiffer-

entiatepeople,on the basisof what criteriado

we observe ethnic groups? For example,

whereas the French and the English inCanadaaredistinguishedn termsof languageand religion,AfricanAmericansand Euro-

peanAmericans n the United Statesarecon-

trastedwith respect o racialdifferences.

A data source for ethnic heterogeneity s

offeredby Gurr (1994) in his Minoritiesat

RiskProject.Gurrprovidesan 'ethnicdiffer-

ence index' that is developedwith respectto

four attributes f ethnicity:anguage, eligion,

race, and custom. However, this measureisnot appropriateorthisstudybecause he unit

of analysis in Gurr'sdataset is the ethnic

group. He assigns ordinal level scores to

ethnic groups basedon the extent to which

they are different from the dominant ethnic

population with respect to their linguistic,

religious, racial,and custom characteristics.

Therefore, it is necessaryto use a national

level measure that assessesnot the extent to

which ethnic groups are differentfrom the

dominant populations, but the degree to

which a nation is ethnicallyheterogeneous.

Therefore, the best availablemeasureof

ethnic heterogeneity or the purposesof this

study seemsto beTaylor& Hudson's 1972)

index of 'ethnolinguistic fractionalization'

for 136 countriescirca 1960.6 In construct-

ing this measure, Hibbs (1973) states that

cultural, ethnic, and linguistic differences

among ethnic populationsare taken as the

main criteria.The differencesamong groupswere determinedwith respectto the groups'

roles and descents ratherthan theirphysical

characteristics. lthoughthismeasuremostly

incorporates cultural and linguistic dimen-

6 Hibbs notes that'. . . the Index of Ethnolinguistic Frac-

tionalization (ELF) is derived from data reported in the

authoritative Atlas Narodov Mira on the numbers of people

in distinctive cultural, ethnic, and linguistic groups' (1973:

68). Taylor & Hudson constitute the Ethnolinguistic Frac-

tionalization Index from these data by using the 'fragmen-

tation' index of Rae & Taylor (1970).

sions of ethnicity, it is said to be highlycorrelated with other attributes of ethnic

diversity (religious or racial) (Taylor &

Hudson, 1972).7

Consistentwith most studies of politicalviolence, this study utilizes this measureas

the indicator of ethnic heterogeneityn a

nation. The time period circa 1960 falls

approximately in the middle of the time

period analyzed (1948-82). Since ethno-

linguisticdifferenceschange very slowly, t is

not expected that any fundamentalchangeoccurredduringthe time-spanof this study.

Ethnic separatism Muller & Seligson(1987), Boswell & Dixon (1990), and others

control for the effectof the 'intensityof sep-

aratism',considering that separatistethnic

groupsconstitute a threat to the integrityof

the state with their potential for violent

mobilization. Potential orseparatist iolence

usually develops when territoriallyconcen-

tratedethnic populationsstartseekingcom-

plete autonomy from the dominant state to

control resources.Brown& Boswell (1997)

argue hatseparatismsrelated o a mobilized

political action, and therefore has different

implications from ethnic heterogeneity.

Whereas the latterreflectsa nation's evel of

ethnic diversity,separatism s consideredas

an organizedactivity hatmayeasilytranslate

into violence. Consistent with previous

studies,I control for ethnic separatism ince

it may have an independentpositive impact

on politicalviolence apartfrom ethnic het-

erogeneity.The data for this variable areobtained from Gurr's (1994) Minoritiesat

RiskProject.Gurr identifiesthe active sepa-

ratistgroups among 233 communal groups

7 The fractionalizationndex s calculatedor each nationwiththefollowing ormula:Fractionalization[1- X Ni/ N) *(Ni- 1/ N - 1)]* 100;whereNi =number f peoplein the ith groupandN = totalpopulation.This indexgivesa ratio variable angingbetween0 and 1. The larger henumberof groupsand the proportionof theirpopulationin the total population,the more differentiated s thenation.

555

Page 11: 5 dec III

8/3/2019 5 dec III

http://slidepdf.com/reader/full/5-dec-iii 11/22

556 journal of PEACE RESEARCH

between 1945 and 1989 on a four-pointordinal scale. Based on this scale, I con-

structed a dummy variablefor ethnic sepa-

ratism, assigning 1 to nation-years that

contain at least one active separatistethnic

group,and 0 otherwise.8

Democracy The democracy variable is

operationalizedby using the annual democ-

racy and autocracyscores provided in the

PolityIIIdataset Jaggers& Gurr,1995). The

PolityIII dataincludea scale for democracy,which measures the availability of insti-

tutions that allow for politicalparticipation

of citizens in the political system, constrainthe executive power in office, and preservefundamental civil liberties. Although civil

libertiesare not measureddirectly,Jaggers&

Gurr(1995: 475) show that theirdemocracyindex correlateshighlywithvariousmeasures

of civil liberties,i.e. FreedomHouse'spoliti-calrightsand civil liberties ndexandBollen's

politicallibertyindex.

Jaggers& Gurr(1995) suggest measuring

democracy by subtracting the 11-point

autocracy corefromthe 11-point democracy

measure in order to increase the range

between extremely autocratic and highly

democraticregimes.I follow this suggestion

here,with the scalerangingfrom -10 to +10

based on a summated index of five insti-

tutional dimensions: competitiveness of

political participation, regulation of political

participation, openness of executive selec-

tion, competitiveness of executive recruit-

ment, and constraintson executiveauthority.

Democratic maturity The hypothesized

curvilinear nverted-U impact of democracy

8 For similar categorizations of a dummy variable for sepa-

ratism, see Muller & Seligson (1987) and Brown & Boswell

(1997). The following countries had missing data on sepa-

ratism: Benin, Central African Republic, Gabon, Gambia,

Jamaica, Kuwait, Libya, Malawi, Mongolia, Nepal, Tan-

zania, Tunisia, Trinidad, and Uruguay. Following Muller &

Seligson (1987: 448), I coded these countries 0 using infor-

mation in Bankss (1976) Political Handbook of the World.

on politicalviolence does not count for the

maturity or duration of a democracy.As

mentioned previously,a 'level' refers to a

democracyscore at a certainpoint in time.

Therefore,any developing or democratizing

nation, such as Turkeyor Nigeria, can have

higher levels of democracyat certainpointsin time perhaps due to extensive institutional

reforms. However, this variable does not

indicateto what extent the regimeis durable

or mature. As discussed, it is plausiblethat

more matureand stable democraticregimeswill have lower levels of political violence

than less maturedemocraticregimes,due to

the consolidation,over time, of democraticnorms of peaceful conflict resolution

(Diamond, Linz & Lipset, 1995). Accord-

ingly, failure to control for the regime ma-

turityeffect may mask the actual impact of

levels of democracy on political violence.

Following Mousseau (2000), I operational-ized democraticmaturityas the product of

theageof regime (logged +1) andthe nation's

democracyscore (+11). I obtained data on

regimeagefromthe DURATION variableof

the Polity III (version 98) dataset, which

measures he ageof the regimeor the number

of years since a nation experienced any

sudden institutional change, starting from

the year 1900.9

Democratization A few studies in the

empiricalliteratureon domestic and inter-

national conflict have operationalizedand

tested the impactof change n thedemocracy

variableon various orms of conflict and war.For example, Gurr (1993) measured the

changevariableas the number and direction

of scale-point changes on the democracy

indicator between 1975 and 1986. In the

internationalconflictliterature,Mansfield&

Snyder (1995) looked at the impact of

9 The durability ata aremissing n the Polity 98 dataset

before heyear 1954. I filled n theyears1948 to 1954 byextrapolatingbackwardand noting abrupt changesof

regimes.

volume 8 / number / september001

Page 12: 5 dec III

8/3/2019 5 dec III

http://slidepdf.com/reader/full/5-dec-iii 12/22

Demet Yalcin Mousseau DEMOCRATIZING WITH ETHNIC DIVISIONS

democratizationon the probabilityof inter-

national wars by classifying the scores of

democracyvariable in the Polity II dataset

within three categories:democracies, auto-

cracies,and anocracies.This third categoryindicates political systems in which demo-

cratic and autocratic characteristics are

mixed. These identifydemocratizing tatesas

thosewith regimechangesfromautocracy o

eitheranocracyor democracy,or from anoc-

racyto democracy. n addition,they catego-rizeautocratizing tates if theirregimemoves

from democracy o autocracyor anocracy,or

from anocracy o autocracy.In a

more recentstudy,Ward& Gleditsch(1998) operationalized he democratization

variable as the change in the scores of the

democracy ndicator n the PolityIII dataset

over ten-year periods. Ward & Gleditsch's

(1998) continuous measureprovidesa better

indicator of change than the categorical

measures,because it makes use of all the

informationavailableand avoids the risks of

arbitrariness n the construction of the cat-

egories. Becausedemocratization akesplace

in agradualprocess,Ward& Gleditschoper-ationalized it during periods of ten years to

capturea meaningfulchange in the democ-

racyvariable.FollowingWard& Gleditsch, I

constructed a changevariableby calculating

the scale-point changes on the democracy

indicatorin the Polity III dataset over ten-

yearperiodsbetween 1948 and 1982.

Economic development Consistent with

the political violence literature, this studygaugesthe impact of different evels of econ-

omic development on political violence

with a measureof energy consumption per

capita(logged).This measure s preferred o

GNP, since some studies demonstrate that

the latter may be an unreliable measure

(Henderson, 1991; Summers & Heston,

1988). The data for this measure were

obtainedfrom Banks's 1992) Data Archive.

With the operationalizations,measures,and

data sources of the dependent and indepen-dent variablesnow described,we can move

on to the methods and research designsection.

Analysis and Results

The dependent variable, the number of

deaths from political violence, is an event-

count measure. The nature of these data

means that Ordinary Least Squares (OLS)

analysis s not preferred, or severalreasons.

First,the minimum andmaximumvaluesof

these data rangefrom 0 to 525,304 deaths.

Most empiricalstudies of political violencethat use thesedatatake the natural og of this

variable to smooth extreme values and to

decrease the large range between the

minimum andmaximumvalues.10However,

this method may generate inefficient esti-

matesfor this study's argesampledue to the

inclusionof an extensive ime period.As dis-

cussed,the presenceof dynamic variables n

the sample requires he analysisof a pooled

time-series sample rather than a cross-

nationalsample.Moreover, see no reason oexcludefromthe samplethe smallnumberof

casesof extremeviolence,such asdeaths rom

civil wars, since our concern is to assess

whether or not higherlevels of violence are a

common pattern among ethnically hetero-

geneousnations.

Owing to the large range between

minimum and maximum values,even after

taking the natural og of the political deaths

variable, hedatado not meet thestandardofnormal distribution. This means that the

mean andstandarderrorsof this measuredo

10 As mentioned before, Boswell & Dixon (1990) and

Muller & Seligson (1987) also establish a censoring point

at 50 deaths per million to eliminate the extreme cases.

Because this adjusted indicator is expressed as a rate, they

use logged OLS (Dixon, Muller & Seligson, 1993). The

technique may not pose a problem for these studies'

samples, since they use a relatively small number of cases in

cross-national designs and thus have fewer if any cases of

extremely high values.

557

Page 13: 5 dec III

8/3/2019 5 dec III

http://slidepdf.com/reader/full/5-dec-iii 13/22

558 journal of PEACE RESEARCH

not meet the assumptionsof OLS estimates.

In addition, as Table I shows, 73% of the

nation-yearsn thisstudyhave hadno deaths

from political violence between 1948 and

1982. The distribution of these datais thus

highly skewedtowards 0 and lower levels of

deaths. Forexample, nTableI the frequencyof occurrenceof more than200 deaths n the

sample is 5% and of more than 50 deaths

9%. Also, becausethe dependentvariable s

continuous and takeson only positivevalues,

it may be problematicto interpretthe co-

efficientswith OLS. King(1989) stresses hat

the adoption of OLS techniques for such

event-count models is problematic,becausethese models producebiasedparameter sti-

mates(seealsoWang, 1993).

Fortunately, he empiricalliteraturepro-vides more sophisticatedtechniques for the

analysisof event-count models. One alterna-

tive suggestedby King (1989) is the Poisson

distributionor NegativeBinomial,with data

aggregatedover time. However,the rangeof

the data (0 to 525,304) is far too wide to fit

the Poissondistribution,and an aggregationover time would lose information in the

crucial but continuous dynamic variables

such as democratization and economic

development.Anotheralternativemightbe to

rescale hedependentvariable nto several at-

egories (as in TableI) and estimatewith an

orderedprobit or logit. The constructionof

thesecategoriesmayseemarbitrary,owever,

TableI. FrequencyDistributionof Deathsfrom

PoliticalViolence n Sample

No. ofdeaths %

0 73

1-3 7

4-10 5

11-25 4

26-50 2

51-100 2

101-200 2

>200 5

andit is not clearhowto accountfortemporal

dependencewith orderedmodels.

All things considered,it seems the best

alternatives a logit model designed for the

analysisof binarydependentvariables King,

1989). While the settingof the dichotomous

cutoff for the dependentvariablemay seem

arbitrary,t is less arbitraryhan the use of

several utoffpointsneededfor orderedanaly-sis. More importantly,Beck, Katz & Tucker

(1998) havedevelopeda simpleprocedure or

accounting for temporal dependence with

binaryvariables. he procedures to construct

a count variableindicatingthe number of

yearswith no events n thedependentvariable(peaceyears), plus three additionalvariables

modeling the non-linearityof the estimator

using a cubic spline (cubicspline 1, 2, 3).

To avoidappearingarbitraryn my choice

of cutoff point, I followed the spirit of

Boswell & Dixon (1990) and constructed

the binary dependent variable at the 50-

death ceiling point, with 1 indicating all

nation-years that experienced at least 50

deaths from political violence and 0 other-

wise. This categorizationis appropriate or

several reasons. First, the cases with 50

deaths and above representrelativelyhighlevels of political violence and occur less

frequently than cases with lower levels of

deaths, and the objectivehere is to account

for higher levels of political violence.

Second, it is necessary o censor the data to

eliminate or reduce the reporting bias on

political deaths in developing countries.

Since50 deaths are ikelyto be reportedevenin developingnations,the binarydependentvariablewill, at least to some degree,reduce

the reportingbiasin the data.

Since more deathsareexpected n nations

with largerpopulations,I introducedcontrol

fornationalpopulation sizeas one of the inde-

pendent variables,with data obtained from

Banks's1992) Data Archive.The mergingof

these data sources produced a sample of

3,172 observations with 249 yearswith at

volume 8 / number / september001

Page 14: 5 dec III

8/3/2019 5 dec III

http://slidepdf.com/reader/full/5-dec-iii 14/22

Demet Yalcin Mousseau DEMOCRATIZING WITH ETHNIC DIVISIONS

least50 deathsfrompoliticalviolencefor 126

nationsfrom 1948 to 1982.11 The basicsta-

tistics for the dependent and independentvariablesaregiven in AppendixA.

The analysesbeginwith a basemodel esti-mating the impact of ethnic heterogeneity,levels of democracy, democratization, and

economic development on the measure of

extreme political violence, controlling for

population size, ethnic separatism,demo-

cratic maturity,and temporal dependence.

Afterwards, the conditional relationships

among the independent variables are

reportedwith the addition of the interactive

terms. All estimateswere performedusingWhite (robust)standarderrors.

As can be seen in Model 1 in TableII, the

base model is significant, confirming the

expected invertedU-shape form for democ-

racyandeconomic development.Democracy(0.302) and its square (-0.011), and econ-

omic development (0.771) and its square

(-0.057), areall significant.Consistentwith

Hypotheses 2 and 6, it seems that nations at

middlelevelsof democracyanddevelopment

are more likelyto experiencehigherlevels ofpolitical violence than nations at low and

high levels of democracyanddevelopment.

As expected with Hypothesis 1, in this

model the variable ethnic heterogeneity

(0.137) is not significant.This finding is con-

trary o theprimordialist iew thatconstructs

ethnic heterogeneityas a structuralcause of

higher levelsof politicalviolence. It appears

that ethnic divisions, alone, do not con-

tribute to extreme political violence. Theinsignificantcoefficient for the democratiza-

tion variable -0.016) indicatesthatnations

with change in theirpoliticalinstitutionsdo

not seem to experience any more violence

than nationswithout change. This outcome

l In calculatingthe change variable (the differencesbetweeneach ten nation-yearntervals),he first tenyearsneeded to be calculatedwith reference o the databefore1948. Some missingdataprior o 1948 led to the deletionof somedatapointsafter1948, sincethey didnot haveanyreference oints forthecalculation f thechange.

is contraryto Hypothesis 4, which predictsdemocratizationto be related to a higher

probabilityof extremepoliticalviolence.All

the remainingvariables n Model 1 in Table

II arecontrolvariables,andallaresignificantin the expecteddirections.While ethnic sep-aratism(0.268) and population size (0.308)

appear o increase he probabilityof extreme

levelsof politicalviolence,democraticmatu-

rity (-0.016) appears o amelioratet.12

The second model in Table II examines

whether the ameliorating mpact of democ-

racy holds also among ethnically hetero-

geneous nations. Recall that Hypothesis 3

predicts increasingviolence in ethnicallydividednationswith higherlevels of democ-

racy. This hypothesis is mathematicallymodeled as ethnicity*democracy2o distin-

guishthe effectsof lowerandhigher evels of

democracy in ethnically heterogeneousnations. As can be seen in Model 2, the co-

efficient for the interactive term eth-

nicity*democrac/ appears positive and

significant (0.003). This indicates that the

estimate without the interactive term in

Model 1 is underspecified,and instead the

impact of each constituent term, ethnicity

anddemocrac/, isconditionalon thevalueof

the other (Friedrich,1982). Figure 1 illus-

trates these conditional impacts for a state

with ethnic separatism, table nstitutionsfor

at least ten years (0 democratization),and

average values of economic development,

institutionalmaturity,population size, and

conflicthistory.

As can be seen in Figure1, the impact ofdemocracy on extreme levels of political

violence takes the form of an inverted U for

ethnically homogeneous nations, but for

12 The democraticmaturityvariable s highlycorrelatedwith the democracy ariable 0.86) andits square 0.89).Multicolinearity,owever,will not affect he interpretationof these coefficients.Multicolinearityproducesinflatederrorerms,ndicatinghat hefindings remore ignificantthan heircalculationshow n theregression. his does notposea significantproblem n our casebecauseall of thesecoefficients restatistically ignificant Friedrich, 982).

559

Page 15: 5 dec III

8/3/2019 5 dec III

http://slidepdf.com/reader/full/5-dec-iii 15/22

Page 16: 5 dec III

8/3/2019 5 dec III

http://slidepdf.com/reader/full/5-dec-iii 16/22

Demet Yalcin Mousseau DEMOCRATIZING WITH ETHNIC DIVISIONS

Figure1. Impactof Levelsof Democracyon the

Probabilityf ExtremePoliticalViolence n Ethni-

callyHomogeneousandEthnicallyHeterogeneousNations

30% - '

8

.2

0.E2

o

I

1

CL

0.

edt

in

an

ca

ge

pr1ab

na

pr

ex

vi(

he

th

les

Filedt

cr;

ate

sta

foi

ra(

eti

be

th

Le

!5%

!0%

*

EI .- -

*UU--*-----

(democracy >= 17) in the analysis include a

myriadof states n the sample,suchasPortu-

gal (1976-82), Mauritius (1978-82), and

Peru (1980-82). Less institutionallymature

democracieswith extreme levels of politicalviolence include Nigeria (1981-82), Spain

(1978-81), andSudan(1965-66), to name a

few.

5X- '; * To what extent is the extremeviolence in5%--- -. /i...-------.------.-----...\- ----- ethnically heterogeneous democracies a func-

/0\;^ tion of recent political or institutional

/. * \: changes?The third model in Table II reports5%T- - - - -- --,------------- --- ---.......... thesamepredictorss in Model1, onlythis

time I introducethe interactive erm democ-0% i Ih Iam Irdcosa

nMdli nyti

-10 -5O

510

ratization*ethnicity to assess the effect ofDemocracy political change on extreme levels of political

-Ethnically homogeneousations violence in ethnically heterogeneous* Ethnically heterogeneous nations societies. Since politicalchange is associated

with increasinguncertaintyabout conditions

of ethnic groups, democratizationmay have

inically heterogeneous nations the adestabilizing ffecton multi-ethnic nations.

verted U takes the form of a wave: the Although the regime may graduallyrelax

lelioratingimpact of democracy s signifi- politicalrestrictionson ethnicgroups,it may

ntly less. While highly democratic homo- not be capable of accommodating risingneous nations have an estimated ethnicdemandsbecauseofa lackof resources

obability of extreme political violence of and democraticpoliticalculture,aspredictedout 6%, highly democraticheterogeneous with Hypothesis 5.

tions have the much higher estimated Indeed,the analysisn Model 3 appears o

obability of 14% - about 1 year in 7 confirm these expectations.The interaction

periencingat least50 deaths frompolitical effectof democratizationand ethnic hetero-)lence. Extreme violence in autocracies, geneity (0.077) is significantin the positive

wever, appearsabout the same (6-8%), direction, indicating that the process of

ough autocraciesappear to have slightly democratization is associated with higher;sviolence, levels of political violence in ethnicallyThe estimate n Model2 andillustratedn heterogeneoussocieties. Note also that the

gure 1 thus confirmsHypothesis3: while negative constituent term democratizationhnicityhas littleimpacton stateswith auto- now appears ignificant(-0.047). This indi-

itic structures,ethnic divisions are associ- catesthat democratizationdoes indeed have

ed with extreme violence in democratic a significant pacifying influence in highlyLtes.However,because he analysiscontrols homogeneous nations (when ethnic hetero-

r the pacifying impact of maturedemoc- geneity= 0), but not for ethnically divided

cy, the positive impact of democracy on nations. This can be seen more clearly in

inicallyheterogeneous ocietiesshould not Figure2, which shows the estimatedimpactinferredto apply to all democracies:only of democratizationand its interaction with

ose with relativelyess matureinstitutions, ethnicityon extremepoliticalviolence from

ss institutionally mature democracies the coefficients n Model 3, fora state whose

561

Page 17: 5 dec III

8/3/2019 5 dec III

http://slidepdf.com/reader/full/5-dec-iii 17/22

562 journal of PEACE RESEARCH

Figure 2. Impact of Democratization on the

Probability of Extreme Political Violence in Ethni-

cally Homogeneous and Ethnically Heterogeneous

Nations

30%

25%

8 20%

* 15%

10%

s 5%

0. 0%

U

U

.U

: -::::.-- .._:::1

\ * e

U

-10 -5 0 5Democratization

Amount nddirection f institutionalhange

Frommedian-leveldemocracyscore of O

-- Ethnically homogeneous nations

* Ethnically heterogeneous nations

democracyscore was 0 ten yearsbefore (and

thus whose regime is ten years old), and

where the nation has ethnic separatismand

average values of economic development,

populationsize, andconflicthistory.At firstglance, Figure2 looks very much

like Figure1, andfor averygood reason: t is

not possibleto separate he effects of changesin democracy rom actualdemocracy evels-

as one cannot democratize while being a

democracy!Hence, 0 on the democratization

scale refers to an anocracy today with no

changen itsdemocracy evelcompared o ten

yearsbefore; 10 indicatesa 10-pointchange

from anocracyto extremeautocracy(in tenyears);and 10 indicates a 10-point changefrom anocracy o extremedemocracy(in ten

years).The striking indingin Figure2 is that, to

reduceextreme ormsof politicalviolence,an

ethnically heterogeneous anocracy is better

off autocratizingratherthan democratizing.Whereas n the long runethnicity appears o

have little impact on violence in autocratic

nations (Figure 1), the short-term mpact of

a change towardsautocratization Figure2)

appears o reduceviolence moreeffectively n

ethnicallyheterogeneousnations (6%) than

in more homogeneous ones (10%). This

seems intuitivelyplausible:greaterautocra-tization allows thestate to controlcompetingethnic groups with extremely repressivemeasures measures hatmayinflamea more

unifiedresistancen homogeneoussocieties.

Still, the focus on the literaturehas not

been on autocratization s a cure forpolitical

violence,but on democratization saconduit

of it (seeabove,Hypothesis5).Turning o the

right side of Figure 2 (the democratizing

side), there appears ittle substantivediffer-ence betweenFigures1 and2. This indicates

that the increased risk of higher levels of

politicalviolencewhendemocratizingn eth-

nicallyheterogeneoussocieties,comparedto

homogeneous societies, is a function not of

the changen institutions tself,butrather he

statusof being democratic (as seen in Figure1). That is, newly democratic ethnically

heterogeneous ocietiesappear o be no more

prone to violencethan other democraticeth-

nically heterogeneous ocieties.13

Model 4 in TableII reports he multivari-

ate analysis with the addition of the inter-

active variableof ethnic heterogeneityand

economic development. As expressed in

Hypothesis 7, higher levels of economic

development areexpected to producea posi-

tive effect on extreme political violence in

ethnicallyheterogeneousnations, asdevelop-

ment may increasegroup grievances hrough

competition. Sincethe objectiveof this studyis to assess the impact of higher levels of

13 I also estimateda combinedmodel that includesboth

ethnicity*democracyndethnicity*democratizationalongwith the remaining ariablesn Model 1), but this model

yieldeda log-likelihood f-669.2, and og-likelihood atio

tests ndicate hat his snot a significantmprovementver

Models 1 (p = 0.1277 at 2 degreesof freedom),2 (p =

0.2182 at 1 degreeof freedom), r3 (p=0.3663 at 1 degree

of freedom).It seems that colinearityamongthese termsrenders he combined estimate inefficient.In any case,

furtheranalysis howed hat this combinedmodel yieldedresults lmost dentical o that illustratedn Figure .

volume 8 / number / september001

w \ - - - - - - w

* - - - - - - \

@ - - - - - - - - - - -

Page 18: 5 dec III

8/3/2019 5 dec III

http://slidepdf.com/reader/full/5-dec-iii 18/22

Demet Yalcin Mousseau DEMOCRATIZING WITH ETHNIC DIVISIONS

development, the interactive

mathematically formulated

nicity*development2.he analysis i

however,does not support this e)

The coefficientfor the interactiv

nicity*development2s not significa

indicating that the impact of hi

of development on extreme

violence does not varywith respecconditions. This conclusion is al

by the insignificanceof the chi-s<

tistic for Model 4, indicatingtha

yields the better estimate of the

economic development on pol

lence.14Accordingly,Figure3 presents

of economic developmentas indi

the coefficientsin Model 1 for a

ethnic separatist movements anvalueson all theremaining ariable

seen in the figure, development

havea polynomial, nverted-Usha]

Figure 3. Impact of Levels of Econon

ment on the Probability of Extreme Po

lence in Nations

40% -

35%-

S 30%-

I 25%

I 15%

110%

0. 5%

2 4 6 8

Economicdevelopment(logged energyconsumptionperc

14 I also estimateda full modelcontaining;

action terms(fromModels2-4), but thism

log-likelihoodof -669.0, not a significantover Models1 (p= 0.2004 at 3 degrees ffre0.3611 at 2 degreesof freedom),or 3 (p

degreesof freedom).

variable is on theprobability f extremepoliticalviolence

as eth- regardlessf ethnic conditions.Highlyunder-

nModel4, developed and highly developedstateshave

cpectation. estimated probabilitiesof violence of only

e term eth- 11%and7%,respectively,ompared o apeaknt (0.012), of 35%amongdeveloping tates.This finding

gher levels strongly upports heoriesof resourcemobiliz-

political ationandcompetitionthatexpectthe process:t to ethnic of economic modernization to invigorateso reached extreme political violence in developingquared ta- nations.

t Model 1 In summary, he overallanalyses ndicate

impact of that whereas ethnic heterogeneitydoes not

itical vio- appearto be an independent condition for

the occurrenceof extremepoliticalviolencethe impact within nations, its interaction with political

catedfrom conditions appears to have a significantstate with impact on higherlevels of politicalviolence.

Id average The political conditions that I analyzedin

s.As can be multi-ethnic societies were levels of democ-

appearsto racyand change in these levels,or democra-

ped impact tization. It seems that the well-established

inverted-U effect of democracyon political

nicDevelop- violence applies only for homogeneous

litical Vio- nations; for ethnically heterogeneous

societies,the inverted U takes the form of a

wave, with democracy'spacifying impact

;~ only about half as potent. Democratization,--:-----~------- in contrast,does not appear o effect extreme-- -------------- levelsof political violence in nations beyond

the effect of thelevel of democracy tself, but,

N in a comparisonof homogeneousandhetero-

geneous nations, autocratization appears

more effective in reducingextremeforms of

violence in the latter.An importantcaveat o

----------------- hese conclusions is that these conditionsi i applyonlyto lessmaturedemocracies.These

10 12 empirical findings thus lend support to

instrumentalistviews of ethnic conflict andcapita)

consolidationistviews of democratic tability.

allthreenter- Conclusionlodelyieldeda

improvement Empirical research on political violenceeedom),2(p=..= 0.5125 at 2 hasfocusedmostly on political andeconomic

factors in explaining violent political

563

_ _? _ * ^ 4 ow 8 o - @

b / @@@X0@X@~~~?I

Page 19: 5 dec III

8/3/2019 5 dec III

http://slidepdf.com/reader/full/5-dec-iii 19/22

564 journal of PEACE RESEARCH

behavior,with tests for ethnic heterogeneityas an independent condition for the occur-

renceof politicalviolence.However,whether

or not ethnically heterogeneous states are

moreproneto extremepoliticalviolence than

homogeneous states - and, if so, why -

remainsan importantquestionin this litera-

ture. Presenting he argumentsthat empha-size how political and economic context

might be relevant n explainingpoliticalvio-

lence in multi-ethnicsocieties,I designedan

empirical ramework o test the independentand conditional impacts of ethnic, demo-

cratic,and economic conditions on extreme

forms of politicalviolencewithin nations.Between the two competing theoretical

frameworks that guided this empirical

investigation, the instrumentalistapproach

logically and empirically gained supportfrom this study. First, contrary to the pri-mordialistexpectation, n none of the analy-ses was the variable or ethnic heterogeneitysignificantlyrelated o higher levels of politi-

cal violence. Rather, ethnic diversity

appeared as insignificant, indicating that

ethnic divisions, alone, are not enough to

fosterpolitical violence.

Second, the analysesconfirmedthe well-

established inverted-U shape impact of

democracyon politicalviolence - but only

forethnicallyhomogeneous nations.Foreth-

nically divided nations the inverted-U takes

on theform of a wave, with democracy's aci-

fying impact only abouthalf as potent. This

effect, however,must be put in the perspec-

tive of theoverallanalysis,asseparate ontrolfor the maturityof democratic institutions

merits only the conclusion that the 'wave'

impact of democracy in ethnically hetero-

geneous societiesappliesonlyto relativelyess

maturedemocracies.

Third, democratization oes not appear o

effect extreme political violence in nations

beyond the effect of the level of democracy

itself but, in a comparison of homogeneous

and heterogeneous nations, autocratization

appears o lowertheprobabilityof violence in

heterogeneousstates.This might be due to

intensive authoritarian measures against

ethnic mobilization, whereas these same

measures n morehomogeneousstatesmight

ignite a more unified resistance. Another

interesting indingof this studywas that the

impact of economic developmenton politi-cal violence did not vary with respect to

ethnic conditions. An inverted U-shapeeffect is obtained for all forms of societies,

indicating that whereas a medium level of

development increases the likelihood of

extreme political violence, low and high

levels of development ameliorate it. Theseresultsgive merit to the alreadymentioned

instrumentalist heories n the ethnic politics

literature, heories that identifypoliticaland

economic processesascontributory actors o

ethnic mobilization and extreme forms of

politicalviolence.

Althoughthese results uggest he despair-ing conclusion that democracy in multi-

ethnic societies will face more intense forms

of politicalviolence than other democracies,

these resultsarereasonable, onsideringthat

the maturityof a democracyand a highly

developed economy appear as important

conditions for peacefulconflictresolution.It

thus seems that the development of an all-

embracing democratic culture and norms

thatreinforce nstitutions s necessary, s well

as bettereconomic conditions. Interestingly,

given that most of the mature democracies

are also highly developed, and given that

both democraticmaturityand high levelsofdevelopmentappear o decreasepoliticalvio-

lence, future studies should be aimed at

deciphering the joint effects of economic

modernization and democratic insti-

tutionalizationon domestic conflict.

It thus appears hat ethnic violence is not

a function of inexplicable and fixed pri-

mordial elements, but rather a function of

explainableenvironmental conditions. The

contributionof ethnicityto politicalviolence

volume 8 / number / september001

Page 20: 5 dec III

8/3/2019 5 dec III

http://slidepdf.com/reader/full/5-dec-iii 20/22

Demet Yalcin Mousseau DEMOCRATIZING WITH ETHNIC DIVISIONS

can be understood better with an analysis of

ethnicity within the political and economic

context of nations. Accordingly, we arrive at

the promising conclusion that ethnic vio-

lence is environmentally influenced and,

thus, can be regulated by changing the con-

ditions that cause it.

References

Auvinen,JuhaY., 1997. 'PoliticalConflict in Less

Developed Countries, 1981-89', Journal ofPeaceResearch4 (May): 177-195.

Banks, Arthur, 1976. Political Handbookof the

World.New York:McGraw-Hill.

Banks, Arthur, 1992. Cross-NationalTime-Series

Data Archive.Binghamton, NY: Center for

SocialAnalysis.

Beck, Nathaniel; Jonathan Katz & Richard

Tucker, 1998. 'Taking Time Seriously in

Binary Time-Series-Cross-SectionAnalysis',

American Journal of Political Science 42

(October):1260-1288.

Bonacich, Edna, 1972. 'A Theory of Ethnic

Antagonism:The Split LaborMarket',Ameri-

can Sociological eview 7 (October):547-559.

Boswell, Terry & William J. Dixon, 1990.'Dependency and Rebellion:A Cross-National

Analysis', American Sociological Review 55

(August):540-559.

Brass, Paul, 1985. Ethnic Groupsand the State.

Totowa, NJ: Barnesand Noble.

Brockett, Charles D., 1992. 'MeasuringPolitical

Violence and Land Inequality in Central

America',AmericanPoliticalScienceReview86

(March): 169-176.

Brown, Cliff & Terry Boswell, 1995. 'Strike-

breakingor Solidarity n the GreatSteelStrikeof 1919: A Split LaborMarket,Game Theo-

retic, and QCA Analysis',American ournalof

Sociology100 (May): 1479-1519.

Brown, Cliff & Terry Boswell, 1997. 'Ethnic

Conflict and Political Violence: A

Cross-National Analysis', Journal of

Political and Military Sociology 5 (Summer):

111-130.

Connor, Walker, 1994. Ethnonationalism:The

Quest for Understanding. Princeton, NJ:

PrincetonUniversity Press.

Diamond, Larry;Juan J. Linz & Seymour M.

Lipset, 1995. Politics n DevelopingCountries:

Comparing Experiences with Democracy.

Boulder,CO: LynneRienner.

Dixon, WilliamJ.;EdwardN. Muller & MitchellA. Seligson, 1993. 'Inequalityand Political

ViolenceRevisited:Response oWang',Ameri-

can Political Science Review 87 (December):

983-993.

Douglass, William A., 1988. 'A Critique of

Recent Trends in the Analysis of Ethno-

nationalism', Ethnic and Racial Studies 11

(April):192-206.

Ellingsen,Tanja,2000. 'ColorfulCommunity or

Ethnic Witches' Brew? Multiethnicity and

Domestic Conflict During and After the Cold

War',JournalofConflictResolution 4 (April):

228-249.

Ellingsen, Tanja & Nils PetterGleditsch, 1997.

'Democracyand ArmedConflict in the Third

World', in Ketil Volden & Dan Smith, eds,

Causesof Conflictin the Third World.Oslo:

North-South Coalition & PRIO (69-81).

Flanigan, William F & Edwin Fogelman, 1970.

'Patternsof PoliticalViolence in Comparative

HistoricalPerspective',Comparative olitics3

(October):1-20.

Friedrich,RobertJ., 1982. 'In Defense of Multi-plicativeTermsin Multiple RegressionEqua-

tions',American ournalof PoliticalScience26

(November):797-833.

Glazer, Nathan & Daniel P. Moynihan, 1975.

Ethnicity:Theoryand Experience.Cambridge,

MA: HarvardUniversityPress.

Greeley,A., 1974. Ethnicity n the United States:

A Preliminary Reconnaissance.New York:

Wiley.

Gurr,Ted R., 1993. 'Why Minorities Rebel:A

Global Analysis of Communal Mobilizationand Conflict Since 1945', InternationalPoliti-

cal ScienceReview14 (April):161-201.

Gurr,Ted R., 1994. Minoritiesat Risk.Washing-

ton, DC: United StatesInstituteof Peace.

Gurr, Ted R. & BarbaraHarff, 1994. Ethnic

Conflict in World Politics. Boulder, CO:

Westview.

Henderson, Conway W, 1991. 'Conditions

Affecting the Use of Political Repression',

Journal of Conflict Resolution35 (March):

120-142.

565

Page 21: 5 dec III

8/3/2019 5 dec III

http://slidepdf.com/reader/full/5-dec-iii 21/22

566 journal of PEACE RESEARCH

Hibbs, DouglasA., 1973. MassPoliticalViolence.

New York:Wiley.

Horowitz, Donald L., 1985. Ethnic Groups n

Conflict. Berkeley, CA: University of Cali-

fornia Press.Horowitz, Donald L., 1994. 'Democracy in

Divided Societies', n LarryDiamond & Marc

F. Plattner,eds, Nationalism,Ethnic Conflict,

and Democracy. Baltimore, MD: Johns

Hopkins UniversityPress(35-55).

Huntington, SamuelP.,1997. 'Democracy or the

Long Haul', in Larry Diamond, Marc F.

Plattner,Yun-Han Chu & Hung-Mao Tien,

eds, Consolidatinghe ThirdWaveDemocracies.

London: Johns Hopkins University Press

(3-13).Jaggers,Keith & Ted R. Gurr, 1995. 'Tracking

Democracy'sThird Wave with the Polity III

Data',Journalof PeaceResearch 2 (Novem-

ber):469-482.

King, Garry, 1989. 'Event Count Models for

InternationalRelations: Generalizationsand

Applications', International Studies Quarterly

33 (June):123-147.

Lipset, SeymourM., 1959. 'Some Social Requi-

sites of Democracy:Economic Development

and Political Legitimacy',American Political

ScienceReview53: 69-105.McKay,James,1982. 'AnExploratorySynthesisof

Primordialand MobilizationistApproaches o

Ethnic Phenomena',Ethnicand RacialStudies

5 (October):395-420.

Mansfield, Edward D. & Jack Snyder, 1995.

'Democratization and the Danger of War',

InternationalSecurity 0 (Summer):5-38.

Melson, Robert & Howard Wolpe, 1970.

'Modernizationand the Politics of Commu-

nalism: A Theoretical Perspective',American

Political Science Review 64 (December):1112-1130.

Mousseau, Michael, 2000. 'Market Prosperity,

Democratic Consolidation, and Democratic

Peace', Journal of Conflict Resolution 44

(August):472-507.

Muller, Edward N., 1985. 'Income Inequality,

Regime Repressiveness,and Political Vio-

lence', AmericanSociologicalReview50 (Febru-

ary):47-61.

Muller,EdwardN., 1988. 'Inequality,Repression,

and Violence: Issues of Theory and Research

Design', American Sociological Review 53

(October):799-806.

Muller,EdwardN. & MitchellA. Seligson,1987.

'InequalityandInsurgency',AmericanPolitical

ScienceReview81 (June):425-451.Newman, Saul, 1991. 'Does Modernization

BreedEthnicPoliticalConflict?',WorldPolitics

43 (April):451-478.

O'Donnell, Guillermo;PhilippeC. Schmitter &

LaurenceWhitehead, 1986. TransitionsromAuthoritarianRule: Prospectsor Democracy.

Baltimore, MD: Johns Hopkins UniversityPress.

Olzak, Susan & Joanne Nagel, 1986. Competi-tiveEthnicRelations.New York:Academic.

Rae, Douglas& Michael

Taylor,1970. The

Analysis of Political Cleavages.New Haven,

CT: YaleUniversityPress.

Rothschild,Joseph, 1981. Ethnopolitics: Con-

ceptualFramework.New York:Columbia Uni-

versityPress.

Rummel, RudolphJ., 1995. 'Democracy,Power,

Genocide, and MassMurder', ournalof Con-

flict Resolution 9 (March):3-26.

Rummel, Rudolph J., 1997. 'Is Collective Vio-

lence Correlated with Social Pluralism?',

Journal of Peace Research 34 (May):

163-175.

Rustow,Dankwart,1970. 'Transitions o Democ-

racy:Towarda Dynamic Model', Comparative

Politics2 (April):357-370.

Schock, Kurt, 1996. 'A Conjuctural Model of

Violent PoliticalConflict', Journalof ConflictResolution 0 (March):98-133.

Scott, George M., 1990. 'AResynthesisof the Pri-

mordial and CircumstantialApproaches to

EthnicGroupSolidarity:Towardsan Explana-

tory Model', Ethnic and Racial Studies 13

(April): 147-171.Sigelman,Lee & Miles Simpson, 1977. 'ACross-

National Test of the LinkageBetween Econ-

omic Inequality and Political Violence',

Journal of Conflict Resolution 21 (March):

105-128.

Summers,Robert & Alan Heston, 1988. 'A New

Set of International Comparisons of Real

Product and Price Levels Estimates for 130

Countries, 1950-1985', ReviewofIncomeand

Wealth 4 (March): 1-25.

Taylor,CharlesL. & Michael C. Hudson, 1972.

volume 8 / number / september001

Page 22: 5 dec III

8/3/2019 5 dec III

http://slidepdf.com/reader/full/5-dec-iii 22/22

Demet Yalcin Mousseau DEMOCRATIZING WITH ETHNIC DIVISIONS

WorldHandbookof Political and Social Indi-

cators,2nd edn. New Haven, CT: Yale Uni-

versityPress.

Taylor,CharlesL. & DavidA.Jodice,1983. World

HandbookofPoliticaland SocialIndicators,rdedn. New Haven, CT: YaleUniversityPress.

Wang, T. Y, 1993. 'Inequality and Violence

Revisited',AmericanPoliticalScienceReview87

(December):979-983.

Ward,MichaelD. & KristianS. Gleditsch,1998.

'Democratizingfor Peace',AmericanPolitical

ScienceReview92 (March):51-62.

Weede, Erich, 1981. 'Income Inequality,

Average Income and Domestic Violence',

Journalof ConflictResolution 5 (December):639-654.

Ziegenhagen,EduardA., 1994. PoliticalConflict,PoliticalDevelopment, nd PublicPolicy.West-

port, CT: Praeger.

AppendixA: DescriptiveStatisticsof the Variables

Variable Mean Std.dev. Minimum Maximum

Dummydeaths 0.08 0.27 0.00 1.00

Ethnicheterogeneity 0.35 0.28 0.00 0.93

Democracy -0.70 7.66 -10.00 10.00

Democracy(+ll)2 164.68 181.53 1.00 441.00

Democratization -0.31 4.78 -17.00 19.00

Development 8.64 1.78 2.49 12.10

Development2 77.88 29.29 6.20 146.41

Ethnicseparatism 0.33 0.47 0.00 1.00

Regimematurity 22.80 21.34 0.00 82.00

Democraticmaturity 29.43 30.23 0.00 92.80

Lnpopulation

9.11 1.46 4.92 13.83

Ethnicity*democracy2 47.13 74.07 0.00 354.40

Ethnicity*democratization -0.23 2.15 -13.04 12.17

Ethnicity*development2 24.48 21.46 0.00 101.77

Peaceyears 9.94 8.85 0.00 34.00

Cubicspline1 -719.55 1010.72 -4224.00 0.00

Cubicspline2 -1438.32 2158.20 -9135.00 0.00

Cubicspline3 -1939.56 3190.35 -13920.00 0.00

N = 3172

DEMET YALCIN MOUSSEAU, b. 1964,

PhD in PoliticalScience(SUNY-Binghamton,

1999); Assistant Professor, Department of

International Relations, Koc University,

Istanbul (1999- ). Current research nterests:

the impactsof political and economic factors

on domestic politicalconflict and ethnic vio-

lence.

567