Professional Documents
Culture Documents
Reynal-Querol (2002) Ethnicity, Political Systems, and Civil Wars
Reynal-Querol (2002) Ethnicity, Political Systems, and Civil Wars
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 support@jstor.org.
Sage Publications, Inc. is collaborating with JSTOR to digitize, preserve and extend access to The Journal of
Conflict Resolution.
http://www.jstor.org
MARTAREYNAL-QUEROL
Institutd'Analisis Economic (IAE), CSIC
Studying social conflict has been considered an issue of political science for
decades. Yet we are recently observing an increasingnumberof contributionsfrom
economics. Partof this literaturehas studiedthe effect of economic and social factors
on the probabilityof civil war. Collier and Hoeffler (1998) investigatethe generic
causes of civil wars with special emphasis on economic factors. Using the index of
ethnolinguisticfragmentation(ELF),they find thatmore fragmentedsocieties arenot
moreproneto civil warthanthe rest,butthe dangerof civil warincreaseswhen society
achievemidlevel valuesof the indexELF.However,afterthose initialfindings,Collier
and Hoeffler (2000) arguethatconflicts arefar more likely to be causedby economic
opportunitiesthanby grievance.However,the studyof the causes of civil wardeserves
additionalstudies,especially on the ethnic and political issues thatmay influence the
incidence of civil war.Ouranalysis,based on these previousstudies,is focused on the
ethnic andpoliticalcauses of civil war,using a theoreticallybased index of latentcon-
flict and giving special importanceto the religious dimensions of ethnicity,using a
new dataset thattriesto overcomethe commoncriticismof the WorldChristianEncy-
clopedia (Barret1982) data.Moreover,most of the literaturehas considereddemoc-
racy the only political variablethat may affect the probabilityof civil war.Here we
arguethatthe level of inclusivenessof the system, togetherwith the level of democ-
racy, matters. These results clarify the role of democracy itself. Elbadawi (1999),
based on the analysisof the resultsof some comparativestaticexperiments,finds that
ethnically polarized societies have a higher risk of suffering a civil war. Ellingsen
(2000) finds that the differentaspects of multiethnicity(the size of the largestgroup,
the numberof groups,the size of the minoritygroup,and ethnic affinities)are impor-
tant in explaining domestic conflicts. Her results give an importantrole to political
regimes and socioeconomic variablesto reducethe level of conflict.
THEROLEOF RELIGION,ETHNICITY,
ANDIDEOLOGY
1. The complete version of this article includes an examinationof the causes of revolutionarycivil
wars.
HOWTO CAPTURESOCIALCONFLICT
To show the importanceof ethnicdiversityin ethnic civil wars,at least two impor-
tantaspectshave to be addressed.First,it is necessaryto clarifythe conceptof ethnic-
ity to select the variablesthat captureethnic diversity and analyze which of these
dimensions of ethnicity are more importantto explain social conflict. Following
Horowitz(1985), the inclusiveconceptionof ethnicitycoversdifferencesidentifiedby
skin color, language,religion, or some otherattributeof common origin. Because of
datarestrictions,in this study we deal with religion and languagedifferences.More-
over,as shown in the previoussection,religionis the most importantethnicdimension
in explaining social conflict.
Second, we have to summarizethe informationfrom each of these dimensions in
one syntheticindex. The most importantissue is the appropriateprocedureto use to
summarizein an index the concept of social conflict. Is it social fragmentationor
social polarizationthatmakesethnictensionsstronger?Thereis no easy answerto this
question.First, we need to study the mechanismthroughwhich these groupsinteract
and analyzein which situationstensionsarisemoreeasily. Second, the threevariables
that define ethnicity (language,religion, and color) can work in differentways. For
instance,the tensions caused by languagedifferencesand the loss of communication
that they generate can emerge in a situation very different from those generatedby
religion.
The measuresusedregularlyin the empiricalliteratureto quantifyethniccharacter-
istics are fragmentationindices, even thoughthereis no theoreticalsupportfor these
kindsof indices. However,rent-seekingmodels suggest thatpolarizationmeasuresare
more appropriatethanfragmentationindices to capturesocial conflict. Any index of
polarizationpoints out thatthe situationthatleads to the point of maximumtension is
when there are two social groups with the same size. These kinds of measuresdiffer
from an index of fragmentationbecause the index of polarizationcapturesto what
Figure 1
the convex region. Whatis the intuitionbehindthis result?In the first case, even that
transferimplies thatthe distributionis more unequalin the new situation:one of the
small groups is largerin respect to the big group, which means thatwe are closer to
polarization.In the secondcase, the transferimplies thatone of the big groupsbecame
smaller,and thereforethe new situationis less polarized.Notice thatthe resultsimply
thatthis index does not satisfy the propertiesof the Lorenzcurveaboutconcavity.In a
Lorenzcurve,this effect of moving people between small or big groupsis the same. It
is importantto notice anotherdifferencecomparedwith the Lorenzcurvedominance:
our measureis global andthe Lorenzcurveis not. Althoughthe Lorenzcriteriaestab-
lish the impacton inequalityof a local transferindependentlyof the shapeof the restof
the distribution,in our case, the effect on polarizationof the transferof population
from one groupto anothercannotbe establishedwithoutknowingthe entiredistribu-
tion. This is a propertythatalso has the measureof polarizationproposedby Esteban
and Ray (1994).
P*(n)=KE E IC
l+a~ jd=IRC2,
i=l j=1
Using this application, the index can be written in the following way:
1
IRC2 = n + ", which is a strict convex function for a > 1. For each possible a,
thereis a differentpolarizationfunction.In practice,we will choose a = 1.5.6The fact
thata is not determinedfromthe theoreticalmodel makesthe use of this secondpolar-
ization index just a matterof comparisonwith respect to the Reynal-Querol(2001a)
index.7The majordifferencebetweenthese two indices is in the form of the function;
whereasIRC1 has a convex andconcavepartas explainedabove,IRC2 is strictlycon-
vex for a > 1.
THEROLEOF POLITICALINSTITUTIONS
This study also analyzes the political characteristicsthat can preventor promote
civil wars. The literaturehas consideredonly repressionas a political characteristic
thatcan affect notonly economic developmentbutalso civil wars.CollierandHoeffler
(2000), Sambanis(2001), Hegreet al. (2001), andEllingsen (2000) find thatmidlevel
democraciesare more proneto civil war than high-level democraciesand high-level
autocracies.Reynal-Querol(2001b) shows theoreticallyand empiricallyhow coun-
tries with alternativepolitical systems have differentprobabilitiesof experiencinga
civil war.She develops a simple theoreticalmodel thatcapturesthe basic relationship
between the political system and rebellion.Accordingto the model, the proportional
system turnsout to have a lower probabilityof rebellionthanthe majoritariansystem.
The intuitionbehindthis resultis thatin the proportionalsystem, the opportunitycost
of rebellion is higher than in a majoritariansystem. The main idea is that the more
inclusive the political system, the higherthe opportunitycost of rebellion,and there-
fore the lower the probabilityof rebellion.Fromthe empiricalanalysis,she shows how
the structureof a politicalsystem is an importantmechanismthatcan affectthe proba-
bility of civil war in a democraticsystem. The result clarifies the role of democracy
itself. She observesthatsome countrieswithhigh levels of democracysufferperiodsof
violence; therefore,havinghigh levels of civil libertiesand freedomdoes not seem to
protectthem againstviolence. She arguesthatthe representationsystem of the voters
in governmentis moreimportantthanthe level of democracyper se. Empirically,she
finds thatthe more inclusive the system, the smallerthe probabilityof civil war.
Moreover,the level of repressionalso influences the opportunitycost of rebellion.
The morepoliticalrightsandcivil libertiesthe countryhas, the higherthe opportunity
cost of rebellion,andthereforethe lowerthe probabilityof groupsto rebel.However,a
little freedom is needed to let groups organize.This is the intuitionthatjustifies the
findings thatmidlevel democraciesare more proneto suffercivil wars.
ETHNICDIVERSITYANDPOLITICALSYSTEMS:
PLURALITYVERSUSCONSOCIATIONALISM
EMPIRICAL ANALYSIS
THEDATA
cases: the animists, the Christians,and the Muslim groups.'4The index of animist
diversity(AD) capturesthe numberof followers of animistcults in each country.'5To
avoid the level effect generatedby this variable,the regressionsthatincludethis index
consider also as an explanatoryvariablethe total populationof the country.
Therearesome differencesbetweenthe dataon religionsused by Barro(1997) and
Reynal-Querol(2000a). Barro'sdataset, whichcomes fromthe WorldChristianEncy-
clopedia (Barret1982), has the advantageof being a time series, providinginforma-
tion for 1970, 1975, and 1980. However,this source has several shortcomings.First,
and probablythe most important,the data do not consider the possibility of double
practice, very common in sub-SaharanAfrica and Latin America countries. Com-
paringthisto the othersourcesof information,we realizethe dataarebiasedtowardthe
Christianreligion. A clear exampleis the case of Zaire,wherethe distributionof reli-
gions is consideredto be similarto Spainor Italy.The distributionof religious groups
between 1970 and 1980 does not change in many countries.There are only approxi-
mately 17 countriesthat recordchanges in proportions.But those changes occur in
countrieswherethereis doublepractice,andtheyusuallyimply an increasein the per-
centage of Christiansand a reduction in the size of animist followers. Moreover,
Barro'sdatado not representwith sufficientdetail all the religions. We have observed
in the WorldChristianEncyclopedia(Barret1982) thatthereis too muchinertiain the
growthrate of some religions, mainly Catholicism.
These shortcomingsmakes the use of Barro's(1997) data somehow problematic.
However,the literatureused Barro'sdata set to compute measuresof religious frag-
mentation. Therefore, we also consider Barro's data when comparing the results.
Because of these shortcomings,we use datain Reynal-Querol(200 la) to constructthe
index of religious polarization.
We wanted to use alternativeethnic variables, as in Vanhaven(1999), mainly
because they capturethe threedimensionsof ethnicity.However,therearetwo impor-
tantshortcomingsin the use of these data.First,theydo not differentiateamongChris-
tians and animist cults, which are very typical in African countries. Second, even
thoughhe reportsthe percentageof the biggest ethnicgroup,he does not providedata
on the size of the othersmall groups,which makesthe constructionof a polarization
measureimpossible.
Another source of data we would like to use is Ellingsen (2000). However, like
Vanhaven(1999), she reportsthe percentageof the biggest ethnic groupbutdoes not
providedataon the size of the othersmallergroups,which makesthe constructionof a
polarizationmeasureimpossible.
14. We do not considerotherreligions because they do not have the featuresthat make these three
groupsinternallyproblematic.Forthe Muslimsandthe Christians,we applythe same measureof polariza-
tion, IRC1.We use the proportionsof differentkind of Christiansfor the polarizationmeasureof Christians
anddataon the votes forMuslimpoliticalpartiesto calculatea polarizationmeasureforthe Muslimgroup.
15. Theanimistgroupsaredifferent.Theyincludea variablenumberof traditionalreligionstypicalof
primitivesocieties. Religion in these societies pervadesall social domains.Therefore,the argumentof the
interactionbetweenreligionsexplainedby therent-seekingmodelsdoes not work.The existenceof so many
differentkinds of organizationsbased on differentbeliefs makes the communicationacross these groups
more difficult. The degree of communicationloss depends on the numberof animist cults. See Reynal-
Querol (2000a) for a complete descriptionof the treatmentof animist cults.
THEECONOMETRIC
SPECIFICATION
REGRESSIONRESULTS
TABLE 1
Logit Pool Estimationfor the Incidenceof Ethnic Civil Warfrom 1960-1995:
Religious Polarizationand Animist Diversity
Model 1 2 3 4 5 6
who reportno effect of polarization.The differencein the resultsis due to the fact that
they analyzethe causes of all civil wars;moreover,the datathey use on religionscome
from Barro (1997), which, as we mentionedbefore, have importantshortcomings.
Moreover,Sambanis(2001), using Vanhaven(1999) data,finds thatethnic heteroge-
neity is linearlyand positively correlatedwith the onset of ethnic civil war.However,
becausethe variablehe uses capturesthe inclusivedefinitionof ethnicity,it is difficult
to comparethe resultshe finds with the effect of religious polarizationfound here.
The resultsin Table1 show the importanceof the religiouscompositionof the soci-
ety as an importantsocial cleavage that affects the incidence of ethnic civil war.'6
These resultsareimportantfor threereasons.First,we considerpolarizationmeasures
insteadof fragmentationmeasuresto capturelatentconflicts, following the results of
the rent-seekingliterature.Second,we use the religiousdimensionof ethnicityinstead
of the linguistic one, as an importantcause of conflict, following some political scien-
tist theories.Finally,we consideranimistdiversity,very typical in sub-Saharancoun-
tries, as a special case of religious conflict.
TABLE2
Logit Pool Estimationfor the Incidenceof Ethnic Civil Warfrom 1960-1995:
Religious and LinguisticFragmentation
Model 1 2 3 4
TABLE3
Logit Pool Estimationfor the Incidenceof Ethnic Civil Warfrom 1960-1995:
Religious Polarizationversus Religious Fragmentation
Model 1 2 3 4
the effect of religious diversityon ethnic civil war.It seems thatthe fragmentationof
the society on religiousgroupsis preventingcountriesfromviolence ratherthaninduc-
ing them to conflict. When analyzing the causes of civil wars, Collier and Hoeffler
(2000) also find evidence that religious fragmentationmakes countriessafer.More-
over, Elbadawiand Sambanis(forthcoming)find a quadraticinteractiontermof reli-
gious andethnicdiversitythatis negativelyassociatedwiththe incidenceof civil war.
TABLE4
Logit Pool Estimationfor the Incidenceof EthnicCivil Warfrom 1960-1995:
Political Rights and Civil Liberties
Model 1 2 3 4 5 6
TABLE5
LogitPoolEstimation
fortheIncidenceof EthnicCivilWarfrom1960-1995:
PoliticalSystem
Model 1 2 3 4 S 6 7 8
TABLE5 Continued
Model 1 2 3 4 5 6 7 8
Policy Implications
In this section, we analyze whetherin religiously polarized and animist diverse
societies, the existence of a political system with a high level of representationof the
populationreducesthe effect of this latentconflict on the incidenceof ethniccivil war.
Table 6 shows the results of this analysis.
Model 1 2 3 4 5 6 7
CONCLUSIONS
This study has analyzed the social and political causes of ethnic civil war.
Threeimportantfindingsarerobustfor the use of differentestimationproceduresand
specifications.
The firstis the importanceof religiouspolarizationandanimistdiversityin explain-
ing the incidenceof ethniccivil warno matterwhatothermeasuresareincluded.These
resultscorroboratethe resultsof the literatureon rent-seekingmodels thatstudy social
conflict andthe importanceof the distributionof the groupsin a countryas a key ele-
ment for understandingthe causes of social conflicts. Second, we find thatreligious
polarizationis moreimportantas a social cleavagethatcan develop into civil warthan
are linguistic differences. This result corroborateswhat Huntington(1996) claims
aboutthe importanceof religiousdifferencesto explaindomestic conflicts. Third,we
find thata consociationaldemocracyis a politicalsystem thatsignificantlyreducesthe
incidence of ethniccivil war.Mainly,it reducesthe probabilityof ethniccivil wargen-
eratedby religious polarization.
The study of the causes of civil war deserves additionalstudies, especially of the
ethnicandpoliticalissues thatmay influencetheincidenceof civil war.Ouranalysis-
based on the study by Collier and Hoeffler (2000), who argue that conflicts are far
more likely to be causedby economic opportunitiesthanby grievance-is focused on
the ethnicandpoliticalcauses of civil war.We use a theoreticallybasedindex of latent
conflict and give special importanceto the religious dimensions of ethnicity,using a
new dataset thattriesto overcomethe commoncriticismto the WorldChristianEncy-
clopedia (Barret1982) data.Moreover,herewe arguethatwhatmattersarethe level of
inclusiveness of the system and the level of democracyinsteadof the level of democ-
racy per se.
These results are importantfor four reasons.First, we considerpolarizationmea-
sures insteadof fragmentationmeasuresto capturelatentconflict. Second, we use the
religious dimensionof ethnicityinsteadof the linguistic one as an importantcause of
conflict. Third,we consideranimistdiversity,verytypicalin sub-Saharancountries,as
a special case of religious conflict. Finally, most of the literaturehas considered
democracyas the only politicalvariablethatmay affectthe probabilityof civil war,and
even the empiricalevidence aboutthis is very week. We arguethatwe need to control
not only for democracybutalso for the political system. The resultsclarifythe role of
democracyitself.
APPENDIX
REFERENCES