By Avital Livny, University of Illinois, Urbana-Champaign
*This memo was prepared for presentation at the Contemporary Turkish Politics Workshop at Rice University’s Baker Institute on October 14, 2016
The success of the pro-Kurdish Peoples’ Democratic Party (Halkların Demokratik Partisi or HDP) in Turkey’s June 2015 general election came as a considerable surprise, even among party members. Although some pre-election polls in late May had the party crossing Turkey’s steep 10 percent electoral threshold, none expected the party to do as well as it did, winning over 13 percent of the national vote and securing 80 parliamentary seats. And even with violent government crackdown in many party strongholds in the lead-up to the early elections called for November of that same year, the HDP once again managed to overcome the threshold, winning 10.8 percent of the national vote and 59 seats. This string of successes represented a major change in the electoral performance of pro-Kurdish parties that, at their best, had captured no more than 5 or 6 percent of the national vote in previous election years.
If we view this as a general trend, a number of questions naturally follow: Does the rising success of pro-Kurdish parties indicate that the size of the Kurdish population in Turkey has itself increased? Or have Kurdish Turks becoming increasingly politicized and therefore more willing to support a pro-Kurdish party? Alternatively, has the liberal, left-wing political agenda of the party been increasingly able to win the support of more ethnic Turks? Tracing such a demographic shift or change in the relationship between ethnicity and political preferences would be relatively straightforward in many contexts where official census data or nationally representative surveys are available. But in Turkey, information about ethnic difference is so tightly controlled by the government that even international surveyors are prohibited from asking about them, to say nothing of official statistics on ethnicity which have not been published since 1965.
To fill the information gap, I use nationally representative survey data on ethnic self-identification and political preferences to assess patterns at the individual-level and in the aggregate, estimating the relative size of the Kurdish population, across space and across time. Though I find no recent increase that could account for the success of Kurdish party politics, I find a steady increase in the percent of Kurdish voters that express support for pro-Kurdish parties, alongside a decline in support for these parties among members of the ethnic Turkish majority. Altogether, the share of Kurdish voters supporting pro-Kurdish parties has increased from only 3 percent of the total electorate in March 2010 to 10.8 percent in September 2015, enough to just pass the threshold without the support of other voters. Preliminary evidence indicates that the largest increases in support came from those Kurds living in districts (ilçeler) with larger Kurdish populations, indicating targeted mobilization by party activists.
1 Measuring Ethnicity in Contemporary Turkey
Despite their considerable political, economic, and historical relevance, ethnic differences have largely been ignored by the Turkish state. While the Ottoman Empire was known for centuries as a place of pluralism, coexistence turned to conflict as the empire began to collapse, and many minority communities emigrated, whether by choice or by force, in the early decades of the twentieth century. When the Turkish Republic emerged from the ashes of the collapsed empire, the mood changed yet again – a singular focus on the Turkish-Sunni nation-state meant a denial of ethnic and religious differences in favor of an illusion of homogeneity, resulting in a paucity of data on identity. Although early Republican censuses continued the Ottoman tradition of enumerating mother tongue (a proxy for ethnicity) and religious denomination, there is some evidence of official manipulation of these results and, starting in 1965, they ceased being published altogether. In the absence of official statistics, scholars of ethnicity in contemporary Turkey have been left to rely on demographic projections from fifty-year-old censuses and data on mother tongue from the Demographic and Health Survey, both of which may introduce some bias into their estimations of group size.[1]
As a corrective to these existing limitations, I look at responses to a series of 54 nationally representative face-to-face surveys conducted by KONDA Research and Consultancy between 2010 and 2015. Altogether, survey data exist for just under 150,000 respondents from nearly 6000 neighborhoods and villages in 490 districts (ilçeler) in 66 provinces across all 12 of Turkey’s geographical regions. Consistently across all surveys, respondents are asked a sensitively-worded subjective question about their ethnic identity: “We are all citizens of the Republic of Turkey, but we may have different ethnic roots. As far as you know or feel, what is your identity?”[2]In Turkish: “Hepimiz Türkiye Cumhuriyeti vatandaşıyız, ama değişik etnik kökenlerden olabiliriz; Siz kendinizi, kimliğinizi ne olarak biliyorsunuz veya hissediyorsunuz?” Answers are transcribed verbatim and later coded as either Turkish, Kurdish, Zaza, Arab, or Other. In addition, respondents are asked how they would vote if a general election were held tomorrow, with answers ranging from one of the major political parties to a minor party or independent candidate to being undecided or being uninterested in voting at all.
Across all respondents, I find that 13.1 percent self-identify as Kurdish, with considerable sub-national variation (Figure 1). (Although some provinces (iller) are not covered in the survey data and, within them, some districts (ilçeler), there is no systematic pattern to this missingness, so that the data can be taken as broadly representative.) Large Kurdish populations have been correctly identified in much of southeastern Turkey, as well as parts of the northeast, while urban centers in western and central Turkey are also identified as having relatively large Kurdish populations. When I disaggregate the national average across time, I find no indication of an increase in the size of Turkey’s Kurdish population over the past five years that might explain the parallel rise in the success of pro-Kurdish parties (Figure 2). If anything, the only group that appears to have grown over time is the catch-all “Other” category, which has seen a slight increase at the expense of self-identified ethnic Turks.[3]
Disaggregation of this catch-all category indicates that the increase is not in self-identified Arabs or Zazas but in those who fall into far smaller categories or who eschew the use of ethnic categories altogether.
2 The Rise of Kurdish Political Identity in Turkey
If the relative size of the Kurdish population has not increased in recent years, how are we to explain the rising success of pro-Kurdish political parties? One concern is response bias that is masking a real change in underlying demographics. If that were the case, stated support for pro-Kurdish parties would likely also be under-reported. But unlike the relative share of the Kurdish population, levels of self-reported support for the pro-Kurdish Peace and Democracy Party (Barış ve Demokrasi Partisi or BDP) and the HDP have steadily increased over time, paralleling the party’s rising success in local and general elections (Figure 3). While only 2.5 percent said that they would support the BDP in March 2010, 9.9 percent indicated their support for its successor, the HDP, in September 2015. This is a considerable increase, especially given the large percentage of undecided respondents (16.9 percent) and those who say they would not vote (almost 4 percent).
An increase in stated support for pro-Kurdish parties should alleviate some concerns about response bias, but it does not address our motivating question: what explains the rising success of Kurdish-based politics? An examination of variation in support for Kurdish parties across ethnic categories over time indicates the surge in support has come almost entirely from voters who identify as Kurdish (Figure 4). While support for the BDP and HDP among Turkish voters has remained consistently small over time – hovering at or under 1 percent – support for the pro-Kurdish parties among Kurdish voters has steadily increased, from under 35.8 percent in March 2010 to nearly 68.0 percent in September 2015. (There is also some indication that support for pro-Kurdish parties is growing among voters who identify as ethnically “Other” – from 0 percent in the earliest surveys to almost 13 percent in 2015 – but the small size of this group – just 5 percent of the Turkish population – limits the extent to which this increase could actually influence the outcome of a general election.) Even without an increase in the relative size of the Kurdish population, this shift in party support has meant that Kurdish voter support for Kurdish parties doubled from just 4.4 percent of all decided respondents in 2010 to 8.8 percent in 2015.
So what explains this increase? One possibility is that a demographic shift within the Kurdish population has increased the extent to which Kurdish identity is politicized. Across all respondents, I find greater support for pro-Kurdish parties among younger male voters who are relatively undereducated. If the Kurdish population has become increasingly younger, increasingly less educated or if there has been a shift in its gender balance, demographics might be able to explain the change over time. I account for this possibility by running multivariate regressions of pro-Kurdish party support including basic demographic characteristics – age, gender, and education – and bivariate indicators of Kurdish and Turkish ethnicity for each year in the dataset. A plot of the ethnicity coefficients over time (Figure 5) reveals that the impact of Turkish ethnicity on pro-Kurdish party support rose and then fell, while the magnitude of the Kurdish coefficient has almost monotonically increased, even after controlling for demographic differences.
To understand what factors have become increasingly important in determining Kurdish voters’ support for pro-Kurdish parties, I ran a number of regression models of party preferences among survey respondents who self-identify as Kurdish. In each, I included interaction between a given variable – individual demographics, including age, gender and education, as well as geographic ones, including the estimated size of the Kurdish population in the respondent’s home district – and bivariate indicators of each survey year after 2010. Most remarkable, perhaps, is what I did not find: the impact of individual-level demographic characteristics on Kurdish voters’ preference for pro-Kurdish parties has remained entirely stable over time. Age, gender and education impact their preference for pro-Kurdish parties in the same way in 2015 as they did in every previous year. The only significant pattern is the importance of local demographics to individual vote-choices: among Kurdish voters, the relative size of the Kurdish population in their home district boosts their level of support for pro-Kurdish parties in all years, but the magnitude of this positive effect increases over time, especially after 2011 (Figure 6).
This interaction effect largely explains the increase in pro-Kurdish party support over time, although there remains a positive intercept shift in 2015. Further, these interaction effects are robust to the inclusion of fixed effects at the region- (bölge) and province-level, indicating that this is more about local environment than geographic variation. Still the reasons for the increasing importance of local demographics remains unclear: it is possible that HDP officials are targeting their mobilization efforts in the most heavily Kurdish areas; alternatively, it may be that Kurdish voters most open to pro-Kurdish politics are choosing to live in areas where they constitute a large share of the population. Using information about survey respondents’ birth province, as well as their father’s birth province, I am able to test for this type of a sorting mechanism. Controlling for respondents’ current home province, I find that internal migration has a direct positive effect on Kurdish support for pro-Kurdish parties but that this effect has decreased over time, evidence against a sorting mechanism.
Conclusion
Using new data on ethnic identification and political preferences in Turkey, I am able to confirm that the recent rise in pro-Kurdish politics is the result not of an underlying demographic shift nor of the increasing support of ethnic Turks. Instead, it appears that the ethnic identity of Kurdish voters in Turkey has become increasingly politicized in recent years, especially in areas with larger Kurdish populations. It is clear, therefore, that ethnic demography is becoming increasingly important in Turkish politics, a trend that warrants more attention from country specialists.
[1] While the former makes a number of assumptions about population growth rates and migratory patterns over a fifty-year period, the latter – like the official censuses – assumes that mother tongue is the same as ethnic identity, although it is, at best, only an imperfect proxy.
[2] In Turkish: “Hepimiz Türkiye Cumhuriyeti vatandaşıyız, ama değişik etnik kökenlerden olabiliriz; Siz kendinizi, kimliğinizi ne olarak biliyorsunuz veya hissediyorsunuz?”
[3] Disaggregation of this catch-all category indicates that the increase is not in self-identified Arabs or Zazas but in those who fall into far smaller categories or who eschew the use of ethnic categories altogether.