Full Terms & Conditions of access and use can be found athttp://www.tandfonline.com/action/journalInformation?journalCode=cper20

Download by: [Koc University] Date: 10 February 2016, At: 10:25

Peace ReviewA Journal of Social Justice

ISSN: 1040-2659 (Print) 1469-9982 (Online) Journal homepage: http://www.tandfonline.com/loi/cper20

Islam, Nationalism, and Kurdish Ethnopolitics inTurkey

Tim Jacoby & Hüsrev Tabak

To cite this article: Tim Jacoby & Hüsrev Tabak (2015) Islam, Nationalism, and KurdishEthnopolitics in Turkey, Peace Review, 27:3, 346-353, DOI: 10.1080/10402659.2015.1063379

To link to this article: http://dx.doi.org/10.1080/10402659.2015.1063379

Accepted author version posted online: 31Jul 2015.

Submit your article to this journal

Article views: 201

View related articles

View Crossmark data

Peace Review: A Journal of Social Justice, 27:346–353Copyright C© Taylor & Francis Group, LLCISSN 1040-2659 print; 1469-9982 onlineDOI: 10.1080/10402659.2015.1063379

Islam, Nationalism, and KurdishEthnopolitics in TurkeyTIM JACOBY AND HUSREV TABAK

According to official figures, the conflict between the Turkish state and pro-Kurdish insurgents (led by the Kurdish Workers’ Party—Partiya KarkareniKurdistan, PKK) killed more than 5,000 soldiers and injured another 11,000between 1984 and 1999. Civilian causalities were similar—5,000 dead and10,000 hurt—while more than 23,000 rebels lost their lives and over 3,000their liberty at an overall economic cost of almost $15 billion. Since then,there have been intermittent ceasefires and a prolonged period of politicalliberalization, but the violence has rumbled on with more than 700 fatalitiesin the last four years.

The state’s principal response has traditionally been to deploy consid-erable military force in defense of what it saw as the unitary character of therepublic. Mainstream religious commentators have generally endorsed theseapproaches, with the result that the “Kurdish Question,” as it is often called,has been couched in terms of an indivisible nation-state to which minorityrights have been marginal.

During the 1990s, Kurdish political parties were subject to banning or-ders and under emergency legislation; efforts were made to bring the

predominantly Kurdish-speaking provinces in the southeast closer to the statethrough an emphasis on what was known as the Turkish–Islamic synthesis.This took many forms—ranging from the mobilization of Hezbollah contra-guerrillas to a strong emphasis on the heresy of the PKK’s Marxism—butit was generally based on a denial of minority identity in favor of an over-arching, apolitical discourse on brotherhood. State authorities, including themilitary and the Diyanet (Turkey’s official religious authority), called on peo-ple to unite behind the government and emphasized the universal, a-nationalexegesis of the Quran and sunnah. In all, it was a strategy that both helped tocurtail the emergence of a pro-Kurdish discourse and to weaken the electoralperformance of local political parties in the southeast.

In 2002, though, a new government, led by the Justice and DevelopmentParty (Adalet ve Kalkınma Partisi, AKP), swept away the two parties thathad dominated the 1980s and ’90s and then won successive general elections

346

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:25

10

Febr

uary

201

6

ISLAM, NATIONALISM, AND KURDISH ETHNOPOLITICS IN TURKEY 347

in 2007 and 2011 (and looks set to win again in June 2015). With around40 percent of the party and its leader (and current President), Recep TayyipErdogan, previously members of the banned Refah party in the 1990s, the AKPwere generally drawn from well outside the traditional state nomenklatura.

Indeed, having been excluded from politics himself (and imprisoned forfour months in 1998) for reading a controversial section of a well-knownreligious poem in the south-eastern city of Siirt, Erdogan has steadfastlyopposed efforts to impose new bans on Kurdish political parties. His partycooperated with Kurdish deputies to return the AKP’s former foreign minister,Abdullah Gul, to the presidency in 2007 and to pass the legal amendment thatlifted the headscarf ban at universities in 2010. The result has been a moveaway from the Turkish–Islamic synthesis and, as writers such Murat Somerand Hakan Yavuz have pointed out, a greater acknowledgment of Kurdishpolitical and cultural interests within religious discourses.

In what follows, we will use Rogers Brubaker’s cognitive approach toethnicity to consider the changing approach that a number of leading religiousintellectuals have taken up in response to this shift in discourse. Arguing that inattempting to move away from the Turkish–Islamic synthesis by increasinglyrecognizing the cultural rights of minority groups in Turkey, they have tendedto endorse a perennial understanding of ethnicity, which has, inadvertently,weakened Islam’s universalist message. For many contemporary scholars ofreligion, this new acceptance of the Kurdish claim to be a national categoryleads to the idea that societal categorizations are real and substantial “things-in-the-world,” which are, in reality, only putatively present and need constantreification in order to have meaning. Such “groupism,” as Rogers Brubakercalls it, has three basic features. The first is to treat ethnic groups, nations,and races as though they are substantial entities to which interests and agencycan be attributed. Second, it is the tendency to see such groups as if theywere internally homogenous and externally bounded actors with commonpurposes. Finally, it represents the social world as a mosaic of such ethnic,racial, and cultural blocs. These strategies are used by what Brubaker calls“ethnopolitical entrepreneurs” to call a group into being.

I n Turkey, the tendency toward groupism among religious scholars haslong been debated. The Ottoman theologian Babanzade Ahmet Naim

(1872–1934), for instance, was very critical of the emerging Turkish repub-licanism in the build-up to the First World War, stating that nationalism isstrictly banned in Islam. He argued that its advocates were in the business ofcreating an alternative to the ummah as the foundational community structurefor Muslims worldwide. As the new Turkish nation-state emerged from theashes of empire, this view was, however, increasingly set aside.

In his response to Babanzade, the Turkish nationalist Ahmet Agayef(1869–1939), for example, maintained that the Quran refers to the inappropri-ateness of tribalism, not national consciousness. Quoting chapter 9, verse 97

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:25

10

Febr

uary

201

6

348 TIM JACOBY AND HUSREV TABAK

with the criticism of the Bedouins’ parochialism, he suggests that a key aim ofthe early Muslims was to achieve the unity of the Arab nation. Accordingly,to Agayef, the dissemination of Islam could not have been achieved withoutthe creation of the Arab national ideal, as embodied in the teachings of theProphet Muhammad.

A similar argument was presented by the founding father of Turkishnationalism, Ziya Gokalp (1876–1924). He pointed to Surat-al Hujurat (verse13) and its famous reference to different peoples being created so that eachmay come to know or recognize one another. He interpreted this to mean thatpeople should be familiar with others’ languages, thereby implying a worldof legitimate national differences—an idea reinforced by chapter 10, verse47 that revealed that God sent each of these nations or peoples a messenger.Gokalp thus concluded that Islam advocates modernize nationalism and itsaim of establishing states composed of homogenous nations, or the kind ofmosaic envisaged by Brubaker.

Agayef and Gokalp’s religiously backed perennial approach to ethnicitywas both instrumental to the founding of the Turkish Republic and key refer-ence points for the conservative nationalists who formed the constituent basisof the Turkish–Islamic synthesis during the 1990s. In these discourses, Islamis generally deployed as not only a justification for national consciousness, butalso as a singular civilization built, in part, by the Turkish “people.” This, sup-posedly, symbiotic relationship between an eternal Turkish nation (with deepprehistoric roots within the central Asian hinterland) and the emergence of agreat world faith is also the imagined basis for Turkey’s “exceptionalism”—itsunique combination of secular unity with irrepressible popular religiosity.

The key here is the idea that such an accommodation can only continueif all political vestiges of both religious practice and minority identity arecarefully subsumed beneath a banal emphasis on “Islamic brotherhood” inwhich dissent has traditionally been seen as simultaneously un-Turkish andun-Islamic. The Kurds were thus asked to go beyond ethnic identity and joinwith Turks in a fraternity in which it was, as one former PKK member putit, always been obvious which one is the abi (a colloquial expression used torefer to an older man).

Many Kurds have responded by campaigning for their contribution to thedevelopment of the Muslim world to be more fully recognized by the

Turkish authorities. The prominent Kurdish pro-Islamic politician Altan Tan,for instance, has long accused the Diyanet of betraying the “the law of broth-erhood” and maintaining a chauvinistically Turkish interpretation of the faith.Similarly, the Kurdish journalist _Ibrahim Sediyani has called for the Kurdishcivilization to gain a greater presence within the Diyanet’s textbooks, whichstill remain heavily focused on the cultures of Turks, Persians, and Arabs.

Zeki Sarıgil and Omer Fazlıoglu make a similar point when reportingthe opinions of the Kurdish local mullahs (mele) who frequently use their

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:25

10

Febr

uary

201

6

ISLAM, NATIONALISM, AND KURDISH ETHNOPOLITICS IN TURKEY 349

addresses to the Friday congregational prayer to accuse the state of favoringTurkish cultural superiority. One such cleric is quoted as commenting that,“they say we are brothers. . . . However, one brother’s language is permit-ted [Turkish], while the other’s is banned. What kind of a brotherhood isthis?” Others have apparently pointed to an ongoing emphasis on Hanefi ju-risprudence (the madhab of the majority of Turks) to the detriment of Shafipractices, followed by the bulk of Kurds.

After the election of the AKP, however, a wide group of Islamic intellec-tuals have begun to move away from the outright denial of non-Turkish cul-tural practices and towards an acknowledgment of Kurdish ethnicity. In otherwords, the ubiquitous “Islamic brotherhood” discourse of the Turkish–Islamicsynthesis is no longer the principal basis for religious scholars’ proposals totransform the conflict between the state and the PKK. They have rather ex-tended Islamic arguments to endorse the idea that the country’s Kurds, similarto Turks, constitute a unitary community with undisputable ethnic rights—a“group” in Brubaker’s terms. These generally propose that the denial of suchrights by the intermittently oppressive and assimilative policies of the sec-ular state is a result of moving away from Islam’s core message, ratherthan a result of the perennial construction of two distinct “groups” withinTurkey.

The journalist and theologian Ali Bulac, for instance, has recently arguedthat, since language and skin color (qavm) are among the ayah of Allah, thenations of the world cannot be deprived of freely expressing their culturesand communal rights. Similarly, the cleric Mustafa _Islamoglu holds that thesubmission to a race does not constitute a contradiction of Islam’s universalitybecause such differences are defined in the Quran. Only the chauvinistictendencies of pre-Islam, such as racism, tribalism, and uncivil nationalism are,he continues, to be considered the work of the devil. Along the same lines,Faruk Beser from the Union of Turkish Ulema (an association that worksclosely with Diyanet), has argued that, since it is Allah who created socialidentities, subordinating the Kurds through imagining a Turkey with a singlerace, nation, and language is, from the religious point of view, inappropriate.

Professor of Islamic Law Hayrettin Karaman presents a comparable argu-ment by pointing to Turkey’s minorities’ legitimate right to protect their

distinctiveness based on the dual markers of difference—biology (skin tone)and culture (language). The latter may rightfully inform religious practicewithin a diverse, yet universal, faith he suggests. This, he continues, is the dis-tinction between what God calls a millet in the Quran and the ummah, definedby Karaman as the followers of a religion and prophet. Such a distinctionis permissible, he argues, even if the former evolves to constitute a distinctnation. In this regard, he concludes, it is not against Islam when a group ofpeople, affiliated by ethnicity or cultural practice, comes together to define

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:25

10

Febr

uary

201

6

350 TIM JACOBY AND HUSREV TABAK

itself collectively and pursue the establishment of a coterminous state—solong as this is undertaken beneath a Muslim supra-identity.

Even the Diyanet, which has long preferred to focus solely on the need forbrotherhood and the harm caused by conflict between Muslims, has startedto recognize the ethnic differences between Turks and Kurds. The currentpresident, Mehmet Gormez, for instance, stated in 2013 that it is said in theQuran that the difference in your language is among the ayahs of Allah.Denying this, he continues, is no different from denying any other ayah ofAllah and that excluding, or being disdainful of, a skin-color, race, or languageis an offense against humanity and Allah.

In doing so, Gormez and the mainstream religious ulema in Turkeyare effectively “bringing a group into being,” as Brubaker might have putit. In keeping with his three characteristics of groupism, these discoursesrest on the premise that nations and races are substantial entities that areperennially present in Quranic terms and that remain the locus of socialagency. Within these, there is also a foundational assumption that the Kurdish“nation” is internally consistent—in other words, that the Kurds are, first,universally conscious of their national identity and that they are, second,uniformly committed to a project of collective self-realization.

The resultant “group,” which in the Turkish case is likely to rest on someform of future political devolution of power to the predominantly Kurdish

south-east of the country, is thus placed alongside other Arab, Turkish, andPersian “things in the world” as part of the kind of mosaic of ethnic/culturalblocs that Brubaker finds in the nationalisms of the West. By stressing theprimordial roots of these through an emphasis on the role of God in theiremergence, the acceptance of sub-ummah groups is rendered a kind of a reli-gious duty, thereby disabling counterarguments and extending the immanentpower of what Brubaker calls “ethnopolitical entrepreneurs.”

Indeed, such a shift in emphasis has certainly proven useful for Kurdishnationalists. The PKK has begun to move away from its traditional discourseon Islam being responsible for spreading linguistic and cultural subordination.The idea that, in the words of its founder, Abdullah Ocalan, religion has had “ananaesthetic effect” on Kurdish speaking communities in the region (therebyhelping to facilitate the cultural hegemony of the Arabs and Turks) has startedto lose ground.

The fact that, by the 2004 local elections, the AKP’s appeal to popularreligiosity was proving more popular than the secularism of the PKK’s politicalwing in the majority of south-eastern provinces (indeed, its national vote fellfrom 2 million in 2002 to 1.6 million in 2007 and, by the last general electionsin 2011, it continued to lose at least half its wards to the AKP in key Kurdishprovinces such as Diyarbakır, Bitlis, Mus, Van, and Mardin) has promotedOcalan’s organization to move closer still to the discourse of the theologians.

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:25

10

Febr

uary

201

6

ISLAM, NATIONALISM, AND KURDISH ETHNOPOLITICS IN TURKEY 351

The result has been a gradual change in the PKK’s approach. Startingin the 1990s with Ocalan’s book entitled A Revolutionary Approach to theQuestion of Religion (Din Sorununa Devrimci Yaklasım) and continuing withthe establishment of “ethnicized” religious organizations, such as the Unionof Kurdistan Patriot Imams (Kurdistan Yurtsever _Imamlar Birligi) and theIslamic Union of Kurdistan (Kurdistan _Islami Hareketi), the PKK concededthat it might be necessary to use religion as part of what it called the anti-colonial struggle. In the 2000s, the PKK begun organizing Islamic funeralservices and mawlids for guerrilla “martyrs,” imams began to appear at itspublic events more often, congregational prayers were organized outside state-controlled mosques as a form of protest, and a greater emphasis was placedon the commemoration of religious festivals.

More recently, Ocalan convened a 2014 “Democratic Islam Congress”under a banner taken from chapter 30, verse 22 of Quran: The differ-

ence of your languages and colors is among the ayah of Allah. In his vicariouswelcome address, he referred to participants as his “brothers in Islam” andconcluded with a discussion of God-given guarantees for the rights of commu-nities to use their mother tongue and to preserve their national culture. Otherprominent figures, such as the Kurdish MPs Altan Tan and Ahmet Turk, con-curred by arguing that Kurds are not only more devout Muslims than theircompatriots, but also the true indigenes of the Anatolian heartland.

This shift toward increasingly using Islam to legitimize ethno-nationalistmessages rests partly on the old, and powerful, idea that a distinct and eter-nal Kurdish “people” gave up their right to become a nation for the sake ofreligious unity after the collapse of the Ottoman Empire. Much is made ofthe assertion that the Kurds embraced Islam at least two centuries earlier thanthe Turks and became one of the first “nations” to institutionalize Muslimgovernance after the divinely chosen Arabs. Ocalan has argued that this pri-mordial “fact” of history has been deliberately obscured by the great empiresthat since colonized Kurdistan. In a meeting in February 2013, he asked AltanTan to unearth and disseminate the “hidden Islam” that he believes lies be-neath the cultural facade imposed by these imperial forces. The overall aim,he explained, is to remove these “adulterations,” return to a more “authentic”moral basis for society, and, perhaps most importantly, to restore the unity ofthe Kurdish nation.

In conclusion, the construction of the Kurdish “group” is, in reality, arhetorical obligation for the Islamic intellectuals who have already acceptedthe presence of a Turkish group. It is impossible to have one without the other.The representative state’s attempts to deny the existence of Kurds was always afig-leaf for what was a process of “otherization” and domination—somethingonly possible once the contours of the target “group” had been identified anddelimited. This is now acknowledged by Bulac, _Islamoglu, Karaman, and

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:25

10

Febr

uary

201

6

352 TIM JACOBY AND HUSREV TABAK

others (and even partially recognized by the Diyanet), but such an acceptanceis, as Babanzade was pointing out, only likely to increase the power of whatBrubaker calls the “ethnic entrepreneurs” for whom religious identity hasbecome a key element of legitimately calling a group into being.

As such, it seems unlikely that simply extending a nationalist rubric tonon-Turkish groups will prove a viable means to transform the country’s

persistent civil war, which continues to rumble on mostly beneath the sur-face, but regularly in the form of overt confrontation. Contrarily, the perennialcorrelation constructed between Islamic exegeses and ethnicity tends to reifynational categories and thus extend the discursive armory of these “ethnop-olitical entrepreneurs.” By treating ethnic categories as primordially presententities and in generating “evidence” from religious texts, the new breed oftheologians and scholars are thus unwittingly promoting the group-formationendeavors that lie at the heart of the “Kurdish question.” This not only rendersthe conflict more grounded, but also makes both the Turkish mythology ofnationhood and Ocalan’s attempts to strengthen his support base more dif-ficult to challenge, thereby reducing the efficacy of a universalist, Islamictransformative agenda.

RECOMMENDED READINGS

Agayef, Ahmet. 1914. “_Islam’da Da’va-yı Milliyet” [The National Cause in Islam]. Turk Yurdu7 (10–11): 2320–2329, 2381–2390.

Babanzade, Ahmet Naim. 1914. “_Islam’da Da’va-yı Kavmiyet” [The Factional Cause in Islam].Sebılurresad Mecmuası; 12 (293): 114–128.

Beser, Faruk. 2012. “Kurt Sorunu Mu, Turk Sorunu Mu, Kavmiyetcilik Sorunu Mu?” [TheKurdish Question, the Turkish Question, the Factionalism Question?]. Yeni Safak (September23).

Bulac, Ali. 2010. Kurtler Nereye? [To Where Kurds Are Heading?]. _Istanbul: Cıra Yayınları.Bulac, Ali. 2014. Ortadogu’dan _Ittihad-I _Islam’a [Beyond Middle East, the Islamic Unity],

Vol. 2. _Istanbul: Inkılap.Diyanet. 2013. “Diyanet _Isleri Baskanı Gormez’den onemli acıklamalar” [Important

Statements by President of Diyanet Gormez] (February 20). Available at <http://www.diyanet.gov.tr/tr/icerik/diyanet-isleri-baskani-gormez%E2%80%99den-onemli-aciklamal-ar/8006>, last accessed April 29, 2015.

_Islamoglu, Mustafa. 1992. “Kurt Sorunu, Sistem ve Muslumanlar” [The Kurdish Question,the System and Muslims], in Mazlumder, Kurt Sorunu Forumu. Conference proceedings.Ankara: Mazlumder. 121–131.

Karaman, Hayrettin. 2009. “Bu Ulke Vatandaslarının Tamamı Turk Mu?” [Are the Nations ofthis Country only Turks?]. Yenisafak (September 4). Available at <http://www.yenisafak.com.tr/yazarlar/hayrettinkaraman/bu-ulke-vatandaslarinin-tamami-turk-mu-18419>, lastaccessed April 29, 2015.

Karaman, Hayrettin. 2012. “Ummeti Birlestirmek Farz, Bolmek Haramdır” [The Unity of theUmma is Obligatory, Division is Prohibited]. Yenisafak (May 17). Available at <http://www.yenisafak.com.tr/yazarlar/hayrettinkaraman/ummeti-birlestirmek-farz-bolmek-haramdir-32425>, last accessed April 29, 2015.

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:25

10

Febr

uary

201

6

ISLAM, NATIONALISM, AND KURDISH ETHNOPOLITICS IN TURKEY 353

Ocalan, Abdullah. 1993. Din Sorununa Devrimci Yaklasım [A Revolutionist Approach to theQuestion of Religion]. _Istanbul: Melsa Yayınları.

Sediyani, _Ibrahim. 2014. “Kurdistan _Islam Devleti _Icin Canlarını Veren Turkler, Cerkesler,Alevıle” [The Turks, Circassians, Alevis Who Sacrificed Their Lives for the Sake of IslamicState of Kurdistan]. Ufkumuz (November 8).

Soylemez, Hasim and Hamza Erdogan. 2008. “Marksist PKK’nın Din Oyunu” [MarxistPKK’s Religious Game]. Aksiyon (April 14). Available at <http://www.aksiyon.com.tr/kapak/marksist-pkknin-din-oyunu 515604>, last accessed April 29, 2015.

Tan, Altan. 2009. Kurt sorunu—Ya tam kardeslik ya hep birlikte kolelik [The KurdishQuestion—Either United in Brotherhood or Slavery]. _Istanbul: Timas Yayınları.

Taraf. 2014. “Diyanet’in dini Diyanet’e, Kurtler’in dini Kurtler’e” [Diyanet Be In Its Re-ligion, and so the Kurds in Theirs] (February 3). Available at <http://arsiv.taraf.com.tr/haber-diyanet-in-dini-diyanet-e-kurtler-in-dini-147000/>, last accessed January 1, 2015.

Tim Jacoby is co-founder of the Humanitarian and Conflict Response Institute and Professor at theInstitute for Development Policy & Management at the University of Manchester. He is author of SocialPower and the Turkish State (Routledge, 2004), Disaster Management and Civil Society: EarthquakeRelief in Japan, Turkey and India (I.B. Tauris, 2005, with Alpaslan Ozerdem), Understanding Conflict &Violence: Interdisciplinary and Theoretical Approaches (Routledge, 2007), and Peace in Turkey 2023: TheQuestion of Human Security and Conflict Transformation (Lexington, 2012, also with Alpaslan Ozerdem).E-mail: tim.jacoby@manchester.ac.uk.

Husrev Tabak is Lecturer in International Relations at Rize RTE University in Turkey. He holds an M.A.in Sociology and Politics from University College London and a Ph.D. in Politics from the University ofManchester. His research interests include norms, ethnicity, ethno-politics, and Southeast European poli-tics. He has been working as an editor for the international journals including Journal of Global Analysis,Political Reflection, and Caucasus International. He has translated and co-edited (with Ozgur Tufekci)the following books into Turkish; Post-War Recovery—Disarmament, Demobilization and Reintegration(Ozerdem 2009); Theory and Practice in Peacebuilding (Ozerdem); New Perspectives on Human Secu-rity (McIntosh and Hunter 2010); Fifty Key Thinkers in International Relations (Griffiths et al. 2008);International Relations: The Key Concepts (Griffiths et al. 2007). E-mail: husrevtabak@cesran.org

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:25

10

Febr

uary

201

6

Full Terms & Conditions of access and use can be found athttp://www.tandfonline.com/action/journalInformation?journalCode=uter20

Download by: [Koc University] Date: 10 February 2016, At: 10:27

Studies in Conflict & Terrorism

ISSN: 1057-610X (Print) 1521-0731 (Online) Journal homepage: http://www.tandfonline.com/loi/uter20

The Polarization Trap

Arzu Kibris

To cite this article: Arzu Kibris (2014) The Polarization Trap, Studies in Conflict & Terrorism,37:6, 492-522, DOI: 10.1080/1057610X.2014.903452

To link to this article: http://dx.doi.org/10.1080/1057610X.2014.903452

Accepted author version posted online: 20Mar 2014.

Submit your article to this journal

Article views: 334

View related articles

View Crossmark data

Citing articles: 1 View citing articles

Studies in Conflict & Terrorism, 37:492–522, 2014Copyright © Taylor & Francis Group, LLCISSN: 1057-610X print / 1521-0731 onlineDOI: 10.1080/1057610X.2014.903452

The Polarization Trap

ARZU KIBRIS

Faculty of Arts and Social SciencesSabanci UniversityIstanbul, Turkey

This article studies the association between ethnic conflicts and political attitudes. Itemploys a new data set on the casualties of the ethnic conflict in Turkey to identify theeffects of the Kurdish insurgency on the electoral behavior of the Turkish voters. Theresults indicate that the conflict leads to the political polarization of the society alongethnic nationalist lines. Further investigation of the data also provides some empiricalsupport for the constructivist argument that the salience of ethnic identities is subjectto change in response to external stimuli, of which ethnic conflict constitutes a drasticexample.

Civil conflicts create a “conflict trap.” They perpetuate themselves by destroying economicand social development, and by crippling political institutions, rendering them unableto address the underlying grievances.1 In this article, we empirically demonstrate thatethnic conflicts lead to another very important political change, which we argue should beconsidered among the self-perpetuating dynamics of these conflicts. We study the Turkishcase to show how an ethnic conflict creates a polarization trap by feeding the politicalpolarization of the society along ethnic-nationalist lines.

Turkey offers a very important opportunity to study how ethnic conflicts affect politicalbehavior in host societies. Since August 1984, the country has been suffering from anethnic insurgency campaign led by the Kurdish separatist guerilla organization KurdistanWorkers’ Party (Partiya Karkaren Kurdistan; PKK). Despite the persistent political violencethe electoral process has continued to unfold in a context of democratic elections, and in apolitical arena that hosts extreme ethnic-nationalist parties. The existence of Kurdish andTurkish ethnic-nationalist parties in the political arena enables us to specifically test theeffects of the ethnic conflict on the political salience of ethnic issues and on the politicalsalience of ethnic identities themselves because these parties are by definition champions ofthe particular interests of their ethnic brethren.2 In this study, relying on a unique casualtydatabase that we have assembled, we analyze the association between the conflict and thevote shares of these ethnic-nationalist parties in the 1995 and 1999 general elections. Theresults reveal a significant positive association between the conflict and both the Turkish-nationalist and Kurdish-nationalist vote shares, and thus, point to an ethnic-nationalistpolarization of the electorate. Given the completely opposing views these ethnic partieshold in terms of the conflict and the ways to solve it, the polarization result is telling in

Received 22 October 2013; accepted 6 February 2014.Address correspondence to Arzu Kibris, Sabanci University, Faculty of Arts and Social Sciences,

Orhanli, Tuzla, Istanbul, 34956, Turkey. E-mail: akibris@sabanciuniv.edu

492

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

Polarization Trap 493

terms of how an ethnic conflict perpetuates itself and remains unresolved for almost threedecades.

We then argue that these results also tie to the constructivist argument that ethnicgroups are fluid and endogenous to a set of social, economic, and political processes, andthat individuals have multiple identities whose salience changes over time in response totheir environment.3 We refer to the ecological inference method developed by King (1997)to infer ethnic Kurdish and ethnic Turkish vote choices from aggregate election results.4

Then we analyze how the ethnic conflict is associated with these estimated ethnic votes. Theresults reveal the significant positive association the conflict has with both ethnic Kurdishvotes for the Kurdish-nationalist party, and ethnic Turkish votes for the Turkish-nationalistparty, and thus, indicate that in response to ethnic political violence voters attach morepolitical salience to their ethnic identities in their vote choices.

Our study is closely linked to some recent works which have provided some clues aboutthe type of the association between ethnic conflicts and electoral choice. In their studies onthe Palestinian conflict both Gould and Klor, and Berrebi and Klor find a positive associationbetween the attacks and the vote share of the intransigent right-wing parties in Israel.5 Kibrisreaches a similar conclusion for the Turkish case.6 These studies are of utmost importancebecause they tell us about how people react to violent conflict, and understanding people’sreactions is crucial if we are to figure out the dynamics of these conflicts. Nevertheless, weargue that they provide us with only a partial understanding because they implicitly assumea structure in which a homogenous society is exposed to violence from a single outsidesource. Ethnic conflicts bid ethnic groups in a society against each other both politicallyand economically. In most cases civilians are exposed to violence from both the rebels andthe state. Consequently, it is difficult to talk about a single unifying threat perception inethnic conflicts. Moreover, ethnic conflicts activate ethnic identities in a society rather thana common national one, and raise the salience of ethnic issues. Therefore, we expect theassociation between ethnic conflicts and voting behavior to vary among different groups,and in that sense deepen and maybe even increase the number of cleavages in a society.Our results support our expectation, and demonstrate that the association manifests itselfas an array of different effects which can not be subsumed under a turn toward the rightwing. It seems, rather than unifying the the society behind a certain ideology as suggestedby the aforementioned previous studies,7 the conflict further fragments and polarizes thesociety behind opposing camps.8 This more comprehensive approach to the associationbetween ethnic conlicts and vote choice is one of the main contributions of our study to theliterature.

Also important is the construction of the data set on security force causalties (SFCs).This is the first empirical study to look into the association between the ongoing ethnicconflict and the electoral fortunes of ethnic parties in Turkey. The dearth of studies acknowl-edging this association was not due to the lack of importance or relevance of the issue butwas rather due to lack of data on which to build credible measures of the conflict. Ourdata set provides the first, and so far the only detailed information on casualties. Relyingon this data set, it is possible to measure the variation in conflict intensity across time andlocations.

Finally, this study contributes to the literature by offering a preliminary empirical testof the constructivist arguments about the positive association between ethnic conflicts andthe political salience of ethnic identity. This is an association that carries great importancefor the study of ethnic conflicts. Nevertheless, it evades empirical study in most cases due todifficulties associated with simultaneous availability of data on ethnic identity and politicalbehavior.

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

494 A. Kibris

In the following section, we present a brief overview of the ethnic conflict in Turkey.Then we discuss the theoretical foundations of our study. In the fourth section, we presentour model and introduce our variables. The fifth section discusses the data. We presentour results in the sixth, conduct robustness tests in the seventh section. Finally, weconclude.

The Conflict in Turkey

Ethnic Kurds are the largest ethnic minority in Turkey. A recent survey study puts theirpercentage in the population at around 16 percent.9 Historically, the Turkish state’s ethnicpolicy has consisted of the assimilation of the Kurdish speaking population under the rubricof the Turkish nation and consequently, this large minority has been denied the rights toself expression in the public sphere.10 Saatci suggests that this dominant and homogenizingTurkish nationalism has in return produced the PKK.11

The PKK is a Kurdish separatist guerilla organization that has been conducting armedattacks in Turkey since August 1984. The organization was first founded with the goal ofestablishing an independent Kurdish state in southeastern Turkey, although by late 1999, itappeared to roll back on its goal to a federational structure that would grant more autonomyto the Kurdish population in Turkey.

The armed conflict between the PKK and the Turkish security forces has been geo-graphically concentrated in southeastern and eastern Turkey which is an underdevelopedpart of the country, and which has traditionally been inhabited by ethnic Kurds. Whileboth sides mainly targeted each other’s forces and facilities, they have also employed re-pressive tactics against the civilian populations. The PKK conducted attacks against publicservants like teachers, clergymen, administrative officers, and against civilians whom theyaccused of being “collaborators with the Turkish Republic.” Also, as a result of the “villageguards” system under which the state employed civilian villagers as armed guards againstthe insurgents, there were numerous incidents where the PKK attacked these guards andtheir villages. On the other hand, the Turkish state’s responses included the legal persecu-tion and intolerance of Kurdish political demands, and a more sinister counterinsurgencycampaign that involved extrajudicial killings in the 1990s.12 Interrupted by short-livedcease-fires the armed conflict between the Turkish security forces and the PKK has beengoing on for almost thirty years now. It has cost the country billions of dollars, and morethan 40,000 lives.13 Our knowledge about civilian and insurgent casualties is limited toaggregate numbers sporadically released by contending sources as there is no credible andpublicly available dataset on them. Nevertheless, this study, provides us with a uniquedatabase on security force (i.e. soldiers and police officers) casualties (SFCs).

As can be clearly seen in Figure 1, which depicts the total number of SFCs over theyears, the 1990s has been the most bloody period of the conflict. The PKK received a majorblow when its leader Abdullah Ocalan was captured in Africa in 1999, brought back toTurkey, tried, and sentenced to life in prison. Headless and divided, the PKK ceased itsattacks in the early 2000s. Unfortunately, peace in the area did not last long. The PKKresumed its armed attacks in the second half of 2004.

Since the early 1990s, the conflict has been considered by the political elite and thepublic as one of the most important problems facing the country,14 and the ethnic-basednationalist cleavage contrasting the Turkish and Kurdish identities has become one of themost dominant cleavages of the ideological competition in the Turkish party system15 withtwo ethnic-nationalist parties, one on the Turkish and one on the Kurdish side, as theflagbearers of this competition.

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

Polarization Trap 495

0

200

400

600

800

1000

1200

19841986198819901992199419961998200020022004200620082010

Figure 1. Total number of security force casualties. (Color figure available online).

The MHP (Nationalist Movement Party) is an extreme Turkish-nationalist party withthe most uncompromising, militarist position toward the Kurdish issue. The party has neveracknowledged a separate Kurdish ethnic identity,16 and consequently defined the problemas one of terrorism supported by foreign governments. Party officials have always supportedmilitary solutions to the conflict, accused those who advocated granting political rights andfreedoms to the Kurdish minority with attempting to disintegrate the country, and threatenedto spill blood to prevent any kind of political concessions.17

HADEP (People’s Democracy Party) was a Kurdish-nationalist political party foundedto take up the Kurdish cause and work for a political solution, after its predecessors HEP andDEP were closed down by the Constitutional Court. HEP (People’s Labor Party) was thefirst pro-Kurdish party in Turkey. It was formed in 1990, but due to some legal constraints,could not enter the 1991 general election. Instead 22 HEP members joined the left-wingSocialist People’s Party (SHP) and were voted into the parliament within the party lists ofSHP.18 During its brief lifespan, HEP criticized the policy of the Turkish state against theKurdish minority, stating that the Kurdish issue could not be reduced to a terrorism problem,and insisted that a political solution was urgently needed. It condemned the military optionand stressed that the PKK was not a terrorist movement, but rather, a political organization.It argued that the government should immediately begin a process of dialogue that wouldalso include the PKK as one of the sides, and HEP as the mediator. They demandedcultural rights and political freedoms for the Kurdish minority. HEP was closed downby the Constitutional Court in 1993. Party members reorganized under the roof of DEP(Democracy Party) whose life span turned out to be even shorter than its predecessor. TheConstitutional Court closed the party down in 1994. Expecting this closure, party memberswere already organized under the banner of HADEP, which then lived long enough tocompete in the 1995 and 1999 general elections.19

Ethnic Conflict and Vote Choice

The theoretical insights for the association between ethnic conflict and vote choice stemsfrom several literatures.20 At the very basic level we expect political behavior to be aproduct of personal characteristics and environmental factors including socioeconomicconditions and social networks.21 A large literature of empirical works which have started

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

496 A. Kibris

to accumulate as early as the 1950s supports this argument and demonstrates the impact ofenvironmantal factors on turnout, vote choice, partisanship, and political preferences.22

Civil conflicts create profound social changes. They reverse socioeconomic develop-ment; trigger emigration; redistribute wealth and power in the society and consequently,change the relative status of groups; and create and intensify hatreds.23 Wood categorizesthese changes under six social processes: political mobilization, military socialization, thepolarization of social identities, the militarization of local authority, the transformation ofgender roles, and the fragmentation of the local political economy.24 And she specificallyargues that wartime polarization may lead to electoral polarization, segregation, and adistrustful political culture.

The theory of issue voting labels social factors as issues and presents them as importantdeterminants of vote choice. According to this theory people vote for those candidates whosepositions on and salience attribution to issues they find similar to their own. Accordingly,parties benefit from the salience of issues to which they are generally viewed as attachinghighest priority.25 The salience of an issue varies over time and over individuals. The moresalient the issue, the greater the expected effect of issue ownership and issue proximity onan individual’s voting decision.26

Civil conflicts raise the political salience of security issues, and also the salience of thoseissues surrounding the conflict.27 Violence creates security threats. A large body of literaturein political psychology shows that threat perception increases prejudice, aggression towardsothers, and voting tendencies toward extreme right-wing parties who support aggressivepolicies to deal with these threats.28 In the Israeli case empirical findings show that feelingsof insecurity increase with every terrorist attack mounted against civilians,29 and thatsecurity threats are important predictors of the vote for radical right-wing parties.30

We can also gather theoretical insights from the constructivist arguments and therelated literature about how identity varies as a function of environmental incentives andstrategic manipulation.31 Chandra and Wilkinson differentiate between nominal identitiesthat are “those identity categories in which an individual is eligible for membership basedon the attributes that she or he possesses”; and activated identities which are those that theindividual choses among her nominal identities to guide her behavior.32 An important pointhere is that identity activation is endogenous: It very much depends on the salience of group-specific attributes, which in turn are determined by the social context.33 Ethnic conflictsprovide drastic examples of how the social context activates certain identities. Wilkinsonand Haid argue that instead of being the product of already solid ethnic identities, ethnicconflicts are best seen as an exercise in increasing the salience of some dimensions andagendas over others.34 Wilkinson even argues that ethnic conflicts are provoked by politicalentrepreneurs to mobilize constituents around one ethnic identity rather than another.35

Indeed, Appadurai analyzes the genocide in Rwanda as a form of community-building,a strategy designed to strengthen the identity of the perpetrators.36 Manning and Royargue that national identification is sensitive to patterns of conflict by providing empiricalevidence that Catholics from Northern Ireland rarely think of themselves as British while amajority of Protestants do.37 Shayo and Zussman show how the Palestinian conflict affectsjudges in Israel and demonstrate that attacks lead Arab judges to favor Arab plaintiffs andJewish judges to favor Jewish plaintiffs.38 Shayo and Sambanis argue that ethnic conflictsnot only make ethnic attributes more salient to individuals but they also tend to makeethnic differences more prominent, and thereby reduce perceived similarity to the nationas a whole.39 These conflicts also destroy and reallocate resources between groups, whichcan directly affect both national and group status. In the Turkish case, for example, Saatcidiscusses how the conflict impoverished thousands of Kurds by dislocating them from

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

Polarization Trap 497

their villages.40 Finally, Byman points to the role of state repression in ethnic conflicts interms of identity creation, and argues how the Sri Lankan government’s violent responseto Tamil Tigers has created the perception that the Sri Lankan state and army act only inthe interest of the Sinhalese.41 Similarly, Saatci depicts the Turkish–Kurdish conflict asa “double-edged sword” that has contributed to the growth of ethnic consciuosness andnationalism on both the Turkish and the Kurdish side.42

To summarize, theory offers us three main arguments based on which we can expect tosee an association between ethnic conflicts and vote choice: civil conflicts have profoundeffects on the socioeconomic and political environment that in turn impacts vote choice;people tend to vote for those candidates whose position on and salience attribution to issuesare similar to their own, and civil conflicts affect the relative salience of security issues andthe issues surrounding the conflict; ethnic conflicts activate ethnic identities and activatedidentities are those that shape our political behavior.

The Model

The data set comprises two observations for each county,43 one for the 1995 general electionand one for the 1999 general election. Note that the data exhibit a multilevel character: theobservations are chronologically nested within counties, which in turn are geographicallynested within provinces. The presence of layers may violate the standard Ordinary LeastSquares (OLS) assumption of independent error terms if unobserved factors at the countyand province levels lead the error terms to be correlated. Ignoring the multilevel characterof data risks erroneously low coefficient standard errors, and consequently, increases therisk of declaring significant effects for predictors which in fact have none.44 To accountfor that risk we use the following multilevel linear model in which unobserved county andprovince level factors are incorporated as random effects:

V{t,i,j} = α + βC{t,i,j} + θN{t,i,j} +ϕK{t,i,j} + γ X{t,i,j} + ωE{t}

+ �M{t,i,j} + ρ{j} + u{i,j} + ε{t,i,j}

where ρ{j} is the province level error component, u{i,j} is the county level error compo-nent, and ε{t,i,j} is the individual observation level error component, with E(ε{t,i,j}) = 0,Var (ε{t,i,j}) = σ 2; E(u{i,j}) = 0 Var(u{i,j}) = τ 2; E(ρ{j}) = 0, Var(ρ{j}) = ς2; and Cov[ε{t,i,j},u{i,j}] = 0, Cov [ε{t,i,j}, ρ{j}] = 0, and Cov [u{i,j}, ρ{j}] = 0.

V{t,i,j} is the vote share of the party on whose votes we explore the effects of theethnic conflict. We first explore the effects of the conflict on the vote share of the Kurdish-nationalist party HADEP, and then we explore the effects on the vote share of the Turkish-nationalist MHP. t denotes the time of election, i denotes the county and j denotes theprovince. C{t,i,j} is the number of SFCs who died in the fight against the PKK since theprevious election; N{t,i,j} is the turnout rate; K{t,i,j} is the percentage of ethnically Kurdishpopulation; X{t,i,j} is a vector of socioeconomic, and demographic control variables; E{t} isan election dummy that takes on the value 1 for the 1995 general election and zero for the1999 general election; and M{t,i,j} is a dummy variable that takes on the value 1 if county iof province j was under state of emergency at the time of election t.45

We employ the number of SFCs who died in the fight against the PKK in each countysince the previous general election as a measure of conflict intensity.46 Admittedly, ourmeasure falls short of the total number of casualties which is the most commonly usedmeasure of conflict intensity in the literature. But, as we mentioned above, there is nodetailed, complete and credible panel data available on civilian and insurgent casualties.

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

498 A. Kibris

Hence our data on the SFCs constitute the only available measure that can be used to assessthe variation in the intensity of the conflict over time and space. Moreover, we argue that itis a good measure first because we expect it to be highly correlated with the total numberof casualties and the total number of attacks, and second because it is a good measureof PKK presence in an area. While lack of data on civilian and insurgent casualties,and PKK attacks prevents us from studying how correlated our measure would be withsuch series across counties or provinces over time, the yearly aggregate numbers reportedby the Turkish General Staff47 and the Federation of American Scientists (www.fas.org)enable us to present some evidence of the correlation at the country level. The yearlycountry-wide aggregates we have for the number of SFCs is 84 percent correlated with theyearly total casualty, and 98 percent correlated with the yearly total PKK attack numbersthat the Turkish General Staff reported. The correlation between the yearly total casualtynumbers reported by the Federation of American Scientists is even higher at 97.4 percent.These high correlations are not surprising considering the fact that attacks by the PKK,and offensive military operations by the Turkish security forces, claim lives from bothsides. Unfortunately they also claim the lives of civilians. Also, village guards make up asignificant portion of total casualties, and in many cases village guards get killed alongsidesecurity forces in PKK attacks on their villages, or during search or pursuit missions inthe area. Another major cause of civilian casualties is landmines laid down by both sides.Needless to say, landmine casualties among security forces and civilians are expected to behighly correlated.

Moreover, we expect SFCs to be a strong predictor of PKK casualties as offensive mil-itary operations by Turkish security forces usually follows PKK attacks, especially so whenthose attacks result in heavy casualties on the security forces’ side. Once again, lack of dataprevents us from supporting our claim by empirical evidence from the Turkish case. Never-theless, Jaeger and Paserman’s48 study reveals a similar dynamic in the Israeli–Palestinianconflict. They find that the incidence and levels of Palestinian fatalities can be predicted bythe incidence and levels of Israeli fatalities in the past two months.

Most importanly, as the correlation between the number of SFCs and PKK attacksclearly demonstrate, SCFs is a good measure of the PKK presence in the area. The presenceof PKK insurgents and activity in an area causes a great deal of inconvenience for thecivilian residents. Not only it means that they can get caught in crossfire, or become alandmine victim, it also means that their daily lives are disturbed by the heightened securitymeasures like the increased number of security personnel in the area, and the frequentsecurity checks and controls that are imposed on the civilians, and also by the frequentinterruption of economic activity as a result of attacks and armed skirmishes betweensecurity forces and the PKK. In many cases it also means that they will be pressured,threatened, or even killed by combatants who may accuse them of collaborating with theenemy. Thus, we argue that, SFCs provide a good measure of the intensity of the conflictand the level of political violence civilians are exposed to.

The vector of socioeconomic and demographic variables X{t,i,j} includes populationsize, population growth rate, level of urbanization, unemployment rate, average householdsize, and literacy rate.

Population measures are frequently used variables in modeling electoral behavior.49

Population size measures the total population of the county in thousands; population growthrate is the average yearly growth rate of the county population since the last census; andlevel of urbanization is the percentage of county population living in the county center.Population size and the level of urbanization in a locality may influence vote choice throughtheir effects on the availability of and level of exposure to alternative sources of information.

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

Polarization Trap 499

Population growth rate reflects the migration patterns in a locality that may have an influenceon the vote choice of inhabitants.

One of the stylized facts of the economic voting literature is that voting reacts to a fewmacroeconomic variables, mainly unemployment and inflation.50 Unfortunately, there areno county- (nor province-) level inflation data available for Turkey. Accordingly, the rateof unemployment, which is the number of unemployed per hundred persons in labor forceserves as our macroeconomic control variable.

The average household size is a proxy to capture the center–periphery distinction inthe Turkish society. In his seminal work, which has heavily influenced the literature onelectoral behavior in Turkey, Mardin argues that the center–periphery relations providea good explanatory scheme to understand Turkish politics.51 According to this scheme,the Turkish political system is composed of a coherent body of nationalist, laicist, etatist,educated, urban, “elite,” which constitutes the “center” and a “periphery” constituted by themore traditional, more conservative, more religious, more rural, “antietatist” populations.The center–periphery divide in the Turkish politics nicely coincides with the left-rightdivide. Recent survey studies confirm that the center–periphery framework still retains itsvalidity for Turkish politics.52 Carkoglu and Hinich based on the results of a 2001 survey,conclude that two dimensions dominate the ideological competition in the Turkish partysystem: the secularist versus pro-Islamist cleavage, which again largely overlaps with thecenter versus periphery formations in Turkish politics; and the ethnic-based nationalistcleavage contrasting the Turkish and Kurdish identities.53

Household size is a good measure of conservatism as it is positively correlated withthe level of religious conservatism, and agricultural employment, and negatively correlatedwith the education level of women. Thus, we argue that it can be used to capture thecenter–periphery dichotomy.

Finally, literacy rate is included as a measure of political awareness, and social devel-opment in a county.

Data

Data on Security Force Casualties

The casualty data set includes the date, and place of death at the county level for a totalof 6,802 SFCs who died in the fight against the PKK between August 1984 and January2012. The period in between the 1991 and 1999 general elections, which is the period wewill be focusing on in our analyses, harbors 4,794 of them. This is a unique data set thatwe constructed by bringing together the information we obtained from numerous sourcesincluding the Turkish General Staff, Ministry of Defense, Administration of Police Forces,local administrations, civil society organizations, and newspaper archieves. We explain thedata collection process and the sources we referred to in detail in the Appendix.

Our data set constitutes the only available detailed information on the casualties of theconflict in Turkey. Nevertheless, it has its limitations. First of all, we only have informationon the security force casualties. Admittely, a better measure would include civilian andinsurgent casualties as well, but unfortunately there is no credible and publicly availableinformation on them. On the other hand, we are confident in the reliability of our data onSFCs. As we explain in more detail in the Appendix, the information for the casualtiesof the 1984–September 1998 period comes directly from the Turkish Ministry of Defensepersonnel archives. This is actually the most reliable source for this type of information,and consequently we do not expect to have any unaccounted or misreported casualties for

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

500 A. Kibris

Figure 2a. The distribution of SFCs, 1991–1995. (Color figure available online).

this period. Furthermore, to guarantee the reliability of our information on the SFCs of theOctober 1998–April 1999 period we referred to several sources simultaneously, in manycases including the families of the deceased, and cross checked the information on eachcasualty.

The second limitation of our data set is that we only have information on place of deathat the county level. While some of these casualties took place in or close to villages, towns,or cities, some of the soldiers were killed in clashes in remote mountainous areas quite farfrom populated centers. Because we do not have the exact location of death, we are not ableto control for the distance of the violent act to the nearest settlement, and consequently, wetreat all casualties the same in terms of their impact on voters.

Figure 2(a) and 2(b) provide visual representations of the distribution of SFCs at thecounty level in the 1991–1995 and the 1995–1999 periods, respectively. They reveal theconcentration of PKK attacks in southeastern Turkey.54

Figure 2b. The distribution of SFCs, 1995–1999. (Color figure available online).

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

Polarization Trap 501

Data on Elections

We employ the county level results of the 24 December 1995 and 18 April 1999 parlia-mentary elections. The data are obtained from the Turkish Institute of Statistics. The usageof county-level data on actual vote choice is an important aspect of this study. Note thatmost studies in the literature employ individual-level survey data to analyze the associ-ation between conflict and political attitudes and behavior. Individual-level survey dataallows researchers to more effectively control for exposure to violence, and personal char-acteristics of voters including psychological and perceptional factors, level of awareness,ideological preferences, and party identification, which are important determinants of votechoice. Nevertheless, it also raises significant concerns in terms of representativeness ofthe sample, and in terms of non-response and response biases. In many cases the intensityof the conlict, security concerns, or limitations imposed by the authorities may render itimpossible to reach certain locations and their residents. Moreover, declarations may al-ways differ from actual behavior, especially when respondents try to give the “appropriate”answers, or refrain from answering certain questions. Usage of county-level actual voteseliminates such concerns. We should also emphasize the importance of studying the actualpolitical reaction of people to conflict because, in nonauthoritarian societies, this reactiongreatly influences government policies.

Turkey had seven general elections since the start of armed attacks by the PKK in1984. We focus on the 1995 and 1999 general elections for several reasons. First of all, aKurdish-nationalist party competed only in the 1995, 1999, and 2002 general elections.55

Our second reason for focusing on the 1995 and 1999 general elections is the limitedavailability of county-level socieconomic indicators. The only source of country-widesocieconomic data at the county level in Turkey is the population censuses, which wereheld every five years between 1965 and 2000, with the unfortunate exception of 1995.After 2000, the five-yearly censuses were completely abandoned to be replaced by a newcomputerized system of population registration. Unfortunately, during the transition period,which lasted until 2007, no county-level data were released. In short, we only have 1990 and2000 values for our county-level socioeconomic controls, and this obliges us to choose twoelections among the feasible three. The choice about which elections to study is actuallydictated by the cease-fire that lasted from the second half of 1999 until 2004. The fact thatthe PKK ceased its attacks between the 1999 and 2002 elections renders our measure ofethnic conflict inappropriate for the period. Consequently, we use the indicators derivedfrom the 1990 census results as our socioeconomic controls for the 1995 general election,and the ones derived from the 2000 census results as our controls for the 1999 generalelections.56

Table 1 displays the major political parties that entered the 1995 and 1999 elections,the percentage of votes they received, and their general ideological placement.57

Data on Socioeconomic and Demographic Indicators

We derive our socioeconomic variables from two county-level development studies con-ducted by the State Planning Agency based on the 1990 and 2000 censuses.58

The data set also contains an estimate of the ethnically Kurdish population percentagein each county. The derivation and inclusion of a control for the ethnic distribution ofpopulation across counties is a rather important aspect of this study. Although it is quiteobvious that a model trying to capture the electoral response to ethnic political violenceshould control for the ethnic make up of the society, the lack of data dictates their omission

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

502 A. Kibris

Table 1Political parties that entered the 1995 and 1999 general elections, and the percentage votes

they received

1995 1999

ANAP (center-right) 19.6 13.22DSP (center-left) 14.6 22.19DYP (center-right) 19.2 12.01HADEP (Kurdish-nationalist) 4.2 4.75MHP (Turkish-nationalist) 8.2 17.98RP/FP (pro-Islamist) 21.4 15.41CHP (center-left) 10.7 8.7

in many cases leading to spuriousness concerns. The inclusion of this control eases suchconcerns for our analyses.

We derive our Kurdish population estimates by combining the results of the 1965 censuswith those of the 1990 and 2000 censuses. The 1965 population census is the last census inTurkey to include questions about mother tongue. Considering those who reported Kurdishas their mother tongue in 1965 as ethnically Kurdish, and aggregating the results at theprovince level, Mutlu (1993) reports the 1965 Kurdish population percentages of provincesin Turkey.59 Combining these 1965 provincial percentages, and the 1990 and 2000 censusresults which provide information on birthplace distribution of county residents, we derivedaverage expected Kurdish population percentages for counties for 1990 and 2000. To dothis, we used the 1965 provincial Kurdish population percentages as probabilities of beingKurdish for people who were born in those provinces.60 These estimates put the total ofethnic Kurds in Turkey to 14 percent of the population in 1990, and to little more than15 percent in 2000. Note that the estimate for 2000 is very close to the result of the 2006survey study we have mentioned before which puts the total to 15.6 percent.61

Table 2 displays the mean values and standard deviations for the control variablesincluded in the analyses. The table distinguishes between counties that suffered at least oneSFC in the periods we are interested in and those with no casualties.

Results

The Vote Share of Kurdish-Nationalists

We estimate our model using the restricted maximum likelihood (REML)62 estimationtechnique.

The column on the left in Table 3 presents the results of the regression analysestesting the effects of the conflict on the vote share of the Kurdish-nationalist HADEP. Theestimated coefficient of SFCs is positive and highly significant. Each additional SFC, onthe average, is associated with a 0.09 percentage point increase in the vote share of HADEP.The magnitude of this result is better understood by considering that the average number ofSFCs in the 109 counties that were under the state of emergency in the 1991–1995 periodwas 24, and that there were 23 counties with more than 50 casualties, and 6 with more than100. In other words, the ethnic conflict was, on the average, associated with a 2 percentagepoint increase in the vote share of HADEP in the 1995 election in counties under the state

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

Polarization Trap 503

Table 2Summary statistics

1995 1999

Number of counties C:NC:

179678

133737

SFCs since the previous election C:NC:

16.99 (32.15)0 (0)

9.86 (16.63)0 (0)

Percentage of Kurdish population C:NC:

43.78 (27.49)5.03 (10.34)

46.45 (27.00)7.10 (13.09)

Turnout rate C:NC:

80.04 (7.12)86.13 (6.18)

84.04 (4.58)90.03 (3.98)

Population size C:NC:

78.80 (12.11)45.92 (63.56)

86.42 (14.48)56.06 (98.56)

Population growth rate (%) C:NC:

1.16 (3.26)0.94 (2.58)

0.28 (2.83)0.49 (1.99)

Level of urbanization (%) C:NC:

37.14 (20.07)37.21 (17.60)

46.01 (20.67)43.59 (18.26)

Household size C:NC:

6.41 (0.93)5.16 (0.92)

6.68 (1.49)4.95 (1.21)

Unemployment rate (%) C:NC:

5.25 (3.75)3.78 (2.52)

5.32 (5.48)5.97 (3.81)

Literacy rate C:NC:

63.37 (14.47)78.63 (7.56)

74. 11 (10.81)85.13 (6.36)

(C) gives the mean value for counties with at least one SFC in between the general elections. (NC)gives the mean value for counties with no casualty in the same period. Values in parentheses arestandard deviations.

of emergency, and with more than 5 percentage point increase in counties with casualtiesover 50. Figure 3 visualizes the results. It depicts what happens to HADEP votes whenyou move from a county with no SFCs to a county with average number of SFCs, holdingeverything else constant.63 The curve on the left plots the density of expected HADEP voteshare in a county with no SFCs. The curve to the right plots the density in a county with17 SCFs, which is the average number of SFCs in counties with positive casualties. AsFigure 3 demonstrates, HADEP votes go up by about 2 percentage points on the average as

Figure 3. Simulated HADEP vote shares. (Color figure available online).

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

504 A. Kibris

Table 3Effect on the vote share of ethnic-nationalist parties

Number of obs.: 1,727 Dependent variable Dependent variable

Number of groupsProvince: 80County: 870

Vote share of HADEP Vote share of MHP

Constant 7.688 (3.456) 24.39∗∗∗ (4.92)SFCs since the previous election 0.09∗∗∗ (0.012) 0.08∗∗∗ (0.017)Percentage of ethnically Kurdish

population0.205∗∗∗ (0.024) −0.137∗∗ (0.037)

Turnout −0.057∗∗ (0.026) 0.058 (0.036)Population size 0.006∗∗∗ (0.002) −0.005∗∗ (0.002)Population growth rate 0.188∗∗∗ (0.053) 0.147∗ (0.078)Urbanization level −0.011 (0.013) −0.013 (0.016)Household size 0.915∗∗∗ (0.194) −1.444∗∗∗ (0.262)Unemployment rate 0.243∗∗∗ (0.053) 0.111 (0.075)Literacy rate −0.084∗∗∗ (0.028) −0.013 (0.034)Election dummy −1.214∗∗∗ (0.284) −10.24∗∗∗ (0.401)State of emergency dummy −0.357 (0.840) 3.325∗∗ (1.482)Variance ComponentsProvince level (ς2) 3.26∗∗∗ (0.363) 6.21∗∗∗ (0.569)County level (τ 2) 4.99∗∗∗ (0.148) 3.30∗∗∗ (0.211)Observation level (σ 2) 2.77∗∗∗ (0.067) 4.97∗∗∗ (0.124)–2 Log Likelihood: 10,327 11,242

Estimates from the mixed effects REML regressions. ∗∗∗significant at 1% level; ∗∗significant at5% level; ∗significant at 10% level.

the number of SFCs increase from zero to 17. The simulation exercise when controls areset at their mean values for 1999 yields a very similar picture.

Note that some of the socioeconomic controls like household size, literacy rate andunemployment rate turn out to be highly significantly and substantially associated withHADEP votes. One explanation for these results might the cultural differences betweenethnic Kurds and Turks, and the religious conservatism of ethnically Kurdish populations inTurkey. Consequently, ethnic Kurds tend to have higher birth rates, live in bigger households,and have lower literacy rates, especially among women. In fact, as can be seen in TableA6 inthe Appendix, our estimates for the ethnically Kurdish population are 65 percent correlatedwith household size, and 69 percent correlated with literacy rate across counties for theperiod we are analyzing.64 In other words, along with our ethnic population estimates, thesecontrols might be picking up the ethnic dimension of the votes for HADEP. The significant,highly substantial and negative association between houshold size and MHP vote sharesupports this argument as well. The significant association between unemployment rateand HADEP votes might be due to the fact that HADEP, in terms of economic policy,was a leftist party that supported socialist ideals and a welfare state. The low correlationbetween unemployment rate and ethnic Kurdish population estimates across counties doesnot support arguments based on ethnic discrimination in employment. Finally, one canargue that the detrimental impacts of the conflict on economic activity might lead to higherunemployment rates in the conflict zone, which might then impact vote choices. While this

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

Polarization Trap 505

Figure 4. Simulated MHP vote shares. (Color figure available online).

is a valid argument, the near zero correlation between unemployment rate and SFCs acrosscounties, and the fact that we fail to see any significant impact on the vote share of Turkishnationalists, leads us to think that it can only be a partial one.

The Vote Share of Turkish-Nationalists

The column on the right in Table 3 displays the results for the second set of regressionsinvestigating the effects of the conflict on the vote share of Turkish-nationalists. Eachadditional SFC is, on the average, associated with a 0.08 percentage point increase MHP’svote share. In other words, the ethnic conflict is, on the average, associated with a 2percentage point increase in the vote share of MHP in the 1995 election in counties underthe state of emergency, and with more than 4 percentage point increase in counties withcasualties over 50. Figure 4 visualizes the results. It depicts what happens to MHP voteswhen you move from a county with no SFCs to a county with average number of SFCs,holding everything else constant. The bell-shaped curve on the left plots the density ofexpected MHP vote share in a county with no SFCs, while the curve to the right plots thedensity of expected MHP vote share in a county with 17 SCFs. As can be seen, MHP votesare on the average 2 percentage points higher in counties with average number of SFCs.The simulation exercise when controls are set at their mean values for 1999 yields a verysimilar picture.

The results on the vote shares of the ethnic nationalist parties reveal that the ongoingethnic conflict in Turkey is associated with a rise in both Kurdish and Turkish nationalismin the country.65

To provide a more complete picture of the effects of the conflict on the electoral land-scape we also examine the individual vote shares of the other parties that competed inthe 1995 and 1999 elections. Table A1 in the Appendix reports the results of the regres-sions when the dependent variable is the individual vote share of ANAP, RP, CHP, DYP,and DSP, respectively. The first three of these parties had a relatively moderate stance onthe Kurdish question during the time period we are analyzing.66 The results clearly demon-strate the negative association between the ethnic conflict and the vote shares of thesemoderate parties. The remaining two, DYP and DSP, ideologically stood somewhere in be-tween the moderates and the extreme MHP on the Kurdish-Turkish nationalist cleavage.67

Note that we fail to find any significant association between the vote shares of these twoparties and the ethnic conflict in the period under study.

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

506 A. Kibris

The results clearly demonstrate how the ethnic conflict in Turkey increases the politicalsalience of ethnic nationalism, and polarizes the electorate along an ethnic nationalistcleavage. In that sense they support the theoretical insights about the positive associationbetween ethnic conflicts and the salience of ethnic issues. They are also in line withthe issue voting and issue ownership arguments. The conflict is pushing the electoratetowards extreme ethnic nationalists who are perceived to be the champions of ethnic issueswhile hurting the political success of those with relatively moderate stances. Moreover,they support Wood’s (2008) argument about the electorally polarizing effects of civilconflicts.68 Polarization has vast ramifications for a country’s political system. As argued inSartori’s seminal work, polarization causes centrifugal pressures that shift away the supportfor centrist parties and inhibit the formation of stable parliamentary majorities.69 Thisdirectly leads to the fragmentation and destabilization of democratic regimes. In addition,polarization is likely to cause both social conflict and marked fluctuations of public policies,thus undermining the country’s political and economic performance. In the case of ethnicconflicts, polarization along an ethnic nationalist cleavage does not really bode well for apeaceful and timely resolution as the poles tend to have irreconcilable views on the issuessurronding the conflict and the ways to deal with them. Polarization then implies that theseirreconcilable views are taking hold in the society more and more as a result of the conflictitself. Collier et al. discuss how civil conflicts through their adverse effects on the economycreate a trap that makes peaceful resolution even more difficult.70 The results we havehere point to another trap ethnic conflicts create, namely, the polarization trap. Given theadverse effects of polarization, these results can help us understand why civil conflicts lastlong; why they do not end in peaceful, negotiated settlements more often; and why theyleave societies marked for future conflict.

Ethnic Conflicts and Ethnic Identity

The task that remains now is to understand the characteristics of the observed electoralpolarization. If the constructivist insight about the endogeneity of ethnic identity to ethnicconflicts is valid then we can expect this polarization to actually be a reflection of an ethnicpolarization that is associated with the conflict. Unfortunately, because we do not have in-dividual level data we can not say anything about the ethnic make up of votes. Nevertheless,the ethnic nationalist appeal of the political parties under investigation here, the ethnic na-ture of the conflict, and the significant positive (negative) association between the Kurdishpopulation percentages and the vote share of Kurdish nationalists (Turkish-nationalists)leads us to hypothesize that what we are seeing is in fact a positive association betweenthe conflict and ethnic Kurdish votes for the Kurdish-nationalists, and similarly a positiveassociation between the conflict and ethnic Turkish votes for the Turkish-nationalists. Andif that is the case then it means that ethnic conflicts not only raise the political salience ofethnic issues in a society in general, but also raise the political salience of ethnic identityitself among ethnic groups in the society, and that the constructivists are making a valid ar-gument. So far, due to lack of individual-level data, and the difficulty of measuring conceptslike identity and salience of identity, literature offers limited empirical evidence to supportthis argument. And in many cases, the evidence does not go beyond anectodal referencesto the “legacy of hatred.”71

Unfortunately, the Turkish case also suffers from the lack of individual level data onethnic identity and vote choice. Nevertheless, King offers us a method of “reconstructingindividual behavior from aggregate data.”72 King’s ecological inference (ei) methodologyallows us to use the ethnic population percentages we derived to estimate the ethnic Kurdish

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

Polarization Trap 507

and Turkish73 votes for HADEP and MHP in the 1995 and 1999 general elections.74 Wethen use these estimates to test whether the conflict increases the political salience of ethnicidentities.

The results of this exercise are presented in Table 4. We reestimate our base model usingthe estimated ethnic Kurdish and Turkish votes for HADEP and MHP, respectively, as ourdependent variable. The estimated parameters reveal the significant positive associationbetween the conflict and the votes the two ethnic nationalist parties receive from theircoethnics. In other words, they give support to the conjecture that the conflict increases thepolitical salience of ethnic identity leading ethnic Kurds to vote for Kurdish-nationalists,and ethnic Turks to vote for Turkish-nationalists.75

Note that these results should be viewed as substantially more speculative than the mainresults presented in Table 3, and should not be taken as conclusive evidence since they arenot based on a very rich set of information to begin with. Given the fact that actual ethnicvotes are not observable, and that ecological inferences expose one to the risk of ecologicalfallacies, we present these results cognizant of their limitations. We argue that the resultsshould be read as preliminary evidence and as a suggestion that future research should takeinto account the constructivist arguments about the endogeneity of ethnic identity to ethnicconflicts.

Robustness Tests

We have performed several robustness checks on the main results presented in Table 3.These tests show that the positive association between the ethnic conflict and the voteshares of ethnic nationalist parties are robust to alternative specifications of the model,and the conflict measure used in the analyses, as well as to excluding outlier observationsfrom the sample. The results of the tests are presented in the Appendix. Table A2 presentsthe results when the number of SFCs include those who died in attacks in neighboringcounties as well.76 Table A3 presents the results when the model is specified with fixedprovince effects and county level clustered errors. Table A4 presents the results whenoutlier observations are excluded from the data set.77 As can be seen, the results remainrobust.

Reverse Causality

One concern in estimating the effects of political violence on vote choice is that theremay be a dynamic interaction between these two variables. In other words, while violencemay influence voting behavior, it may also be staged in response to electoral choices. Insuch a case, estimates of the effects of political violence on vote choice will be biased.Note that the situation in Turkey does not warrant such a concern. As we have alreadymentioned, PKK attacks have been concentrated in southeastern Turkey. The area is chosennot because of its residents’ electoral profile but because it was claimed as part of theKurdish homeland by the PKK, and also because the mountainous terrain, which bordersSyria, Iraq, Iran, and Armenia, helped guerilla warfare and provided the insurgents withshelters, and escape routes. Moreover most attacks targeted military facilities rather thancivilian targets. Nevertheless, to scientifically assess that reverse causality is not creatingany specification problems in our analyses, we estimated the coefficients of the followingequation to see if the number of SFCs in between two elections is determined by the results

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

Tabl

e4

Eff

ects

ofth

eet

hnic

confl

icto

net

hnic

vote

sfo

ret

hnic

-nat

iona

lists

Dep

ende

ntva

riab

le

#of

obse

rvat

ions

:1,7

27#

ofgr

oups

:870

Tur

kish

vote

sfo

rH

AD

EP

Kur

dish

vote

sfo

rH

AD

EP

Tur

kish

vote

sfo

rM

HP

Kur

dish

vote

sfo

rM

HP

Con

stan

t5.

366∗∗

∗(1

.46)

48.3

2∗∗∗

(7.3

7)24

.25∗∗

∗(5

.99)

2.42

8∗∗∗

(0.6

80)

SFC

ssi

nce

the

prev

ious

elec

tion

−0.0

08(0

.005

)0.

080∗∗

∗(0

.026

)0.

103∗∗

∗(0

.021

)0.

011∗∗

∗(0

.002

)

Perc

enta

geof

ethn

ical

lyK

urdi

shpo

pula

tion

−0.0

22∗∗

∗(0

.008

)−0

.035

(0.0

57)

0.02

5(0

.042

)0.

003

(0.0

05)

Tur

nout

−0.0

01(0

.011

)−0

.167

∗∗∗

(0.0

55)

0.06

6(0

.045

)0.

008

(0.0

05)

Popu

latio

nsi

ze−0

.001

(0.0

01)

0.01

8∗∗∗

(0.0

04)

−0.0

05∗

(0.0

03)

−0.0

004

(0.0

003)

Popu

latio

ngr

owth

rate

0.14

3(0

.095

)0.

356∗∗

∗(0

.111

)0.

143

(0.0

95)

0.01

5(0

.011

)U

rban

izat

ion

leve

l0.

022

(0.0

23)

−0.0

14(0

.027

)−0

.028

(0.0

19)

−0.0

03(0

.002

)H

ouse

hold

size

0.31

8∗∗∗

(0.0

79)

1.59

9∗∗∗

(0.4

14)

−1.7

37∗∗

∗(0

.321

)−0

.189

∗∗∗

(0.0

36)

Une

mpl

oym

entr

ate

−0.0

09(0

.023

)0.

482∗∗

∗(0

.112

)0.

146

(0.0

93)

0.01

6(0

.011

)L

itera

cyra

te−0

.066

∗∗∗

(0.0

11)

−0.2

74∗∗

∗(0

.060

)0.

006

(0.0

4)0.

0002

(0.0

04)

Ele

ctio

ndu

mm

y−0

.693

∗∗∗

(0.1

19)

−3.2

07∗∗

∗(0

.603

)−1

0.60

∗∗∗

(0.4

87)

−1.1

89∗∗

∗(0

.056

)St

ate

ofem

erge

ncy

dum

my

−0.2

25(0

.353

)−1

.013

(1.7

81)

0.94

0(1

.725

)0.

108

(0.1

97)

Var

ianc

eco

mpo

nent

sar

eno

tre

port

ed.

–2L

ogL

ikel

ihoo

d7,

372

12,9

2711

,929

4,46

3

Est

imat

esfr

omth

em

ixed

effe

cts

RE

ML

regr

essi

ons.

∗∗∗ s

igni

fican

tat1

%le

vel;

∗∗si

gnifi

cant

at5%

leve

l;∗ s

igni

fican

tat1

0%le

vel.

508

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

Polarization Trap 509

of the prior election:

C{t,i,j} = α + βC{t − 1,i,j} + ηV{t − 1,i,j} + θN{t − 1,i,j}

+ ϕK{t − 1,i,j} + γ X{t − 1,i,j} + �M{t − 1,i,j} + ρS{i,j} + ε{t,i}

where t denotes the election year, i denotes the county, and j denotes the province. C{t,i.j} isthe number of SFCs who died in the fight against the PKK insurgency between the electionat (t−1) and the election at t in county i of province j; C{t−1,i,j} is then the number ofSFCs who died the fight against the PKK insurgency between the election at (t−2) andthe election at (t−1); V{t−1,i,j} is the vote share of the party we are interested in in the(t−1) election; N{t−1,i,j} is the turnout rate; K{t−1,i,j} is the percentage of ethnically Kurdishpopulation; X{t−1,i,j} is the vector of socioeconomic variables; M{t−1,i,j} is a dummy thattakes on the value 1 if the county was under state of emergency in (t−1); and finally S{i,j}is a dummy variable that takes on the value 1 if the county has a border with Iraq, Syria,Iran, or Armenia.

We estimate the above equation for the 1995–1999 period (t = 1999) using OLS withclustered errors at the province level.78 Unfortunately we are not able to conduct a similaranalysis for the 1991–1995 period to see whether the SFCs of the period was determinedby the Kurdish and/or Turkish-nationalist votes in the 1991 general election since neitherMHP nor HADEP entered the 1991 general elections. Table A5 in the Appendix displaysthe results. None of the estimated βs are significant indicating that the data do not supportthe reverse causality hypothesis.

Conclusion

This study analyzes the effects of the Turkish–Kurdish ethnic conflict on the electoralchoices of Turkish voters in the 1995 and 1999 general elections. The results demonstratea significant positive association between the conflict and the vote shares of Kurdish andTurkish-nationalist parties, and as such, they indicate the polarization of the electoratealong an ethnic nationalist cleavage. Given the completely opposing views of these partieson how to resolve the conflict, such a polarization does not actually bode well for a peacefuland timely resolution. It seems the conflict is leading people to mass behind two opposinggroups with irreconcilable claims about the reasons of and solutions to the conflict. Whatthe Turkish case is presenting us here is, in a sense, a circular structure with ethnic conflictfeeding political polarization that then renders a peaceful and timely resolution moredifficult; in other words, a polarization trap. Horowitz makes a similar argument in hisseminal work on ethnic conflicts when he talks about how ethnic parties may exacerbateethnic conflict by coming up with “mutually incompatible” claims to power.79 We arguethat, along with the other results in the literature on the detrimental sociopolitical effects ofcivil conflicts, the polarization trap can shed further light on the self-perpetuating dynamicsof these conflicts, and can help us better understand the reasons behind their resilience.

We also present some preliminary evidence that what we are observing is not just apolitical polarization along an ethnic nationalist dimension, but also a rise in the politicalsalience of ethnic identities that then reflects on vote choice. The results of the ecologicalregressions reveal how the conflict increasingly brings ethnic identities to the forefront invote choice, and thus, further deepens the ethnic cleavage in the society.

The observed ethnic nature of most civil conflicts has led scholars to devise and testvarious hypotheses on how conflict and the number of ethnic groups in a society arerelated.80 Interestingly, a good portion of these failed to establish a clear association, which

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

510 A. Kibris

then led scholars on a quest for better ethnic heterogeneity measures.81 Note that whilethese studies significantly advance our understanding of ethnic conflicts, they all, implicitlyor explicitly, assume that the salience of ethnic identity for any two co-ethnics will be thesame for both and will remain fixed. In other words, they ignore the constructivist insightabout the endogeneity of the salience of ethnic identities to ethnic conflict itself. The resultsin this article support this insight. If it is not just ethnic identity itself but rather the salienceof that identity that drives people’s behavior, then this endogeneity might actually explainwhy scholars have so far not been able to establish a clear link between ethnicity and ethnicconflict even though they strongly suspect that it exists. We therefore close by emphasizingthe need to find new and novel ways to incorporate the dynamic link between ethnic identityand ethnic conflict into our studies on such conflicts.

Acknowledgments

This article was written while I was visiting the Political Science Department of DukeUniversity. I thank the members of this institution for their hospitality. I especially thankMichael C. Ward for all his help and mentorship in my research. Finally, I thank the seminarparticipants at University of North Carolina at Chapell Hill and Duke University for theirhelpful comments and suggestions.

Notes

1. Paul Collier, “On the Economic Consequences of Civil War,” Oxford Economic Pa-pers 51(1) (1999), pp. 168–183; Paul Collier et al., “Breaking the Conflict Trap, Civil War andDevelopment Policy”, A World Bank Policy Research Report (A copublication of World Bankand Oxford University Press, Washington DC and New York); Arzu Kibris, “The Conflict TrapRevisited: Civil Conflict and Educational Achievement,” forthcoming in the Journal of ConflictResolution; Hazem Adam Ghobarah et al., “Civil Wars Kill and Maim People Long After theShooting Stops,” American Political Science Review 97(2) (2003), pp. 189–202; Anke Hoefflerand Marta Reynal-Querol, “Measuring the Costs of Conflict,” unpublished working paper. Avail-able at www.conflictrecovery.org/bin/2003 Hoeffler Reynal-Measuring the Costs of Conflict.pdf(accessed 13 April 2014); Elisabeth Jean Wood, “The Social Processes of Civil War: The WartimeTransformation of Social Networks,” Annual Review of Political Science 11 (2008), pp. 539–561.

2. Kanchan Chandra, “What is an Ethnic Party,” Party Politics 17(2) (2011), pp. 151–169;Donald L. Horowitz, Ethnic Groups in Conflict, 2nd edition (Oakland, CA: University of CaliforniaPress, 2000).

3. Kanchan Chandra, “Cumulative Findings in the Study of Ethnic Politics,” APSA Newsletter12(1) (2001), pp. 7–11; Kanchan Chandra, “What is Ethnic Identity And Does It Matter?” AnnualReview of Political Science 9 (2006), pp. 397–424; Steven I. Wilkinson, “Constructivist Assumptionsand Ethnic Violence,” APSA Newsletter 12(1) (2001), pp. 17–20; Daniel N. Posner, “The Implicationsof Constructivism for Studying the Relationship Between Ethnic Diversity and Economic Growth,”in Ethnicity, Politics, and Economics, ed. Kanchan Chandra, forthcoming; Ravi Bhavnani and DanMiodovnik, “Ethnic Polarization, Ethnic Salience, and Civil War,” Journal of Conflict Resolution53(1) (2009), pp. 30–49.

4. Gary King, A Solution to the Ecological Inference Problem (Princeton, NJ: PrincetonUniversity Press, 1997).

5. Eric D. Gould and Esteban F. Klor, “Does Terrorism Work?” The Quarterly Journal ofEconomics (November, 2006), pp. 1459–1510; Claude Berrebi and Esteban F. Klor,, “Are VotersSensitive to Terrorism: Direct Evidence from the Israeli Electorate,” American Political ScienceReview 102(3) (August, 2008), pp. 279–301; Claude Berrebi and Esteban F. Klor, “On Terrorism

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

Polarization Trap 511

and Electoral Outcomes: Theory and Evidence from the Israeli-Palestinian Conflict,” The Journal ofConflict Resolution 50 (December 2006), pp. 899–925.

6. Arzu Kıbrıs, “Funerals and Elections: The Effects of Terrorism on Voting Behavior inTurkey,” Journal of Conflict Resolution 55(2) (2011), pp. 220–247.

7. Berrebi and Klor, “Are Voters Sensitive to Terrorism”; Gould and Klor, “Does TerrorismWork?”; Berrebi and Klor, “On Terrorism and Electoral Outcomes”; Kıbrıs, “Funerals and Elections.”

8. Wood, “The Social Processes of Civil War.”9. KONDA Arastırma ve Danısmanlık, Biz Kimiz, Toplumsal Yapı Arastırması (2006).

10. Gunes Murat Tezcur, “When Democratization Radicalizes: The Kurdish Nationalist Move-ment in Turkey,” Journal of Peace Research 47(6) (2010), pp. 775–789; Nergis Canefe, “TurkishNationalism and Ethno-Symbolic Analysis: The Rules of Exception,” Nations and Nationalism 8(2)(2002), pp. 133–155; Mesut Yegen, “The Kurdish Question in the Turkish State Discourse,” Journalof Contemporary History 34(4) (1999), pp. 555–568.

11. Mustafa Saatci, “Nation-States and Ethnic Boundaries: Modern Turkish Identity andTurkish–Kurdish Conflict,” Nations and Nationalism 8(4) (2002), pp. 549–564.

12. Hamit Bozarslan, “Human Rights and the Kurdish Issue in Turkey: 1984–1999,” HumanRights Review 3(1) (2001), pp. 45–54; TBMM, Faili Mechul Cinayetleri Arastırma KomisyonuRaporu (Report of the Turkish Parliament Commission Investigating the Unidentified Murders)(Ankara, 1995); Gunes Murat Tezcur, “Judicial Activism in Perilous Times: The Turkish Case,” Law& Society Review 43(2) (2009), pp. 305–336; Tezcur, “When Democratization Radicalizes”; JoostJongerden, “Contested Spaces in Landscapes of Violence: Displacement and Return in Diyarbakır atthe Turn of the 21st Century,” Kurdische Studien 4/5 (2004/2005), pp. 61–89.

13. Nedim Sener, “26 Yilin Kanli Bilancosu,” Milliyet, 24 June 2010.14. In various surveys conducted in the last decade, PKK terror was ranked the most important

problem of the country by large percentages of respondents. See Yilmaz Esmer, Devrim, Evrim,Statuko: Turkiye’de Sosyal, Siyasal ve Ekonomik Degerler (Istanbul: TESEV, 1999); Ali Carkogluand Binnaz Toprak, Turkiye’de Din, Toplum ve Siyaset (Istanbul: TESEV, 2000); Ali Carkogluand Binnaz Toprak, Degisen Turkiye’de Din, Toplum ve Siyaset (Istanbul: TESEV, 2006); ErsinKalaycıoglu, “Public Choice and Foreign Affairs: Democracy and International Relations in Turkey,”New Perspectives on Turkey (2009): 59–83; Eurobarometer Autumn 2004 and onward. For example,in a 1993 survey conducted by the Turkish Foundation for Social and Economic Research (TUSES),45 percent of the respondents ranked the conflict as the most important problem of the country,making it the most highest ranked problem. In a 1996 survey conducted by the same institution, 17percent ranked it as the most important problem, making it the second highest ranked problem. InNovember 1991, the then–prime minister, Suleyman Demirel described the “Kurdish situation” asTurkey’s top problem. In 1993, the then-president, Turgut Ozal, described the situation as “perhapsthe most significant problem in the republic’s history.” See Michael M. Gunter, The Kurds and theFuture of Turkey (New York: St. Martin’s Press, 1997).

15. Ali Carkoglu and Melvin Hinich, “A Spatial Analysis of Turkish Party Preferences,” Elec-toral Studies 25 (2006), pp. 369–392.

16. MHP leaders have upheld the argument that Kurds are “overwhelmingly of Turkish descent.”17. Gunter, The Kurds and The Future of Turkey; Kemal Kirisci and Gareth M. Winrow, The

Kurdish Question and Turkey: An Example of Trans-state Ethnic Conflict (London: Frank CassPublishers, 1997); Nimet Beriker, “The Kurdish Conflict in Turkey: Issues, Parties, and Prospects,”Security Dialogue 28(4) (1997), pp. 439–452.

18. In Turkey, people vote for parties and not for individual candidates. Thus, there is actuallyno way to identify the Kurdish-nationalist votes within the total votes SHP got in the 1995 elections.

19. Gunter, The Kurds and The Future of Turkey; Kirisci and Winrow, The Kurdish Questionand Turkey; Gulistan Gurbey, “The Development of the Kurdish Nationalism Movement in Turkeysince the 1980s,” in Robert Olson, ed., The Kurdish Nationalist Movement in the 1990s: Its Impacton Turkey and the Middle East (Lexington, KY: The University Press of Kentucky, 1996); JoostJongerden and Ahmet Hamdi Akkaya, “Born From the Left: The Making of the PKK,” in Marlies

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

512 A. Kibris

Casier and Joost Jongerden, eds., Nationalism and Politics in Turkey: Political Islam, Kemalism, andthe Kurdish Issue (New York: Routledge Studies in Middle Eastern Politics, 2011).

20. “What are the factors that are associated with the vote share of a specific politicalparty?” and “What makes an individual voter more likely to vote for a specific political party?”are two different, nonetheless, very related questions. If there are certain individual characteris-tics that make a person more likely to vote for a specific party, then we would expect that partyto get a high vote share in a society in which a significant number of voters share those char-acteristics. In that sense, understanding how ethnic conflicts are associated with individual votechoice can inform us about the association between ethnic conflicts and vote shares of politicalparties.

21. Admittedly, not all agree on the importance of environmental factors in political behavior.As Zuckerman argues “the founders of the behavioral analysis of political preferences and electoralchoices institutionalized a research agenda that departed from the social logic of politics.” The heavyreliance on survey data that examined individual characteristics without much attention to the socialcontext, coupled with the introduction of rational choice theory into the subject matter, led, for aperiod, to a focus on the individual level of analysis, and on individual attitudes and calculations asthe main determinants of vote choice. Nevertheless, in the last decades the study of the social logic ofpolitical behavior has regained interest. For a comprehensive account of the academic debate on theindividual versus social logic of political behavior and a detailed review of the related literature, pleasesee Alan Zuckerman, “Returning to the Social Logic of Political Behavior,” in Alan S. Zuckerman,ed., The Social Logic of Politics.Personal Networks as Contexts for Political Behavior (Philadelphia:Temple University Press, 2005).

22. Bernard R. Berelson, Paul F. Lazarsfeld, and William N. McPhee, Voting: A Study ofOpinion Formation in a Presidential Campaign (Chicago: University of Chicago Press, 1954); RobertHuckfeldt, Politics in Context: Assimilation and Conflict in Urban Neighborhoods, (New York:Agathon Press, 1986); Paul Allen Beck et al., “The Social Calculus of Voting: Interpersonal, Media,and Organizational Influences on Presidential Choices,” The American Political Science Review96(1) (2002), pp. 57–73; Zuckerman, “Returning to the Social Logic of Political Behavior”; MichaelS. Lewis-Beck and Martin Paldam, “Economics of Voting: An Introduction,” Electoral Studies 19(2000), pp. 113–121.

23. Collier, “On the Economic Consequences of Civil War”; Collier et al., “Breaking the ConflictTrap.”

24. Wood, “The Social Processes of Civil War.”25. D. Roderick Kiewiet, “Policy-Oriented Voting in Response to Economic Issues,” American

Political Science Review 75 (June 2001), pp. 448–459; G. Rabinowitz, J. W. Prothro, and W. Jacoby,“Salience as a Factor in the Impact of Issues on Candidate Evaluation,” Journal of Politics 44 (1982),pp. 41–63.

26. George C. Edwards III, William Mitchell, and Reed Welch, “Explaining Presidential Ap-proval: The Significance of Issue Salience,” American Journal of Political Science 39(1)(1995), pp.108–134; Eric Belanger and Bonnie Meguid, “Issue Salience, Issue Ownership, and Issue-Based VoteChoice,” Electoral Studies 27 (2008), pp. 477–491.

27. Edwards et al., “Explaining Presidential Approval”; Berrebi and Klor, “Are Voters Sensitiveto Terrorism: Direct Evidence from the Israeli Electorate,” American Political Science Review 102(3)(August 2008), pp. 279–301.

28. For a review please see Sivan Hirsch-Hoefler, Daphna Canetti, and Ami Pedahzur, “Twoof a Kind? Voting Motivations for Populist Radical Right and Religious Fundamentalist Parties,”Electoral Studies 29 (2010), pp. 678–690.

29. Carol Gordon and Asher Arian, “Threat and Decision Making,” The Journal of ConflictResolution 45 (2001), pp. 196–215; Daphna Canetti et al., “An Exposure Effect? Evidence from aRigorous Study on the Psycho-political Outcomes of Terrorism,” in Samuel Justin Sinclair and DanielAntonius, eds., The Political Psychology of Terrorism Fears (New York: Oxford University Press,2013).

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

Polarization Trap 513

30. A. Arian, Security Threatened: Surveying Israeli Opinion on Peace and War (New York:Cambridge University Press, 1995); A. Arian, “And Then There Were Two: On Demography andPolitics in the Jewish Future,” Society 36 (1999), pp. 21–26; Berrebi and Klor, “Are Voters Sensitive.”

31. Chandra, “Cumulative Findings in the Study of Ethnic Politics”; Chandra, “What is EthnicIdentity”; Wilkinson, “Constructivist Assumptions and Ethnic Violence.”

32. Kanchan Chandra and Steven Wilkinson, “Measuring the Effect of ‘Ethnicity,”’ Compara-tive Political Studies 41 (2008), pp. 515–563.

33. Chandra and Wilkinson, “Measuring the Effect of Ethnicity”; Nicholas Sambanis and MosesShayo, “Social Identification and Ethnic Conflict,” American Political Science Review 107(2) (2013),pp. 294–325.

34. Steven I. Wilkinson and Christopher J. Haid, “Ethnic Violence as Campaign Expenditure:Riots, Competition, and Vote Swings in India,” working paper, Department of Political Science, YaleUniversity.

35. Wilkinson, “Constructivist Assumptions and Ethnic Violence.”36. Arjun Appadurai, Fear of Small Numbers: An Essay on the Geography of Anger (Durham,

NC: Public Planet Books/Duke University Press, 2006).37. Alan Manning and Sanchari Roy, “Culture Clash or Culture Club? National Identity in

Britain,” The Economic Journal 120 (2010), pp. 72–100.38. Moses Shayo and Asaf Zussman, “Judicial Ingroup Bias in the Shadow of Terrorism,” The

Quarterly Journal of Economics 126 (2011), pp. 1447–1484.39. Sambanis and Shayo, “Social Identification and Ethnic Conflict.”40. Saatci, “Nation-States and Ethnic Boundaries.”41. Daniel Byman, “The Logic of Ethnic Terrorism,” Studies in Conflict and Terrorism 21

(1998), pp. 149–169.42. Saatci, “Nation-States and Ethnic Boundaries.”43. Note that due some administrative changes, some localities gained county status after the

1995 general election. For those counties the data set includes only the 1999 observation.44. Marco R. Steenbergen and Bradford S. Jones, “Modeling Multilevel Data Structures,”

American Journal of Political Science 46(1) (2002), pp. 218–237.45. As a result of the heavy and incessant PKK activities and attacks, a number of provinces

were put under a state of emergency in 1987, and most of them remained so all through the 1990s.State of emergency brought many restrictions on the daily lives of the residents of these places, andthe freedoms they should have normally enjoyed. Moreover, it meant transfer of military personnel tothe area, which meant some change in the voter profile of these locations. Thus, a state of emergencyitself can be an important determinant of electoral behavior.

46. Note that because the conflict is geographically concentrated, SFC distributions over theyears are very highly correlated. Consequently, the usage of number of SFCs over shorter periods oftime before the elections leads only to scale effects rather than any substantial changes in the results.

47. Sener, “26 Yilin Kanli Bilancosu.”48. David A. Jaeger and M. Daniele Paserman, “The Cycle of Violence? An Empirical Anal-

ysis of Fatalities in the Palestinian-Israeli Conflict,” American Economic Review 98(4) (2008), pp.1591–1604.

49. Benny Geys, “Explaning Voter Turnout: A Review of Aggregate Level Research,” ElectoralStudies 25 (2006), pp. 637–663.

50. Lewis-Beck and Paldam, “Economics Voting.”51. Serif Mardin, “Center Periphery Relations: A Key to Turkish Politics?” in Engin Akarlı and

Gabriel Ben-Dor, eds., Political Participation in Turkey (Istanbul: Bogazici University Press, 1975),pp. 8–18.

52. Ersin Kalaycıoglu, “Elections and Party Preferences in Turkey, Changes and Continuitiesin the 1990s,” Comparative Political Studies 27(3) (1994), pp. 402–424; Ersin Kalaycıoglu, “PublicChoice and Foreign Affairs: Democracy and International Relations in Turkey,” New Perspectives onTurkey 40 (2009), pp. 57–81; Yılmaz Esmer, “At the Ballot Box,” in Sabri Sayarı and Yılmaz Esmer,eds., Politics, Parties, and Elections in Turkey (Boulder, CO: Lynne Rienner, 2002); Yılmaz Esmer,

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

514 A. Kibris

“Parties and the Electorate: A Comparative Analysis of Voter Profiles of Turkish Political Parties,” inCigdem Balım et al., ed., Turkey-Political, Social, and Economic Challenges in the 1990s (Leiden,New York: E.J.Brill, 1995).

53. Carkoglu and Hinich, “A Spatial Analysis of Turkish Party Preferences.”54. The red areas have more than 100 SFCs; the pink areas have 50 to 100; the grey areas have

30 to 50; the orange areas have 10 to 30; the green areas have 1 to 10; and the yellow areas have none.55. The 1995 general elections were the first in which a Kurdish nationalist party competed.

The Kurdish-nationalist party did not compete in the 2007 and 2011 general elections either; instead,a number of its members ran as independent candidates from a limited number of electoral regions.Note that the Turkish-nationalist MHP did not enter the 1991 elections either.

56. Because violence may affect demographic variables like population growth rate and urban-ization rate through its effect on migratory flows, the usage of year 2000 values of socioeconomicindicators as controls for the 1999 elections may lead to an identification problem. We argue that inthis case such a concern is not warranted for several reasons: First of all, the census was taken only ayear after the elections, and we would not expect the demographic variables to change dramaticallyin just one year. Relatedly, the population growth rate is an average for the 1990–2000 period andconsequently is not expected to be much different from an average for the 1990–1999 period. Second,the PKK announced a ceasefire shortly after the 1999 elections. Because of the cease-fire we wouldnot expect to see a significant amount of violence-related migratory outflows from conflict areasbetween the 1999 elections and the 2000 census. Moreover, we would not expect those who hadescaped from violence before to return home as one year would not really be enough to convincepeople to do so.

57. The ideological classification of parties follows closely the widely used classification inTurkish electoral studies (Kalaycıoglu, “Elections and Party Preferences in Turkey”; Esmer, “At theBallot Box”; Ali Carkoglu and Ersin Kalaycıoglu, Turkish Democracy Today: Elections, Protest, andStability in an Islamic Society (London: I. B. Tauris, 2007)).

58. Bulent Dincer and Metin Ozaslan, Ilcelerin Sosyo-Ekonomik Gelismislik SıralamasıArastırması, Devlet Planlama Teskilatı (1996; 2004).

59. Servet Mutlu, “Ethnic Kurds in Turkey,” International Journal of Middle East Studies 28(4)(1996), pp. 517–541.

60. For example, if 50 percent of the 1990 residents of county X were born in province A,which, according to Mutlu, had 30 percent ethnic Kurds in 1965, and the other 50 percent was bornin province B, which had 10 percent ethnic Kurds in 1965, then our expectation for county X is tohave a population with 20 percent ethnic Kurds in 1990.

61. KONDA Arastırma ve Danısmanlık, “Biz Kimiz.”62. REML and unrestricted/full maximum likelihood estimation (MLE) produce identical fixed

effects estimates. Their estimates for variance components will also be the same for large samples, butbecause REML corrects for the degrees of freedom consumed by estimation of the fixed effects, forsmall samples, MLE is biased toward lower variance estimates (Steenbergen and Jones, “ModelingMultilevel Data Structures”; Jeremy J. Albright and Dani M.Marinova, Estimating Multilevel ModelsUsing SPSS, Stata, SAS, and R, mimeo (Indiana University, 2010)).

63. To create this figure, we used the estimated coefficients and their covariances as the meansand the covariances of a multivariate normal distribution from which we drew 100,000 simulatedcoefficients. Then we fixed SFCs at zero, the election dummy at 1, the martial law dummy at 0,and the remaining controls at their mean values for 1995, and by multiplying each of the simulatedcoefficient vectors with this fixed control variables vector, we calculated 100,000 expected HADEPvote shares, whose density is depicted by the bell-shaped curve on the left. Then we drew another setof simulated coefficients, fixed SFCs at 17 and holding all other controls constant, calculated anotherset of expected HADEP vote shares, whose density is depicted by the bell-shaped curve on the right.

64. Note that these correlations are not high enough to cause any concern of multicollinearity.In any case, multicollinearity does not bias the estimates, it may only render them statisticallyinsignificant, which is not a problem here.

65. Also note that the election dummy signals a sizable jump in especially MHP’s vote share inthe 1999 election. Actually this jump is also very much related to the conflict. As mentioned before

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

Polarization Trap 515

PKK’s leader was captured by Turkish security forces shortly before the election in 1999. This ledto a rise in ethnic nationalist sentiments in both ethnic groups. The success of the security forces andthe heavy blow the PKK received as a result gave the MHP the opportunity to argue that they wereright all along in their support for military solutions and to surf to the parliament on the raising waveof Turkish ethnic-nationalist sentiments in the 1999 election.

66. Hans-Dieter Klingemann, Andrea Volkens, Judith Bara, Ian Budge, and Michael McDonald,Mapping Policy Preferences II. Estimates for Parties, Electors, and Governments in Eastern Europe,the European Union and the OECD, 1990–2003 (Oxford: Oxford University Press, 2006); Carkogluand Hinich, “A Spatial Analysis of Turkish Party Preferences”; Gunter, The Kurds and The Futureof Turkey; Kirisci and Winrow, The Kurdish Question and Turkey; Beriker, “The Kurdish Conflict inTurkey.”

67. Klingemann et al., Mapping Policy Preferences II; Carkoglu and Hinich, “A Spatial Analysisof Turkish Party Preferences.”

68. Wood, “The Social Processes of Civil War.”69. Giovanni Sartori, Parties and Party Systems: A Framework for Analysis (Cambridge, UK:

Cambridge University Press, 1976).70. Collier et al., “Breaking the Conflict Trap.”71. Chaim Kaufmann, “Possible and Impossible Solutions to Ethnic Civil Wars,” International

Security 20(4) (1996), pp. 136–175; Collier et al., “Breaking the Conflict Trap”; Paul Collier andAnka Hoeffler, “Civil War.” Draft chapter for the Handbook of Defense Economics (2006).

72. Due to space limitations we are not able to go into a methodological description.Those interested are kindly requested to refer to King, A Solution to the Ecological InferenceProblem.

73. Note that ethnic Turks here refer to ethnically non-Kurdish populations of Turkey. While itis true that there exist other small ethnic minorities in Turkey, we base our referral of non-Kurdishpopulations as “ethnically Turkish” on the fact that these other groups are relatively very small.Moreover, according to a recent survey study, most of them consider themselves coming from aTurkic ethnic background (like Balkan Turks, Gagauz Turks, Asian Turks, or Caucasian Turks)(KONDA, Biz Kimiz). Consequently, that study puts the total number of ethnic Turks and Kurds toaround 98.5 percent of the population.

74. King’s methodology requires that we make three assumptions: (a) that ethnic HADEP(MHP) votes are distributed according to a truncated bivariate normal distribution; (b) that theirexpected values are independent of the ethnic composition; (c) that HADEP (MHP) votes acrosscounties are independent after conditioning on ethnic composition. Admittedly,the second assumptionis the most likely one to fail in the Turkish context. King suggests plotting estimated county levelethnic Kurdish (Turkish) votes for HADEP (MHP) against the percentage of ethnic Kurds (Turks)across counties as a diagnostic for the presence of an aggregation bias. The plots indicate some smallbias but not enough to warrant discarding the analysis.

75. Note that the estimated coefficient for the effect of casualties on ethnic Kurdish votes forMHP is positive (albeit very small) and significant. One explanation for this might be the villageguards system. Because there are ethnic Kurds who make money out of this system, MHP’s strongsupport for the continuation of the system might have appealed to them.

76. Neighboring counties to a county are those that border that county.77. We have also conducted jackknife cross validation, which re-estimates the model parameters

by resampling serially leaving one county out each time. The mean and the variance of the distributionof the parameter estimates from these iterations are very close to the original results. The results areavailable on request.

78. Note that because we can only analyze the casualties of the 1995–1999 period, we have onlyone observation at the county level, and consequently the possible correlation of error terms withincounties is not a problem anymore. But we still have the clustering of counties within provinces. Weuse OLS with clustered errors to account for the possibility of intraprovince correlation of errors.Note also that the existence of the lagged dependent variable in the model renders a random or fixedeffects model unsuitable.

79. Horowitz, Ethnic Groups in Conflict.

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

516 A. Kibris

80. Paul Collier and Anke Hoeffler, Greed and Grievance in Civil War, World Bank PolicyResearch Working Paper 2355 (World Bank, 2000); Nicholas Sambanis, “What is Civil War? Con-ceptual and Empirical Complexities of an Operational Definition,” Journal of Conflict Resolution 48(2004), pp. 814–858; Havard Hegre and Nicholas Sambanis, “Sensitivity Analysis of the EmpiricalLiterature on Civil War Onset,” Journal of Conflict Resolution 50 (2006), pp. 508–535; Randall J.Blimes, “The Indirect Effect of Ethnic Heterogeneity On The Likelihood of Civil War Onset,” Jour-nal of Conflict Resolution 50 (2006), pp. 536–547; Ibrahim Elbadawi and Nicholas Sambanis, “HowMuch War Will We See? Explaining The Prevalence of Civil War,” Journal of Conflict Resolution 46(2002), pp. 307–334; James D. Fearon and David D. Laitin, “Ethnicity, Insurgency, and Civil War,”The American Political Science Review 97(1) (2003), pp. 75–90; James Fearon, Kimuli Kasara, andDavid Laitin, “Ethnic Minority Rule and Civil War Onset,” American Political Science Review 101(1)(2007), pp. 187–193.

81. James Fearon, “Ethnic and Cultural Diversity by Country,” Journal of Economic Growth8(2) (2003), pp. 195–222; Posner, “The Implications of Constructivism”; Marta Reynal-Querol,“Ethnicity, Political Systems, and Civil War,” The Journal of Conflict Resolution 46(1) (2002), pp.29–54; Jose G. Montalvo and Marta Reynal-Querol, “Ethnic Polarization, Potential Conflict, andCivil Wars,” The American Economic Review 95(3) (2005), pp. 796–816.

82. Security force casualties are referred to as “martyrs” in Turkey.83. The list of these websites are available on request.

Appendix

Data Sources for Security Force Casualties

Our main source for military casualties is a publication, named “Our Martyrs(Sehitlerimiz),” published by the Turkish Ministry of Defense in late 1998. The publi-cation lists the names, ranks, and place of death of all Turkish soldiers who died in the fightagainst the PKK insurgency in the August 1984–September 1998 period. In total, the listcontains information on 5,554 soldiers. This publication actually provides the most credi-ble, and accurate information on military casualties as the information is directly taken fromthe personnel archives of the Ministry of Defense. Unfortunately, the list covers casualtiesup to mid-1998, and detailed information on place of death is missing for gendermariecasualties. In order to bring the database up to date with detailed information on places ofdeath, we referred to various other sources. We contacted the Associations of the Familiesof Martyrs82 in 28 provinces. These associations are civil society organizations founded byfamilies of soldiers and police officers who died in service, to help each other cope withthe situation. They helped us greatly in gathering information about the SFCs from theirarea. We have also contacted and obtained information from numerous military bases indifferent parts of the country, and also the Gendarmerie Museum in Ankara, which hasan extensive database on gendarmerie casualties. Another important source has been thearchives of daily newspapers. We searched through the daily archives of five major Turkishnewspapers (Milliyet, Cumhuriyet, Zaman, Sabah, Hurriyet) for news on PKK attacks.Even though the newspapers did not cover all the attacks and all the casualties, they stillprovided information on a good deal of them. The Internet has been another importantsource of information. We have conducted searches using related words like PKK, attack,Kurdish conflict, terror, martyr, and so on and skimmed through thousands of websites thatcame up on these searches. We referred to some 359 of these websites.83 Most of them arecommemorative sites maintained by local administrations including governorships, mu-nicipalities, village associations, and civil society organizations. Finally, we obtained theinformation on the casualties of the police forces directly from the public relations officeof the Administration of Security Forces.

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

Tabl

eA

1E

ffec

tsof

the

ethn

icco

nflic

ton

the

vote

shar

eof

the

othe

rpo

litic

alpa

rtie

s

Num

ber

ofob

s.:1

,727

Num

ber

ofgr

oups

:870

Dep

ende

ntva

riab

le:

AN

AP’

svo

teD

epen

dent

vari

able

:R

P’s

vote

Dep

ende

ntva

riab

le:

CH

P’s

vote

Dep

ende

ntva

riab

le:

DY

P’s

vote

Dep

ende

ntva

riab

le:

DSP

’svo

te

Con

stan

t11

.80∗∗

(5.1

0)32

.82∗∗

∗(5

.33)

−1.8

1(3

.64)

−2.5

9(5

.48)

39.7

4∗∗∗

(4.2

0)SF

Cs

inth

ear

easi

nce

the

prev

ious

elec

tion

−0.0

56∗∗

∗(0

.018

)−0

.094

∗∗∗

(0.0

19)

−0.0

24∗

(0.0

13)

−0.0

13(0

.019

)0.

009

(0.0

15)

Perc

enta

geof

ethn

ical

lyK

urdi

shpo

pula

tion

−0.0

67∗∗

(0.0

34)

−0.0

64(0

.041

)0.

034

(0.0

25)

−0.0

29(0

.033

)−0

.074

∗∗(0

.036

)

Tur

nout

0.01

3(0

.038

)−0

.079

∗∗(0

.039

)−0

.059

∗∗(0

.027

)0.

179∗∗

∗(0

.041

)−0

.141

∗∗∗

(0.0

31)

Popu

latio

nsi

ze−0

.006

∗∗∗

(0.0

02)

0.00

5(0

.003

)−0

.002

(0.0

02)

−0.0

06∗∗

∗(0

.002

)0.

008∗∗

∗(0

.002

)Po

pula

tion

grow

thra

te0.

081

(0.0

81)

0.10

1(0

.082

)−0

.248

∗∗∗

(0.0

55)

−0.1

84∗∗

(0.0

88)

−0.0

08(0

.065

)U

rban

izat

ion

leve

l0.

005

(0.0

16)

0.07

7∗∗∗

(0.0

19)

0.00

1(0

.013

)−0

.031

∗(0

.017

)−0

.032

∗∗(0

.014

)H

ouse

hold

size

0.18

2(0

.271

)1.

110∗∗

∗(0

.296

)0.

023

(0.2

03)

0.17

6(0

.287

)−2

.193

∗∗∗

(0.2

24)

Une

mpl

oym

entr

ate

−0.0

6(0

.079

)−0

.254

∗∗∗

(0.0

81)

−0.0

52(0

.056

)−0

.062

(0.0

85)

0.00

2(0

.063

)L

itera

cyra

te0.

011

(0.0

35)

−0.2

24∗∗

∗(0

.042

)0.

183∗∗

∗(0

.029

)0.

047

(0.0

29)

0.02

9(0

.03)

Ele

ctio

ndu

mm

y6.

155∗∗

∗(0

.422

)4.

883∗∗

∗(0

.431

)2.

795∗∗

∗(0

.297

)6.

451∗∗

∗(0

.468

)−5

.54∗∗

∗(0

.333

)St

ate

ofem

erge

ncy

dum

my

2.64

7∗(1

.534

)2.

528∗

(1.3

37)

0.26

8(0

.897

)0.

531

(1.7

08)

1.76

9(1

.179

)V

aria

nce

Com

pone

nts

Prov

ince

leve

l(ς

2 )4.

96∗∗

∗(0

.442

)6.

92∗∗

∗(0

.629

)3.

63∗∗

∗(0

.362

)4.

25∗∗

∗(0

.407

)7.

04∗∗

∗(0

.637

)C

ount

yle

vel(

τ2 )

3.27

∗∗∗

(0.2

21)

6.67

∗∗∗

(0.2

10)

4.92

∗∗∗

(0.1

49)

2.30

∗∗∗

(0.3

59)

3.53

∗∗∗

(0.1

52)

Obs

erva

tion

leve

l(σ

2 )5.

31∗∗

∗(0

.129

)4.

34∗∗

∗(0

.106

)2.

95∗∗

∗(0

.072

)6.

30∗∗

∗(0

.155

)3.

83∗∗

∗(0

.093

)−2

Log

Lik

elih

ood

11,3

7811

,695

10,4

4711

,660

10,6

90

Est

imat

esfr

omth

em

ixed

effe

cts

RE

ML

regr

essi

ons.

∗∗∗ s

igni

fican

tat1

%le

vel;

∗∗si

gnifi

cant

at5%

leve

l;∗ s

igni

fican

tat1

0%le

vel.

517

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

518 A. Kibris

Table A2Effects of the ethnic conflict on electoral behavior using an alternative conflict measure

# of observations: 1,727 Dependent variable: Dependent variable:# of groups: 870 HADEP’s vote share MHP’s vote share

Constant 5.875∗∗(3.63) 17.77∗∗∗ (5.02)SFCs in the area since the

previous election0.019∗∗∗ (0.003) 0.034∗∗∗ (0.005)

Percentage of ethnicallyKurdish population

0.204∗∗∗(0.024) −0.166∗∗∗ (0.037)

Turnout −0.056∗∗(0.026) 0.071∗∗ (0.036)Population size 0.006∗∗∗(0.002) −0.004∗∗ (0.002)Population growth rate 0.180∗∗∗(0.053) 0.116 (0.077)Urbanization level −0.012(0.013) −0.016 (0.015)Household size 0.971∗∗∗(0.197) −1.213∗∗∗ (0.263)Unemployment rate 0.246∗∗∗(0.053) 0.122 (0.075)Literacy rate −0.067∗∗(0.030) 0.040 (0.035)Election dummy −1.141∗∗∗(0.288) −10.06∗∗∗ (0.398)State of emergency dummy −0.478(0.842) 3.260∗∗ (1.460)Variance ComponentsProvince level (ς2) 3.40∗∗∗(0.373) 6.31∗∗∗ (0.577)County level (τ 2) 5.02∗∗∗(0.148) 3.35∗∗∗ (0.203)Observation level (σ 2) 2.77∗∗∗(0.067) 4.87∗∗∗ (0.121)–2 Log Likelihood 10,347 11,211

Estimates from the mixed effects REML regressions. ∗∗∗significant at 1% level; ∗∗significant at5% level; ∗significant at 10% level.

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

Polarization Trap 519

Table A3Effects of the ethnic conflict on electoral behavior under province level fixed effects and

county level clustered errors specification

Number of observations: Dependent variable: Dependent variable:1,727 HADEP’s vote share MHP’s vote share

Constant 8,651 (7.750) 21.809∗∗∗ (5.827)SFCs since the previous

election0.097∗∗∗ (0.039) 0.072∗∗∗ (0.017)

Percentage of ethnicallyKurdish population

0.134∗∗∗ (0.118) 0.156∗∗∗ (0.090)

Turnout −0.116∗∗ (0.046) 0.095∗∗ (0.039)Population size 0.006∗∗ (0.003) −0.005∗∗ (0.002)Population growth rate 0.116 (0.090) 0.080 (0.086)Urbanization level −0.033∗ (0.016) −0.013 (0.016)Household size 1.024∗∗ (0.403) −1.256∗∗∗ (0.317)Unemployment rate 0.452∗∗∗ (0.122) 0.059 (0.079)Literacy rate −0.068∗∗∗ (0.076) 0.079∗ (0.040)Election dummy −1.037 (0.643) −9.500∗∗∗ (0.439)State of emergency dummy −0.79 (0.665) 6.092∗∗ (0.849)Parameter estimates for

province dummies are notreported.

R-Squared 0.6657 0.6757

Estimates from linear regressions, standard error adjusted for 870 clusters in county. ∗∗∗significantat 1% level; ∗∗significant at 5% level; ∗significant at 10% level.

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

520 A. Kibris

Table A4Effects of the ethnic conflict on HADEP’s vote share when outlier observations are excluded

Dependent variable: Dependent variable:HADEP’s vote share MHP’s vote share

Estimates from the mixedeffects REMLregressions

# of observations: 1,719;# of provinces: 80; # of

counties: 868

# of observations: 1714;# of provinces: 80; # of

counties: 870Constant 6.870∗∗ (2.983) 24.48∗∗∗ (4.682)SFCs since the previous

election0.088∗∗∗ (0.011) 0.079∗∗∗ (0.016)

Percentage of ethnicallyKurdish population

0.207∗∗∗ (0.022) −0.142∗∗∗ (0.036)

Turnout −0.037∗ (0.022) 0.057∗ (0.035)Population size 0.006∗∗∗ (0.002) −0.004∗∗ (0.002)Population growth rate 0.137∗∗∗ (0.045) 0.145∗∗ (0.075)Urbanization level −0.001 (0.011) −0.013 (0.015)Household size 0.834∗∗∗ (0.170) −1.313∗∗∗ (0.251)Unemployment rate 0.213∗∗∗ (0.045) 0.091 (0.071)Literacy rate −0.094∗∗∗ (0.025) −0.021 (0.033)Election dummy −1.05∗∗∗ (0.249) −10.31∗∗∗ (0.383)State of emergency dummy −0.35 (0.696) 3.477∗∗ (1.401)Variance ComponentsProvince level (ς2) 3.22∗∗∗ (0.360) 6.23∗∗∗ (0.566)County level (τ 2) 5.01∗∗∗ (0.141) 3.22∗∗∗ (0.198)Observation level (σ 2) 2.25∗∗∗ (0.055) 4.66∗∗∗ (0.118)–2 Log Likelihood: 9,903 10,983

∗∗∗significant at 1% level; ∗∗significant at 5% level; ∗significant at 10% level.

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

Polarization Trap 521

Table A5Effects of nationalist votes on SFCs

Dependent Variable: Number ofSecurity force casualties in the 1995–1999 period observations: 857

Constant 2.487 (4.215) 2.507 (4.536)Vote share of HADEP in

1995 elections0.076 (0.046)

Vote share of MHP in 1995elections

−0.033 (0.027)

SFCs in the 1991-1995 period 0.391∗∗∗ (0.029) 0.400∗∗∗ (0.031)Percentage of ethnically

Kurdish population−0.024 (0.021) −0.010 (0.019)

Turnout −0.059 (0.029) −0.054 (0.030)Population size −0.001 (0.001) −0.001 (0.001)Population growth rate −0.055 (0.044) −0.039 (0.045)Urbanization level −0.002 (0.009) −0.005 (0.009)Household size −0.150 (0.202) −0.105 (0.227)Unemployment rate 0.032 (0.087) 0.074 (0.090)Literacy rate 0.043 (0.032) 0.039 (0.034)Border dummy −1.672 (1.463) −1.235 (1.493)State of emergency dummy 2.596∗ (1.347) 2.346 (1.486)R-Squared F(12, 78) 0.7912 35.29 0.7879 30.62

Linear regressions with robust, clustered errors at the province level. ∗∗∗significant at 1% level;∗∗significant at 5% level; ∗significant at 10% level.

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

Tabl

eA

6C

orre

latio

nsbe

twee

nco

ntro

lvar

iabl

es

SFC

sU

rban

izat

ion

Popu

latio

ngr

owth

Kur

dish

popu

latio

n

Stat

eof

emer

genc

ydu

mm

yL

itera

cyra

teH

ouse

hold

size

Popu

latio

nT

urno

utE

lect

ion

dum

my

SFC

s1

Urb

aniz

atio

n−0

.01

1Po

pula

tion

grow

th0.

020.

281

Kur

dish

popu

latio

n0.

46−0

.03

0.06

1

Stat

eof

emer

genc

ydu

mm

y

0.47

−0.0

30.

020.

871

Lite

racy

rate

−0.3

90.

350.

10−0

.69

−0.6

01

Hou

seho

ldsi

ze0.

26−0

.24

0.02

0.65

0.53

−0.6

61

Popu

latio

n−0

.01

0.40

0.28

0.04

−0.0

070.

12−0

.09

1T

urno

ut−0

.26

0.06

0.06

−0.3

7−0

.31

0.41

−0.4

4−0

.07

1E

lect

ion

dum

my

0.08

−0.1

80.

11−0

.000

30.

03−0

.38

0.09

−0.0

4−0

.34

1

522

Dow

nloa

ded

by [

Koc

Uni

vers

ity]

at 1

0:27

10

Febr

uary

201

6

Is Islam the Solution? ComparingTurkish Islamic and SecularThinking toward Ethnic andReligious Minorities

Murat Somer and Gitta Glupker-Kesebir

Can Islamic political actors manage ethnic diversity better thansecular political actors? From Muslim Brothers to ISIS (Islamic Statein Iraq and Syria), Islamists with very different orientations havelong claimed that they can “absorb and resolve ethnic conflicts onthe basis of Muslim unity and brotherhood.”1 In short, they assertthat “Islam is the solution.” At first sight validating these claims,Turkey’s ruling pro-Islamic Justice and Development Party (AKP)has presided over major initiatives to resolve the country’s long-festering Kurdish conflict. It launched major legal-political reforms

MURAT SOMER (PhD, University of Southern California, Los Angeles) is associateprofessor of political science and international relations at Koc University inIstanbul, Turkey. He is the author of Return to Point Zero: From Nation-State toState-Nation, the Three Dilemmas of the Turkish and Kurdish Question (2015).His articles have appeared in journals including Comparative Political Studies,Democratization, Turkish Studies, Journal of Church and State, Middle EastJournal, and Third World Quarterly. Special interests include religious andsecular politics, democratization, and ethnic identities and conflict. GITTAGLUPKER-KESEBIR (BA, MA, University of Osnabruck, Germany) is a doctoral can-didate in the Department of International Relations at Koc University in Istanbul,Turkey. Her articles are forthcoming and have appeared in Journal of Women, Pol-itics and Policy, and Journal of Contemporary European Research, in respectiveorder. Special interests include public opinion research, European integration,EU policies and institutions, women and politics, and women’s rights. Drafts ofthe present essay were presented by Murat Somer at the 18th Annual Associationfor the Study of Nationalities World Convention, Columbia University, New York,April 18–20, 2013, and the 14th Mediterranean Research Meeting, organized bythe European University Institute, Robert Schuman Centre for AdvancedStudies and Mersin University, Mersin, Turkey, March 20–21, 2013.

Journal of Church and State; doi:10.1093/jcs/csv031# The Author 2015. Published by Oxford University Press on behalf of the J. M. DawsonInstitute of Church-State Studies. All rights reserved. For permissions, pleasee-mail: journals.permissions@oup.com

1. R. Hrair Dekmejian, “The Anatomy of Islamic Revival: Legitimacy Crisis, EthnicConflict and the Search for Islamic Alternatives,” Middle East Journal 34, no. 1(1980): 1–12.

1

Journal of Church and State Advance Access published June 3, 2015 by guest on June 4, 2015

http://jcs.oxfordjournals.org/D

ownloaded from

and an ongoing peace process with the Kurdistan Workers Party (PKK).These initiatives have been undermined by developments in neighbor-ing IraqandSyria;ambiguitiesofvision, trustandcommitmentonbothsides; and cycles of government oppression and PKK militancy. Never-theless, thegovernment’sboldstepsandthelimitedyetunprecedentedcultural and educational rights it has secured have led many observersto conclude that Turkish Islamists have more ideological potential tosuccessfully manage ethnic diversity than their secular counterparts.2

Cross-country evidence does not suggest that majority-Muslimcountries and, for that matter, Islamic regimes such as Iran orSudan necessarily fare better in terms of resolving ethnic-nationalproblems; notably, the Islamic Republic of Iran has a major Kurdishconflict, just like the constitutionally secular Turkey.3 Still, Islamistscould argue that the domestic and international political settings in

2. HakanYavuz,andNihatAli Ozcan, “TheKurdishQuestionandTurkey’s Justiceand Development Party,” Middle East Policy XIII, no. 1 (2006): 102–19; HuseyinYayman, Turkiye’nin Kurt Sorunu Hafizasi [Turkey’s Memory of the KurdishProblem] (Istanbul, Turkey: Dogan Kitap, 2011); Gabriel Mitchell, Gabriel, “Islamas Peacemaker: The AKP’s Attempt at a Kurdish Resolution,” The WashingtonReview of Turkish and Eurasian Affairs, May 2012, http://www.thewashingtonreview.org/articles/islam-as-peacemaker-the-akps-vision-of-a-kurdish-resolution.html. Other analyses reach more nuanced and qualified but similarconclusions regarding the relative potential of Islamist ideology to resolve theKurdish conflict and other minority problems. See Sener Akturk, Regimes ofEthnicity and Nationhood in Germany, Russia, and Turkey (Cambridge, UK:Cambridge University Press, 2012); Jenny White, Muslim Nationalism and theNew Turks (Princeton, NJ: Princeton University Press, 2013).3. For the relation between ethnic divisions and conflict, see, for example, CarstenJensen and Svend-Erik Skaaning, “Modernization, Ethnic Fractionalization, andDemocracy,” Democratization 19, no. 6 (2012), 1117–37; Klaus Desmet, IgnacioOrtuno-Ortın, Romain Wacziarg, “The Political Economy of Linguistic Cleavages,”Journal of Development Economics 97 (2012): 322–38; Jose G. Montalvo and MartaReynal-Querol,“EthnicPolarization,PotentialConflict,andCivilWars,”TheAmericanEconomic Review 95, no. 3 (June 2005): 796–816. For Iran, see Nader Entessar, “TheKurdishNationalMovement in IranSince the IslamicRevolutionof 1979,” inTheEvo-lution of Kurdish Nationalism, ed. Mohammed M. A. Ahmed and Michael M. Gunter(Costa Mesa: Mazda Publishers, 2007), and Gunes M. Tezcur, Muslim Reformers inIran and Turkey: The Paradox of Moderation (Austin: University of Texas Press,2010). Numerous other Muslim states experience ethnic conflicts—for example,Indonesia: EdwardAspinall, IslamandNation:SeparatistRebellion inAceh, Indonesia(Stanford,CA:StanfordUniversityPress,2009);Malaysia:JensenandSkaaning,“Mod-ernization,” 1117–37; Pakistan: Theodore P. Wright Jr., “Center-Periphery RelationsandEthnicConflict inPakistan:Sindhis,Muhajirs,andPunjabis,”ComparativePolitics23,no.3 (1991):299–312;AlyssaAyres,SpeakingLikeAState: LanguageandNation-alism in Pakistan (Cambridge, UK: Cambridge University Press, 2009); and FarhanHanif Siddiqui, in Julien Levesque, “Managing Diversity in Pakistan: Nationalism,Ethnic Politics and Cultural Resistance [review essay],” South Asia MultidisciplinaryAcademic Journal, 2013, http://samaj.revues.org/3551; Algeria, Syria, Sudan, Iraq,Lebanon, and Yemen: Saad Eddin Ibrahim, “Ethnic Conflict and State-Building inthe Arab World,” International Social Science Journal 156 (1998): 229–42.

Journal of Church and State

2

by guest on June 4, 2015http://jcs.oxfordjournals.org/

Dow

nloaded from

which they have operated in modern times have not allowed them tofully put their ideas into practice.4 Thus, it is hard to accept or refutethe Islamist peace claims based on the actual performance record ofIslamic versus secular political actors.

Hence, this article evaluates the merit of the “Islam is the solution”thesismainlyonan ideational levelbycomparingTurkish Islamicandsecular actors in terms of their ideas and proposals. We do so basedon a comprehensive content analysis of Islamic and secular press,complemented by other evidence such as political party reportsand opinion polls. The content analysis compares representativesamples of Islamic and secular writers. It examines the contents ofmore than forty-one thousand newspaper articles between 1996and 2004, and in 2007 and 2010, with respect to subjects includingnationalism, ethnic-religious pluralism, conflict resolution, andissues related to Kurds and the Kurdish question itself.5 We drawon wide samples and specific issues—rather than abstract theologi-cal principles or the visions of a few renowned intellectuals. Bydoing so, our goal is to assess how well Turkish Islamic actors are pre-paredtoaddressethnicproblems intermsofvaluesand ideasthatarewidely circulated within their group discourse and in response to realpolitical events.

While asking these questions, we will not investigate whether secu-larism or secular ideas played any historical role in the emergence ofethnic-national problems. Islamist claims often combine two caus-ally and analytically separable arguments. The historical argumentcan be called the “secularism was to blame” argument. Islamistsmaintain that modern ethnic conflicts in Muslim majority statesresulted from the establishment of nation states and are productsof secular nationalist ideologies and national identities, which theytend to ipso facto associate with ethnicity, race, and often, antireli-giouspolitics.TheprescriptiveargumentassertsthatasharedIslamicidentity and ideology and ideas, values, and principles inspired byIslam would be an effective recipe to address contemporary

4. In thewordsofoneauthor,Bengalinationalismgrewin“EastPakistan”becauseof the “publicpolicy failure to develop a religious identity based on Islamic ideas,”even though Islam constitutionally became the basis of nationhood for allIndian Muslims. See Muhittin Ataman, “Islamic Perspective on Ethnicity andNationalism: Diversity or Uniformity?” Journal of Muslim Minority Affairs 23,no. 1 (2003): 98.5. For the methodology and other findings from the content analysis, see MuratSomer, “Media Values and Democratization: What Unites and What DividesReligious-Conservative and Pro-Secular Elites?” Turkish Studies 11, no. 4 (Decem-ber 2010): 555–77; Murat Somer, “Does It Take Democrats to Democratize?:Lessons From Islamic and Secular Elite Values in Turkey,” Comparative PoliticalStudies 44, no. 5 (2011): 511–45. Research on 2007 and 2010 were conducted in2012 and 2013.

Is Islam the Solution?

3

by guest on June 4, 2015http://jcs.oxfordjournals.org/

Dow

nloaded from

ethnic-regional conflicts. In fact, even if the historical argument wereshown to be true, this would not imply that the prescriptive claimshould necessarily be valid. Thus, we evaluate the prescriptiveclaim by itself.

Prior to further examination, a brief note on the terminology is inorder. We use the terms Islamic and secular loosely and for lack ofbetter terms whenever an Islamic or secular orientation forms amajor dimension of a social-political actor’s identity, respectively.We use the terms Islamist or secularist whenever a pro-Islamicor pro-secular ideology or social-political agenda seems to be evenmore pronounced. These categories may share many traits andare internally diverse. For example, Islamists, who invariably comefrom Sunni Islam in Turkey, include actors who fight to replacesecular democracy with a Sharia-based system,6 as well as peoplewho pursue social-cultural and ideological Islamization withoutnecessarily overhauling existing state structures. Similarly, whileauthoritarian-secularists defend the legacy of Ataturk’s (1881–1938) secular-Westernizing reforms by trying to forcefully excludeIslam from the public sphere, more democratic-minded secularistsuphold the same legacy without necessarily denying Islam a public-political role. Nevertheless, our empirical analysis shows that theelites writing in the newspapers we name Islamic and secular formtwo distinct groups in terms of their average views on issues relatedto religion and secularism.

The Political and Academic Debate and Our SpecificHypotheses

In an attempt to achieve regime legitimacy and social-political unitybased on an overriding common identity, the Turkish Constitutiondefines all Turkish citizens simply as Turks, regardless of ethnicity,race, and religion. But the actual practice heavily favors SunniMuslims and Turkish ethnicity, culture and identity. For example,while the “teaching of” minority languages is now possible as electivecourses, the constitution bans “teaching in” minority languages,demanding that Turkish remains the language of education.7 Nolegal-formal recognition, separate rights, or autonomy are offered

6. A numerically and politically marginal group in the Turkish case.7. Non-Muslimminoritieswhomakeuplessthan1percentof thepopulationhavelegal protection and their own cultural and educational institutions but sufferfrom many problems of discrimination and oppression. See Peter AlfordAndrews, Ethnic Groups in the Republic of Turkey, compiled and edited with theassistance of Rudiger Benninghaus (Wiesbaden: Reichert, 1989); Ayhan Kaya,Europeanization and Tolerance in Turkey: The Myth of Toleration (New York:Palgrave Macmillan, 2013).

Journal of Church and State

4

by guest on June 4, 2015http://jcs.oxfordjournals.org/

Dow

nloaded from

to numerous ethnic-linguistic and sectarian groups Turkey harborssuch as the Alevis, Arabs, Bosnians, Circassians, and the Laz. Manyof these groups have politically mobilized, demanding more recogni-tion and rights, especially since the 1990s. However, by far the mostimportant ethnic-national question has been the Kurdish conflict.

Turkish national censuses do not include questions on ethnicity andrace. Depending on whether Kurdishness is defined based on mothertongue or self-identification, Kurds are estimated to constitutebetween15and19percentofTurkey’spopulation.Approximatelytwo-thirds of the Kurdish population are concentrated in eastern Turkeyand form the majority in areas bordering the predominantly Kurdishregions of neighboring Iraq and Syria.8 Turkey’s Kurdish conflict goesback to the late Ottoman period when state elites tried to fostercommon nationhood among ethnically Turkish, Kurdish, and otherMuslims in the name of Ottoman state centralization-modernization.After the foundation of modern Turkey in 1923, even more radicalsteps of nation building and nation-state formation were implementedin Kurdish areas. This time, thesewere enforced in the name of Turkishnationalism and within a narrower geography because the OttomanKurds were divided between Turkey and the British and French man-dates of Iraq and Syria.9 A seriesofKurdish rebellions and state repres-sion followed during the 1920s and 1930s.10 An assimilationist statepolicy was continued under successive military interventions andelected center-right and center-left governments, which denied eventhe existence of a Kurdish minority until the 1990s. The Turkish-Islamic synthesis ideologyof the 1980–83 military rule reinforced thispattern in the name of both religion and nationality. Until the last de-cade, mainstream Turkish actors mainly perceived Turkish nationalismas nonethnic and inclusive, Kurds as actual or “potential” Turks, and

8. Server Mutlu, “Ethnic Kurds in Turkey: A Demographic Study,” InternationalJournal of Middle East Studies 28, no. 4 (1996): 517–41; Ayse Gunduz-Hosgor,and Jeroen Smits, “Intermarriage between Turks and Kurds in ContemporaryTurkey: Interethnic Relations in an Urbanizing Environment,” European Sociolog-ical Review 18, no. 4 (2002): 417–32; Martin van Bruinessen, Aga, Seyh, Devlet[Agha, Sheik, State] (Istanbul: Iletisim, 2003); Konda, Kurt Meselesi’nde Algi veBeklentiler [Perceptions and Expectations in the Kurdish Question] (Istanbul:Iletisim Yayinlari, 2011).9. Robert Olson, The Kurdish National Movement in the 1990s: Its Impact onTurkey and the Middle East (Lexington: University Press of Kentucky, 1996);David McDowall, A Modern History of the Kurds (New York: St. Martin’s Press,1997); Kemal H. Karpat, Ottoman Past and Today’s Turkey (Leiden: Brill, 2000).10. MesutYegen, “TheTurkishStateDiscourseandtheExclusionofKurdishIden-tity,” Middle Eastern Studies 32, no. 2 (1996): 216–29; Kemal Kirisci, and GarethM. Winrow, The Kurdish Question and Turkey: An Example of a Trans-StateEthnic Conflict (London: Frank Cass, 1997); Hakan Ozoglu, Kurdish Notablesand the Ottoman State: Evolving Identities, Competing Loyalties, and ShiftingBoundaries (New York: State University of New York Press, 2004).

Is Islam the Solution?

5

by guest on June 4, 2015http://jcs.oxfordjournals.org/

Dow

nloaded from

the public expression of ethnic pluralism as a threat to national andterritorial unity.11

Up to forty-two thousand people lost their lives due to the war withthe PKK, which has fought the Turkish state since 1984 for Kurdishrights and self-rule from both within Turkey and neighboring Iraqand Syria.12 Short-lived attempts to recognize Kurdish identityduring the 1990s failed to solve the conflict in an environment of vio-lence, weak governments, and military tutelage over civilian politics.A period of relative peace and democratic reforms began in 1999 withthe PKK leader Ocalan’s capture—whose death sentence was laterconverted to a life sentence—and the EU’s recognition of Turkeyas a candidate for membership. This period witnessed relativedemocratization, the beginning of EU accession negotiations in2005, and many unprecedented EU-inspired reforms regarding theKurdish question, even though these reforms fell far short ofKurdish nationalists’ expectations.13 The PKK resumed its armedstruggle in 2004, as the regional context of the conflict changedwith the war and the emerging autonomous Kurdish entity inIraq.14 The AKP has ruled the country since the end of 2002. Helpedby the weakening of military supervision over Turkish politics, theparty tried to address the Kurdish question through culturalreforms such as a new public TV channel in Kurdish and political“Kurdish openings” in 2005, 2009, and 2013.15 These reforms andsteps challenged many assimilationist creeds of mainstreamTurkish nationalism and state ideology.

11. Murat Somer, “Turkey’s Kurdish Conflict: Changing Context and Domesticand Regional Implications,” Middle East Journal 58, no. 2 (2004): 235–53; MesutYegen, “Turkish Nationalism and the Kurdish Question,” Ethnic and RacialStudies 30 (2007): 119–51.12. Gunes M. Tezcur, “The Ebb and Flow of Armed Conflict in Turkey: AnElusive Peace,” in Conflicts, Democratization and the Kurds in the Middle East,ed. David Romano and Mehmet Gurses (New York: Palgrave Macmillan, 2014),171–88.13. Somer, “Turkey’s Kurdish Conflict: Changing Context and Domestic andRegional Implications,” 235–53; Akturk, Regimes of Ethnicity and Nationhoodin Germany, Russia, and Turkey; White, Muslim Nationalism and the New Turks.14. Gunes M. Tezcur, “When Democratization Radicalizes: the Kurdish National-ist Movement in Turkey,” Journal of Peace Research 47, no. 6 (2010): 775–89.15. Yavuz and Ozcan, “The Kurdish Question and Turkey’s Justice and Develop-ment Party,” 102–19; Murat Somer and Evangelos G. Liaras, “Turkey’s NewKurdish Opening: Religious and Secular Values,” Middle East Policy XII, no. 2(2010): 152–65; Akturk, Regimes of Ethnicity and Nationhood in Germany,Russia, and Turkey; Teczur, “The Ebb and Flow of Armed Conflict in Turkey: AnElusive Peace,” 171–88.

Journal of Church and State

6

by guest on June 4, 2015http://jcs.oxfordjournals.org/

Dow

nloaded from

All this ledmanypoliticalactorsandobserverstounequivocallycel-ebrate the AKP’s Islamic identity and ideology.16 Many nuanced andbalancedstudiesalsoreckonedthatTurkey’snewIslamicor “Muslim-nationalist”political actorshavemore ideological promise toaddressethnic conflicts; they ostensibly displayed “greater sensitivity tominority and individual rights” than their secular rivals, whichmakes “[Kurdish openings] possible.”17 White concluded: “[W]hilenot immune from militant nationalism [. . . and sharing] with Kemal-ist nationalism a belief in the superiority of Turkishness . . . theMuslim definition of the Turkish nation appears to be less boundary-and less blood-driven than that held by Kemalistnationalists. Muslimnationalists consciously model their concept of the nation on the his-torical, flexibly bounded, and multinational Ottoman Empire.”18

Akturk agreed that a new “counterelite” armed with an “Islamist mul-ticulturalist new thinking about ethnicity” established its politicalhegemony and moved Turkey to multiculturalism.19

The debate is not new. Turkish Islamist thinkers have long main-tained that “by representing a secular, ‘antireligious nationalism’based on ethnicity, official Turkish nationalism contributed to therise of its twin sister, Kurdish nationalism.”20 “Once Islamic valueswere replaced by Western rules of culture and thinking, Islam lostits roleasareligioncalling forpeaceandpromotingmutualsympathyamong the fellow Muslim Turks and Kurds. As a result, brotherly feel-ingstowardeachothergavewayto ‘stranglingeachother.’”21 Necmet-tin Erbakan (1926–2011), former prime minister and the founder ofTurkey’s first explicitly Islamist political movement and party, putit succinctly in 1994:

Forcenturies, the children of this country have started the school day citingthe besmele (in the name of God). What did you replace it with? “I amTurkish, I am righteous, I am diligent.” When you did this, it became rightful

16. Yavuz and Ozcan, “The Kurdish Question and Turkey’s Justice and De-velopment Party,” 102–19; Yayman, Turkiye’nin Kurt Sorunu Hafizasi [Turkey’sMemory of the Kurdish Problem]; “Islam as Peacemaker: The AKP’s Attempt at aKurdish Resolution,” Washington Review of Turkish and Eurasian Affairs,http://www.thewashingtonreview.org/articles/islam-as-peacemaker-the-akps-vision-of-a-kurdish-resolution.html.17. White, Muslim Nationalism and the New Turks, 53.18. Ibid., 53, 95–96.19. Akturk, Regimes of Ethnicity and Nationhood in Germany, Russia, andTurkey, 39.20. Umit Cizre Sakallioglu, “Kurdish Nationalism from an Islamist Perspective:The Discourses of Turkish Islamist Writers,” Journal of Muslim Minority Affairs18, no. 1 (1998): 77. Sakallioglu offers a detailed and insightful examination ofIslamic writing on the Kurdish question in Turkey.21. Ibid.

Is Islam the Solution?

7

by guest on June 4, 2015http://jcs.oxfordjournals.org/

Dow

nloaded from

forour Muslim child of Kurdish origin to say “I am Kurdish, I am righteous, Iam more diligent.”22

How would Islamic actors address the Kurdish conflict, which alleg-edly resulted from secular-nationalism? Hayrettin Karaman, an Islam-ist theologian influential among AKP circles maintains that “under anIslamic state, citizenship would be replaced by religion, and full equal-ity would be established based on religious unity and being just.”23

Both Erbakan’s party and the AKP have been popular among Kurdishvoters.24 It has been argued that the AKP identifies secularism “asa cause of division between Turks and Kurds” and offers “its ownsolution—‘Islam as cement’—to end the societal polarization ofTurkey.”25 President and former AKP leader and Prime MinisterErdogan condemned both Turkish and Kurdish “ethnic nationalisms”and highlighted the value of unification and brotherhood on the basisof “common citizenship” in the Republic of Turkey.

Erdogan’s understanding of common “citizenship” (Turkish vatan-das) has a historical and emotive content rooted in Islamic brother-hood and common Ottoman-Islamic ancestry.26 His 2011 speech inDiyarbakır, the symbolic capital of Turkey’s predominantly KurdishSoutheast, began with a long opening filled with references toMuslim ancestry and brotherhood:27

We are all descendants of the great soldiers of the great leader SelahaddinEyyubi [a historical Kurdish-Muslim hero who battled against theCrusaders and founded the Ayyubid dynasty] who conquered Palestineand Jerusalem. . . . I am against both Turkish and Kurdish nationalisms.

22. Rusen Çakir, “Erbakan 20 yil once ne yapmisti?” [“What Did Erbakan Do 20Years Ago”], T24, January 3, 2012, http://t24.com.tr/haber/rusen-cakir-erbakan-20-yil-once-ne-yapmisti,189772. Also see Yayman, Turkiye’nin KurtSorunu Hafizasi [Turkey’s Memory of the Kurdish Problem], 417.23. Hayrettin Karaman, “Altan Tan Kardesime” [“To My Brother Altan Tan”], YeniSafak, May 24, 2012.24. Seda Demiralp, “The Odd Tango of the Islamic Right and Kurdish Left inTurkey: A Peripheral Alliance to Redesign the Centre?” Middle Eastern Studies48, no. 2 (2012): 287–302; Zeki Sarigil and Omer Fazlioglu, “Religion and Ethno-nationalism: Turkey’s Kurdish Issue,” Nations and Nationalism 19, no. 3 (2013):551–71.25. Yavuz and Ozcan, “The Kurdish Question and Turkey’s Justice and Develop-ment Party,” 103.26. Rusen Çakir, “Islam kardesligi” Kurt sorununa cozumun anahtari olabilirmi?” [“Can Islamic Brotherhood Be the Key to a Solution for the Kurdish Ques-tion”], Vatan, June 2, 2011. Also see, “Erdogan: Milliyetcilik ayak altinda,”Hurriyet, February 18, 2013, http://www.hurriyet.com.tr/gundem/22621388.asp.27. AK Party (2011), 1 Haziran Diyarbakir Mitingi konusmasinin tam metni (full textof the speech in the June 1rally), http://www.akparti.org.tr/site/haberler/1-haziran-diyarbakir-mitingi-konusmasinin-tam-metni/8230.

Journal of Church and State

8

by guest on June 4, 2015http://jcs.oxfordjournals.org/

Dow

nloaded from

All the Kurds and Turks are my brothers. . . . We turn our faces towards thesame qibla [the Kaaba in Mecca, the holiest place of Islam].

But many scholars have approached the employment of Islam as ameans to address ethnic conflicts with suspicion and reservation.Sakallıoglu observes that even though Turkish and Kurdish Islamistwriters find a common discourse and inspiration in Islam, theirperceptionsoftheKurdishquestionandcoursesofactiondiffersignif-icantly: “Kurdish-Islamist writers tend to search for a ‘space’ forKurdish ethnic distinctiveness within the framework of the suggestedformula of ummah, the Islamic community of the faithful, while theposition of the Turkish-Islamist writers leans heavily toward defend-ingthe integrityoftheTurkishstateratherthantowardacknowledgingKurdishethnic distinctiveness.”28 Houston remarksthat both“statist”Islamists such as those associated with the Gulen community and“Turkish Islamists” want the Kurds to subordinate their Kurdish dif-ference to “Turkish-Muslim” and “Muslim” identities, respectively.29

Similarly, Sarigil and Fazlioglu, and Gurses note that Kurdish Islam-ists—including those groups who draw on the very same textualsources as their Turkish counterparts—rely on Islamism to furtherKurdish causes and think that Turkish Islamists instrumentalize theIslamic fraternity discourse to suppress Kurdish interests.30 Yavuzand Sakallıoglu maintain that the employment of Islamic medium inKurdish politics results in the ethnicization of Islamic politics as muchas the other way around.31 Sarigil and Fazlioglu also find that calls forIslamic unity are received differently based on sectarian differencesamong Kurds. Other authors such as Kandiyoti add that, even if Islamicpolicies can help to bridge ethnic differences, a byproduct would be theundermining of rights and freedoms for groups such as women.32

Amoresubtle criticism bySakallıogluassertsthat Islamist thinkingon complex ethnic-national questions lacks sufficient philosophicaland historical depth and overlooks the importance of institutionsand distributional conflicts:

28. Sakallioglu, “Kurdish Nationalism from an Islamist Perspective: The Dis-courses of Turkish Islamist Writers,” 74.29. Christopher Houston, Islam, Kurds, and the Turkish Nation State (New York:Berg, 2003), 161, 172.30. Zeki Sarigil and Omer Fazlioglu, “Religion and Ethnonationalism: Turkey’sKurdish Issue,” Nations and Nationalism 19, no. 3 (2013), 551–71; MehmetGurses, “Is Islam a Cure for Ethnic Conflict? Evidence from Turkey,” Religionand Politics 8, no. 3 (2015), 135–54.31. Hakan Yavuz, “Five Stages of the Construction of Kurdish Nationalism inTurkey,” Nationalism & Ethnic Politics 7, no. 3 (2001): 1–24; Sakallioglu,“Kurdish Nationalism from an Islamist Perspective,” 73–89.32. Kandiyoti, “BetweentheHammerandtheAnvil: Post-ConflictReconstruction,Islam and Women’s Rights,” Third World Quarterly 28, no. 3 (2007): 503–17.

Is Islam the Solution?

9

by guest on June 4, 2015http://jcs.oxfordjournals.org/

Dow

nloaded from

Bychanneling thediscussionof theemergenceofKurdishnationalismto its“roots causes” identified as Ottoman-Turkish modernization, [Islamists]shift the focus away from distribution and concentration of politicalpower, the role of institutions and strategies of control. . . . Instead, spread-ing . . . European consciousness . . . is assumed to give rise to what it doeseverywhere else, that is, the dissipation of religious values and beliefs asthe most pervasive guide to moral conduct. The precondition for national-ism [and ethnic conflicts] is then reduced to a decline in moral values.33

Another critique, Sami Zubaida holds that across different coun-tries Islamists in power have poor records of managing ethnic con-flicts within a framework of pluralism:

As an opposition movement, Islamism may coincide with particularaspects of Kurdish struggles. But Islamists in power are no more attractedtodemocracy,pluralism,andtheruleof lawthantheirsecularcounterparts.They can be self-righteous in their rejection of democracy and pluralismas imperialist, Western divisive poisons. The organicist emphases of Islam-ist ideologies preclude pluralism, including national pluralism. Notethe antagonism of Algerian Islamists to Berber national expressions. . . .[Islamism] is . . . more likely to attempt to eliminate the question in thename of Islamic identity, much as Ataturk denied Kurdish ethnic identityin favor of a national unity.34

Which side is right in this ongoing academic and political debate?The question is not only to what extent Islamists can transcendnationalism. From Israel to Northern Ireland to Muslim cases, it iscommon that religious actors can embrace nationalism as much asthe secular.35 Regardless of their level of nationalism, another ques-tion is whether Islamic actors may be readier to address ethnic-regional questions than secular actors because they have moreideational preparation for specific reforms. Hence, the followinghypotheses can be put forward:

(1) Islamic elites and/or constituencies are less nationalist thantheir secular counterparts.(2) Islamic elites and/or constituencies talk and deliberate onspecific policies and institutional solutions such as Kurdish edu-cation and autonomy more frequently and more favorably thantheir secular counterparts.

33. Sakallioglu, “Kurdish Nationalism from an Islamist Perspective,” 80.34. Sami Zubaida, “Introduction,” in The Kurds: A Contemporary Overview, ed.Philip G. Kreyenbroek and Stefan Sperl (London: Routledge, 1991), 6–7.35. James P. Piscatori, Islam in A World of Nation-States (Cambridge, UK:Cambridge University Press, 1986); Steve Bruce, Politics and Religion (Cambridge,UK: Polity Press, 2003); Uri Ram, “Why Secularism Fails? Secular Nationalism andReligious Revivalism in Israel,” International Journal of Politics, Culture andSociety 21 (2008): 57–73.

Journal of Church and State

10

by guest on June 4, 2015http://jcs.oxfordjournals.org/

Dow

nloaded from

A third hypothesis can also be formulated. The Islamic peace argu-ment is based in the premise that ethnic minority identities competewith secular national identities because they are both ethnic, but notwith Islamic common identities. If this is the reason why Islamicactors are not threatened by ethnic minority identities, then the rela-tionship should be reversed with respect to religious, sectarian, andsecular minorities:

(3) Islamic actors are less open to accommodating non-Muslimminority and Muslim sectarian differences, perceived “heterodoxMuslim” minorities, and perceived “un-Islamic” minorities suchas gays.

Do Islamists Have More Ideational Potentialto Manage Ethnic Conflict?

The empirical evidence in this section comes from three sources. Themain one is a systematic content analysis of Turkish newspapers.36

Trained graduate student coders analyzed the contents of morethan forty thousand articles in printed versions of 4,850 newspaperissues for the time period from 1996 to 2004. Three Islamic newspa-pers (Zaman, Yeni Safak, and Milli Gazete) and two secular newspa-pers (Cumhuriyet and Milliyet) were analyzed.37 A supplementaryfourteen hundred articles from 384 newspaper editions of Zaman,Yeni Safak, Milliyet, and another secular newspaper (Hurriyet) werecoded later for the years 2007 and 2010.38 Coders covered thewhole content of the papers except for the sports sections and adver-torials. In the interest of space, we only show the findings from 2007and 2010 whenever theyadded significant information to those from1996 to 2004. In categorizing the papers into religious and secular,we went by their average views on religion and secularism andacknowledge that there is no shared ideology encompassing all ofthe writers in a particular paper.39

36. Somer, “Media Values and Democratization: What Unites and What DividesReligious-Conservative and Pro-Secular Elites?” 555–77; Somer, “Does It TakeDemocrats to Democratize?” 511–45.37. Ideologically and to some extent organically, Zaman is associated with theIslamic Gulen movement, Yeni Safak with the AKP, and Milli Gazetewith the afore-mentioned Erbakan’s National Outlook movement. Cumhuriyet and Milliyethave leftist-nationalist and social democratic–liberal orientations, respectively.38. This second analysis covered fewer newspapers and issues due to budgetarylimitations.39. Somer, “Media Values and Democratization,” 555–77. Unless stated other-wise, all the differences between secular and religious newspaper groups andbetween the individual papers were found to be statisically significant at leastwithin the 99 percent confidence interval, according to standard two-tail t tests.

Is Islam the Solution?

11

by guest on June 4, 2015http://jcs.oxfordjournals.org/

Dow

nloaded from

Complementing the findings from this research,we invoke findingsfrom Konda, a nationwide representative survey that was conductedon July 17–18, 2010, with 10,393 respondents.40 The third source isseventeen reports commissioned by secular and Islamic politicalparties during the 1990s and 2000s.

Hypothesis 1: Are Islamists Less Nationalist?

Are Islamists Less Nationalist in General?

The subject of nationalism was defined prior to the content analysis asanyviewthatfullyorpartiallyaddressedtheideathatTurksshoulduniteand/or defend their identity, interests, culture, nation-state, nationalsecurity, national unity, territorial integrity against domestic and exter-nalthreatscomingfromnon-Turkishactorsand/orenemies.Thecodersfirst identified any articles that fully or partially addressed this idea ordiscussed the idea of nationalism in general. Then they coded whetherthe article included views that were supportive (positive), critical (nega-tive), orneutral towardnationalism. Ifanarticle includedmorethanonetypeofthesenormativejudgments,thencodersmarkedall thatapplied.

At first sight, the findings seem to support the Islamic peace thesis.In figure 1, the numbers in parentheses indicate the total numbers ofcoding in each category. On average, the religious newspapers con-tained less supportive and more critical views on nationalism thanthe secular papers.

Closer examination, however, reveals a different picture. Figure 2ranks individual papers according to their support of nationalism.“S” and “R” next to the name of each paper indicate whether thepaper is secular or religious, respectively. In terms of positive evalu-ations of nationalism, the most nationalist newspaper is secularistCumhuriyet, and the next two most nationalist are Islamic Zamanand Milli Gazete.41 The least nationalist are the Islamic Yeni Safakand the secular Milliyet. Thus, views on nationalism differ along aspectrum that cuts across the religious–secular cleavage.42 And, in2007 and 2010, the secular Hurriyet and Milliyet were the mostcritical of nationalism, although they also entailed a higher percent-age of supportive views than Zaman and Yeni Safak.

40. Konda, Kurt Meselesi’nde Algi ve Beklentiler [Perceptions and Expectations inthe Kurdish Question].41. The difference between Milli Gazete and Zaman was significant only at the90 percent confidence interval.42. See Somer and Liaras, “Turkey’s New Kurdish Opening: Religious and SecularValues,” 152–65, for additional findings on the use of the term separatists(bolucu).

Journal of Church and State

12

by guest on June 4, 2015http://jcs.oxfordjournals.org/

Dow

nloaded from

Figure 1 The normative views on nationalism in secular and Islamic press,1996–2004.

Figure 2 Normative evaluation of nationalism in individual newspapers.

Is Islam the Solution?

13

by guest on June 4, 2015http://jcs.oxfordjournals.org/

Dow

nloaded from

Are Islamists Less Turkish Nationalist in the Sense of Their Conceptionof National Identity?

DoIslamicactors imagine national identity inaway thatenablesthemto be more open to ethnic-national diversity? As figure 3 shows, thereis limited evidence supporting this possibility. As expected, Islamicpapers put more emphasis on Islam than secular papers in definingnational identity. But thedifference on Turkishness isnot great, espe-cially if one leaves out Milli Gazete, the most Islamist among the threeIslamic papers. Sixty-two percent of the references to national iden-tity drew on Turkishness in Yeni Safak and Zaman combined, com-pared with 79 percent in Milliyet and Cumhuriyet combined.43

In terms of constituency values, pro-secular CHP (Republican Peo-ple’s Party) voters’ conception of Turkish citizenship seems to beslightly more open to ethnic diversity. According to Konda (2011),the majority of CHP voters (56.7 percent) believed that it was not

Figure 3 The content of national identity in secular and Islamic press,1996 to 2004.

43. The difference between Yeni Safak and Zaman was statistically significantonly at the 95 percent interval.

Journal of Church and State

14

by guest on June 4, 2015http://jcs.oxfordjournals.org/

Dow

nloaded from

necessary to be ethnically Turkish to become a citizen of Turkey. Thecorresponding figure was slightly lower for AKP voters.

Are Islamists Less Nationalist in the Sense of Being More Welcoming toKurds and More Open to Kurdish Ethnic Politics?

An important barrier to finding a political solution for the Kurdishquestion has been the Turkish mainstream fear of ethnic politics as adivisive force and aversion to negotiating and cooperating withKurdish parties. Ethnic politics is a constitutional crime that can causethe abolition of political parties, although legal reforms during the2000s raised the standards of proof for doing so. Seven Kurdishparties were either shut down or dissolved themselves under threat ofbeing banned between the 1960s and 2008. The currently main pro-Kurdishparty (Peoples’DemocracyParty,HDP)wasestablished in2012.

Hence, ideational openness to ethnic politics may help a politicalactor to manage the Kurdish question. This is an area where Islamicelites appear to have an advantage over their secular counterparts.44

Figure 4 captures normative views on the right of ethnic parties toexist in Islamic and secular press.

Nationalism can also be measured indirectly via normative refer-ences to Kurds. The twenty-five-year-long war with the PKK can beexpected to have fed anti-Kurdish sentiments. However, most refer-ences to Kurds were found to have a neutral tone in all newspapers.The majority were positive in Yeni Safak in both periods and inZaman in the second period (2007 and 2010) (figure 5).

In terms of constituency values, anti-Kurdish feelings seem to bemore present among Islamic AKP’s voters than among secularCHP’s voters. According to Konda (2011), almost half of the AKPvoters (47.8 percent) stated that they would eschew Kurdish

Figure 4 The right to form ethnic parties in the secular and Islamic press,1996–2004.

44. Due to the relatively low number of observations, however, the differencebetween the two groups of newspapers was not found to be statisticallysignificant.

Is Islam the Solution?

15

by guest on June 4, 2015http://jcs.oxfordjournals.org/

Dow

nloaded from

neighbors, business partners, or spouses—even though presumablymoreKurdsareAKPvoters than CHP voters.45 AmongCHP voters, thepercentage was lower (43 percent).

Hypothesis 2: Do Islamic Actors Have More IdeationalPreparationforSpecificReformstoResolve theKurdishQuestion?

Regardless of their normative predispositions toward nationalismand ethnic politics, political actors must implement specific policies

Figure 5 References to Kurds in individual newspapers.

45. Demiralp, “The Odd Tango of the Islamic Right and Kurdish Left in Turkey,”287–302.

Journal of Church and State

16

by guest on June 4, 2015http://jcs.oxfordjournals.org/

Dow

nloaded from

to address Kurdish demands. Did Islamic actors discuss these spe-cific reforms more widely and more supportively than the seculardid?

Figure 6 demonstrates that, per newspaper and between 1996 and2004, the secular papers had thrice as many references to Kurdisheducation and broadcasting—two major demands of the Kurdishminority—as the Islamic papers. They also discussed the subjectmore approvingly, although the majority of the discussion in YeniSafak was also positive. The 2007 and 2010 research included moredetailedandseparatecodings forcultural rightssuchasKurdishedu-cation and political rights such as political-administrative autonomy.Figure 7 illustrates that, although religious papers seem to be slightlymore open to cultural rights, they had strikingly little discussion ofpolitical rights (N ¼ 5).

According to Konda (2011), the pro-secular CHP’s voters were moreaware of, and more sympathetic to, Kurdish ethnic-cultural demandsthan AKP voters were. A higher percentage (58.5 percent) of CHPvoters than AKP voters (52.1 percent) endorsed the idea of state

Figure 6 References on education and TV in Kurdish in the secular andIslamic press, 1996–2004.

Is Islam the Solution?

17

by guest on June 4, 2015http://jcs.oxfordjournals.org/

Dow

nloaded from

support forethnicgroups. Similarly, 44.6 and31.1percent of theCHPvoters thought that the Kurdish question resulted from restrictionson “Kurds’ ability to express their own identity” and from “state dis-crimination against Kurds,” respectively. By comparison, only 39.7and 24.8 percent of the AKP voters concurred.

Table 1 shows that reformist reports produced by or presented tosecularpartiesappear tohavebeen relativelymore frequentandelab-orate.However,mostof their reportswerepreparedduringthe1990s;during the 2000s, they became more conservative and nationalistbecause of their distrust of AKP-led reforms, among other reasons.46

Hypothesis 3: Are Islamic Actors Less Open toNon-Muslim and Perceived “Un-Islamic” Minorities?

As figures 8 and 9 show, compared with secular elites, Islamic elitesfeel more threatened by non-Muslim minorities and their public

Figure 7 Discussion of cultural and political rights, 2007 and 2010.

46. Sinan Ciddi, and Berk Esen, “Turkey’s Republican People’s Party: Politics ofOpposition under a Dominant Party System,” Turkish Studies 15, no. 3 (2014):419–41.

Journal of Church and State

18

by guest on June 4, 2015http://jcs.oxfordjournals.org/

Dow

nloaded from

activism—in terms of negative views and in the case of Christian mis-sionaries also greater attention—even if one leaves out the mostmilitant-Islamist newspaper, Milli Gazete. However, the other tworeligious newspapers also contain more positive general views onnon-Muslim minorities than the secular newspapers.

The 1996–2004 research found that, compared with secular elites,Islamic elites were considerably more suspicious of social pluralismin general, whether it resulted from religious, Muslim-sectarian,gender-based, or lifestyle differences.47 Figure 10 shows thatIslamic elites were more critical of social pluralism than the secular

Table 1 Secular and Islamist party reports on the Kurdish issue

Suggestion to solvethe Kurdish question

Secular politicalparties and actors

Islamic politicalparties and actors

Right to education in nativelanguage

SHP (1990, 1992)CHP (1996, 1999,2001)

RP (1991)AKP (2010)

Right to teach Kurdishlanguage

SHP (1990, 1992)CHP (1996, 1999)

RP (1991)Ozdemir (1993)AKP (2010)

Foundation of KurdishInstitute

SHP (1992)CHP (1996, 1999)

AKP (2010)

Kurdish Broadcasting SHP (1992) RP (1991)Kahveci (1992)Ozdemir (1993)AKP (2010)

Public names SHP (1992) AKP (2010)Right to give Kurdish namesto children

SHP (1992) —

Support for furtheradministrative (localadministration) reforms

SHP (1992)CHP (1996, 1999)

AKP (2010)

Support for participatorypluralist administration

SHP (1992)CHP (1996, 1999)

—

Constitutional recognition — —

The data in the table are acquired from the following reports: SHP SoutheastReport (1990), SHP Newroz Report (1992), CHP Tunceli Report (1996) , CHPEast and Southeast Report (1999), CHP Democratization and Human RightsReport (1999), Algan Hacaloglu Report (2000), CHP Democratization andHuman Rights Report (2001), ANAP Southeast Report (1993), ANAPDemocratization Declaration (1999), Kaya Toperi-Aslan Guner Report(1992), Adnan Kahveci Report (1992), Kemal Yamak Report (1993), HikmetOzdemir Report (1993), MCP East and Southeast Report (1991), RP KurdishReport (1991), RP Southeast Report (1994), AKP Democratic Opening Hand-book (2010). SeeYayman,Turkiye’ninKurtSorunuHafizasi [Turkey’sKurdishProblem Memory], for further references.

47. Somer, “Does It Take Democrats to Democratize?” 511–45.

Is Islam the Solution?

19

by guest on June 4, 2015http://jcs.oxfordjournals.org/

Dow

nloaded from

elites in 2007 and 2010, but critical views had decreased for bothtypes of press compared with 1996–2004.

These findings lend limited indirect support to the “Islamicethnic peace” hypothesis via the Islamic elites’ greater suspicion ofreligious, sectarian, and secular diversity (in line with hypothesis 3).But they also suggest that, even if they were successful in addressingethnic conflicts, Islamic identity-based models would reinforceproblems of nonethnic groups such as secular groups, women, andAlevis.

Figure 8 Normative views on Christian and Jewish minorities in Islamic andsecular press.

Figure 9 Normative views on Christian missionaries in Islamic and secularpress.

Journal of Church and State

20

by guest on June 4, 2015http://jcs.oxfordjournals.org/

Dow

nloaded from

If the Islamic Peace Thesis is Invalid, Why Do IslamistsSeem to Do More to Resolve the Kurdish Question?

Table 2 summarizes the particular reforms enacted by secular andIslamic political actors since the 1990s. Few reforms were madeduring the 1990s in the midst of a bloody war with the PKK. Butthe rather short-lived center-left and right coalition government(May 1999–November 2002) led by the secular Democratic LeftParty (DSP) initiated reforms.48 The coalition became reasonablyreformist once the political environment improved after PKK leaderOcalan was captured in February 1999 and the EU recognizedTurkey’s candidacy for membership in December 1999. This evalu-ation seems warranted, especially if one takes into account the frac-tured nature of the government, the disastrous financial crises in2000–01,and thestrengthof thenationalistmilitary,whichwassuspi-cious of Kurdish rights. Thus, structural conditions—such as single-party governments and strong electoral mandates, the EU anchor, rel-ative lack of violence with the PKK, and civilian government autonomyvis-a-vis the weakened military—rather than Islamic ideology mayexplain at least some of the reforms under the AKP.

During the AKP era, many reforms were facilitated by the EU acces-sion process.49 Resolving the Kurdish question was consideredcrucial to prove Turkey’s respect for minorities, a major conditionof the Copenhagen criteria for candidate countries.50 After 2005,

Figure 10 Values on social pluralism among Islamic and secular elites, 2007and 2010.

48. The DSP had been in power in a minority government since January 1999when Ocalan was captured.49. Ziya Onis, “Conservative Globalists versus Defensive Nationalists: PoliticalParties and Paradoxes of Europeanization in Turkey,” Journal of SouthernEurope and the Balkans Online 9, no. 3 (2009): 252.50. European Council, European Council in Copenhagen, June 21–22, 1993, Con-clusions of the Presidency. SN 180/1/93 REV 1; Frank Schimmelfennig, StefanEngert, and Heiko Knobel, “Costs, Commitment and Compliance: The Impact of

Is Islam the Solution?

21

by guest on June 4, 2015http://jcs.oxfordjournals.org/

Dow

nloaded from

Table 2 Reforms accomplished on Kurdish issue by secular and Islamist partiesa

Reform or measure to addressthe conflict

Secular parties Islamic parties

Lifting the ban on the use of Kurdishas forbidden language/nonofficiallanguage

ANAP (1991, Law no. 2932)DSP-MHP (Nationalist Action Party)-ANAP(Motherland Party) coalition 2001: 1st har-monization package, law no. 4709 articles9 and 10; 2002, 2nd harmonizationpackage: article 16/5 in press law no. 5680DSP-MHP-ANAP coalition, August 2002:3rdharmonizationpackage: constitutionaladjustments allowing broadcasting andKurdish in private courses

AKP (2002), extension of the Law on ForeignLanguage Education and Teaching to “Lan-guages and Dialects Used Traditionally byTurkish Citizens in their Daily Lives”2004, Legal changes allowing public broad-casting in radio and television2009 Launch of first TV channel in Kurdish2010 and 2013 right of electoral campaignsin Kurdish2011 Kurdish language and literature pro-grams at university level2012 Kurdish courses at fifth school grade2013 Kurdish teaching at private schools,limited defense in Kurdish at courts, allow-ance of letters q, w, and x on keyboardsb

Abolition of death penalty (convert-ing Ocalan’s—and potentially otherPKK members—sentence to life inprison)

August 2002: 3rd harmonization package

Lift of prohibition of Kurdish per-sonal names

AKP (2003)

Possibility for villages to revert theirnames to their original Kurdishnames

AKP (2013)

Limiting possibilities for the ban ofpolitical parties

DSP-MHP-ANAP coalition (2001, 1st har-monization package, law no.4709 onfreedom of thought and basic rights)

AKP (2010, constitutional change)

Jou

rnalo

fC

hu

rchan

dState

22

by guest on June 4, 2015 http://jcs.oxfordjournals.org/ Downloaded from

Cooperation with Kurdish politicalparties

SHP (Social Democratic People’s Party)-HEP(People’s Labor Party) electoral alliance(1991)c

Reducing the military’s powers inNational Security Council

DSP-MHP-ANAP coalition 2001, 1st har-monization package: law no. 4709 onfreedom of thought and basic rights:change of status

AKP (2003, 7th EU harmonization package:change of duties, meeting, and membershiprules; 2004, 9th EU harmonization package:change of article 118 of constitution)

Lifting state emergency rule (OHAL)in southeast Turkey

1994–July 2002: in nine provinces, variousgovernments; July 2002: two provinces,DSP-MHP-ANAP coalition

AKP (November 2002): Diyarbakır andSırnak provinces

Reform of the criminal code withconsideration for principles of uni-versal human rights.

AKP (2004)

Transfer of range of crimes fromState Security Courts to the RegionalHigh Criminal Courts

AKP (2004, article no. 5109)

Anti-Terror Law AKP (2004, amendment of Anti-Terror LawNo. 3713)

Public discourse on Kurdish identityand question

Demirel/SHP-DYP coalition (1991) AKP (2009, parliamentary and public dis-cussion on the Kurdish problem)

Abolishment of the “Oath ofAllegiance” of primary students tothe Turkish nation

AKP (2013)

aSee Gozde Yilmaz and Didem Soyaltin, “Zooming into the ‘Domestic’ in Europeanization: Promotion of Fight against Corruptionand Minority Rights in Turkey,” Journal of Balkan and Near Eastern Studies 16, no. 1 (2014); 11–29; Teczur, “The Ebb and Flow ofArmed Conflict in Turkey,” 171–88.bQ, w, and x are not part of the Turkish alphabet, but they are used in Kurdish.cMurat Somer, “Why Aren’t Kurds Like the Scots and the Turks Like the Brits? Moderation and Democracy in the Kurdish Ques-tion,” Cooperation and Conflict 43, no. 2 (2008): 220–49; Gulistan Gurbey, “The Development of the Kurdish NationalismMovement inTurkey,” inRobertW.Olson,TheKurdishNationalistMovement in the1990s: Its ImpactonTurkeyand theMiddleEast(Lexington: University Press of Kentucky, 1996), 26; Kirisci and Winrow, The Kurdish Question and Turkey: An Example of aTrans-State Ethnic Conflict.

IsIslam

the

Solu

tion

?

23

by guest on June 4, 2015 http://jcs.oxfordjournals.org/ Downloaded from

when accession negotiations formally began, the credibility of the EUweakeneddue tovarious factorsonbothsides.51 But strategic consid-erations help toexplainwhytheAKPcontinued the reform path.First,fulfilling theEUdemandssignaled to theAKP’scriticsthat the Islamicgovernment would not interrupt Turkey’s Westernization, helping tosafeguard the party’s position vis-a-vis Western allies and investors,and domestically, vis-a-vis skeptical, secularist elites in the adminis-tration, military, media, and business, who still dominated the polit-ical systemthen.52 Second, theelectoral successof theAKPwasbasedon maintaining its anti–status quo reputation.53

But as the EU’s credibility weakened, the AKP consolidated itspower, and violence resurged between the PKK and the state,54 atten-tion to EU reforms gradually waned. As table 2 indicates, even thoughthe government’s generally supportive attitude continued towardethnic minority rights, few programmatic, major advancementswere made targeting the Kurdish question after 2005. Instead, thefocus has been “selective,”55 targeting the electoral support for thegovernment among Kurdish voters, while leaving out the more con-troversial political and constitutional reforms. Electoral shiftsbetween Islamist and pro-Kurdish parties are possible because,amongother reasons,bothappeal toKurdishvotersaspartiesoppos-ing urban elitism in Turkish politics.56 Both the 2009 and 2013 open-ings were motivated to considerable extent by domestic and externalstrategic interests and expediencies, and the AKP’s framing of theopenings oscillated between inclusive narratives of peace and pros-perity for all and exclusive narratives vilifying major Kurdishgroups and opposition parties.57

EU Democratic Conditionality on Latvia, Slovakia and Turkey,” Journal ofCommon Market Studies 41, no. 3 (2003): 496.51. Yilmaz and Soyaltin, “Zooming into the ‘Domestic’ in Europeanization,”11–29; Onis, “Conservative Globalists versus Defensive Nationalists,” 253, 257;Neill Nugent, “The EU’s Response to Turkey’s Membership Application: Not Justa Weighing of Costs and Benefits,” Journal of European Integration 29, no. 4(2007): 481–502.52. Onis, “Conservative Globalists versus Defensive Nationalists,” 252–53.53. Ibid., 253; Yilmaz and Soyaltin, “Zooming into the ‘Domestic’ inEuropeanization.”54. Onis, “Conservative Globalists versus Defensive Nationalists,” 254, 256;Yildiz, EU Conditionality and Minority Rights: A Comparative Study on Romaniaand Turkey (Master Thesis: Uppsala University, 2011), 44.55. Yilmaz and Soyaltin, “Zooming into the ‘Domestic’ in Europeanization,”11–29.56. Demiralp, “The Odd Tango of the Islamic Right and Kurdish Left in Turkey,”287–302.57. Johanna Nykanen, “Identity, Narrative and Frames: Assessing Turkey’sKurdish Initiatives,” Insight Turkey 15, no. 2 (2013): 85–101.

Journal of Church and State

24

by guest on June 4, 2015http://jcs.oxfordjournals.org/

Dow

nloaded from

Public support for EU membership declined among both support-ers and skeptics of the AKP government.58 Findings of the contentanalysis from 2007 and 2010 are consistent with these observations.Figure 11 shows that positive evaluations of the EU decreased in bothtypes of the press, although they were slightly more frequent in theIslamic press. But negative evaluations of the EU almost doubled inthe Islamic papers, while they dropped significantly in the secularpapers.

These findings suggest that political opportunities and strategicconsiderations contributed to the support for the EU and for theEU-linked reforms among Islamic actors.

Figure 11 EU support in secular and Islamic newspapers, 2007 and 2010.

58. Support for EU membership and trust in the government in Turkey (Eurobar-ometer surveys 2004 to 2011). On the interaction of partisanship, religiosity, andEU support, also see Ali Çarkoglu, “Who Wants Full Membership? Characteristicsof Turkish Public Support for EU Membership,” Turkish Studies 4, no. 1 (2003):171–94.

Is Islam the Solution?

25

by guest on June 4, 2015http://jcs.oxfordjournals.org/

Dow

nloaded from

Conclusions

As ethnic and sectarian wars involve Islamist contenders for powerfrom Yemen to Libya to Nigeria and international military campaignsarewagedagainst theradical Islamist ISIS inSyriaand Iraq, examiningthe relationshipbetween Islamism andethnic divisions isastimelyasever. In Turkey, the PKK leader Ocalan stressed that Turks and Kurdshad lived side by side under the “banner of Islam” for thousand yearsbefore the proliferation of “mini nation states” among Turkish,Kurdish, and Arabic peoples under Western imperialism.59 Then–Prime Minister Erdogan highlighted Turkey’s Islamic-Ottoman heri-tage and the Muslim brotherhood of Turks and Kurds. Thus, bothleaders were to differing degrees invoking Islamic values as a basisfor peace.

These statements echo expectations prevailing among many schol-ars as well as Muslim social-political actors. But the Islamic peacethesis is rarely put to the test of dispassionate theoretical and empir-ical analysis. The evidence we offer in this article indicates that theIslamic AKP’s activism in addressing the Kurdish question is insuffi-cient to support the Islamic peace thesis. While relatively more opento ethnic politics and the public expression of ethnicity than theirsecular counterparts, Turkish Islamic actors are not necessarilylessnationalist.Theyhave less ideationalpreparationforspecificpol-icies and reforms to address ethnic grievances. They are also consid-erably more suspicious of social pluralism in general, sectariandifferences, and non-Muslim and “un-Islamic” diversity. These find-ings are consistent with earlier research indicating that Islamicelites—who embrace many creeds of democracy—remain highly crit-ical of secular democracy. They prefer more religious involvement inboth social and state affairs and often oppose the pursuit of fullequality for various groups such as women and Muslim minorities.60

This may also undermine their ability to manage ethnic diversitybecause, many in Western democracies maysee such social and polit-ical norms about equality as the normative basis of the political-institutional frameworks within which modern polities addressethnic conflicts more or less successfully.

Thus, rather than ideological and ideational superiority, structuralfactors such as the reduction of the military tutelage and electoralstrategies may better explain the AKP’s Kurdish peace initiatives.61

59. Ocalan’s statement reflected an attempt to appeal to both religious andsecular sensitivities as he also announced that he embraced Western values ofenlightenment, equality, freedom, and democracy.60. Somer, “Does It Take Democrats to Democratize,” 511–45.61. Somer and Liaras, “Turkey’s New Kurdish Opening: Religious and SecularValues,” 152–65.

Journal of Church and State

26

by guest on June 4, 2015http://jcs.oxfordjournals.org/

Dow

nloaded from

Similarly, the root cause of the Kurdish question may not necessarilybe republicanism, secularism, or Islamism per se. While rejectingsecular nationalism and the idea that loyalty to nation should super-sede all other loyalties, Islamism does not necessarily eliminatehierarchy between different identities and loyalties. Arguably, itmerely substitutes the supremacy of Islam—or nationhood basedin Islam—for the supremacy of secular nation. Hence, according toone sympathetic writer, there would be “some constraints on thefreedom of ethnic minorities in an Islamic state.” Their freedomsshould not “contradict with the basic principles of religion,” mustserve“public interest,”and“ethnicitycannotbeusedforethniccompe-tition that potentially causes fragmentation and discrimination.”62

These constraints are remarkably similar to the constraints imposedby secular nationalists in the name of national unity, harmony, andinterest. Simple replacement of secular nation with religion does notprovide the principles and institutions that would prevent majoritar-ian practices and successfully accommodate minority interests.

We are not arguing that Islamic actors cannot resolve the Kurdishquestion or other ethnic-regional problems. We are merely maintain-ing that the ability to successfully manage ethnic diversity dependson structural opportunities, normative and ideational preparationfor concrete solutions, and political ability and willingness to imple-ment theserecipes. Islamicaswell assecularactorscanresolveethnicconflicts if they meet these conditions.

Hence, the key steps, which societies should take in order to suc-cessfullyaddress ethnic minority problems, involve choices regardingpolitics and institutions, not necessarily choices between religion andsecularism and between religious and secular ideologies. Neithersecularnorreligious identitiesautomaticallyeliminateethnic-regionalconflicts of interest. Rather than secular or religious ideologies perse, the root causes of ethnic-regional conflicts lie in majoritarian andauthoritarian political decisions, institutions and values. They sacri-fice minority interests for majority interests. And even with politicalwill tocherish difference, there isoften a lackof ideas and institutionalsolutions to effectively accommodate ethnic-regional differences.

62. Ataman, “Islamic Perspective on Ethnicity and Nationalism,” 89–102.

Is Islam the Solution?

27

by guest on June 4, 2015http://jcs.oxfordjournals.org/

Dow

nloaded from