Uncultured: Civil War and Cultural Policy

Abstract

How does cultural policy affect violence? While cultural discrimination is frequently cited as a potential grievance motivating political violence, the relationship remains under-theorised and largely untested. I weave theoretical literatures with interviews and secondary sources on the experience of Kurds in Turkey to understand the socio-economic and psychological pathways through which cultural policies impact intrastate conflict. I then analyse cross-national data on political violence, demonstrating that cultural grievances increase support for violence, raise the chance and severity of conflict and prolong violent conflicts. In short, policy matters: cultural restrictions exacerbate violence through multiple pathways.

Acknowledgements

Phil Powell provided valuable research assistance. I also deeply appreciate the generosity of my interviewees and hosts and my translators’ and other contacts’ assistance, but I have omitted their names due to the current political situation in Turkey. Comments from Amita Shastri, Jacques Bertrand and anonymous reviewers have substantially improved this article; any errors are my own.

Notes

1. See Marquardt (2018) for one exception.

2. See Fox (e.g., 1997, 2002) on religious policy, Brown and Ganguly (2003) on language or Kirschner (2015) on the mechanisms linking discrimination and violence for exceptions.

3. Numerous groups peacefully press for Kurdish rights; the PKK, founded in 1974, is the dominant militant organisation. This partly results from its success in internecine fighting. Despite complex attitudes towards the PKK, it continues to be the single most important actor representing Kurds, and even those who disagree with its goals or strategy recognise that it cannot be disregarded.

4. All interview data below are anonymised due to the sensitivity of the topic. Some interviewees were randomly selected by approaching individuals in tea-houses; others agreed to speak because of my own or my translators’ contacts. Some interviews were formal, structured conversations and others became group discussions as they proceeded, with friends, colleagues and family members of the interviewees entering and leaving the conversation. All interviews were conducted in the interviewee’s choice of language: Kurdish (Kurmanji dialect), Turkish or English.

5. See Coşkun et al. (2011) or McDowall (2007) for detailed histories.

6. The legislation and military policies coincide with striking underdevelopment in the south-east. For example, in 2009, national income per capita was $8,200; in Diyarbakır Province, it was below $1,500, and unemployment was as high as 20 per cent (International Crisis Group 2012b, Karakoç and Sarıgil 2013). Infant mortality among Kurds is double the rate for Turks (Koc et al. 2008).

7. See Uçarlar (2009) for more on Turkish language policy and responses by Kurdish groups.

8. Turkey lifted the ban on Q, W and X in late 2013, although the letters are still not formally recognised. The letters Î, Ê and Û remain banned.

9. Coşkun et al. (2011) recorded similar observations by both students and teachers.

10. Skutnabb-Kangas (2005) asserts that assimilationist subtractive education, such as Turkey practices, is a form of genocide.

11. One of the most important advances has been the shift from referring to Kurdish as ‘unknown sounds’ in court records to recording it as an ‘unknown language’ or ‘a language that is supposed to be called Kurdish’. Obviously, this symbolic shift is of little practical use for defendants. Recently, the government proposed allowing Kurdish language defences at specific points in the trial process, but this can be denied by individual judges (Derince 2013).

12. See Robichaud and De Schutter (2012) for more on this concept.

13. Because Kurdish has been so denigrated, many Kurds are deeply appreciative of even basic consideration for their language and culture. For example, several interviewees expressed surprise and gratitude for what they deemed the respect of my distributing business cards with a Kurdish translation of my contact information on the back.

14. Indeed, Robichaud and De Schutter (2012) assert that using one’s native language in itself promotes dignity and self-esteem. Thus, when language and culture are explicitly denigrated, this effect should be magnified many times over.

15. At the extreme, Zeydanlıoğlu (2009) asserts that required recitation of Turkish nationalist slogans and renunciation of Kurdish identity was part of a broader campaign of torture in the notorious Diyarbakır prison in the 1980s.

16. Turkish language facility also tends to be gendered, in part because women attend fewer years of school, but also because men are more likely to learn Turkish at their jobs or during mandatory military service (Öpengin 2012, Çağlayan 2014).

17. See also International Crisis Group (2011) and Karakoç and Sarıgil (2013).

18. The primary limitation is due to missing data, particularly from the earlier conflicts in the sample. While this could introduce some bias, the missing cases (and missing data from certain years within other cases) are randomly distributed geographically, and the observations have little in common beyond the dates during which they occurred.

19. The full scale is: no conflict; political banditry or sporadic terrorism; campaigns of terrorism; local rebellion; small-scale guerrilla activity; intermediate guerrilla activity; large-scale guerrilla activity; and civil war. Coding rules are in the MAR codebook (2009).

20. These data are from Kirschner (2015).

21. Due to substantial amounts of missing data for some of the individual metrics, I use average grievances. While I would ideally explore the effect of concerns surrounding each component, such as religion or language, separately, currently available data are simply not fine-grained enough to analyse these differences.

22. Many of the concepts can be operationalised in multiple ways. For example, to capture political discrimination, I tested measures of legal status and civil and judicial rights. The results are generally robust to different operationalisations.

23. While these scales differ somewhat, I ultimately used as much of the EPR data as possible to highlight the relationship to other recent research on grievances and political violence. The trade-off is a smaller data-set because of the difference in scope: some MAR variables are coded from 1940, others from 1985 and still others only in the mid-2000s, while the EPR data begin in 1946 but have gaps on some variables and groups, particularly in the early years. Despite these challenges, the analyses still include several hundred observations, mostly covering conflict years from the late 1980s on. The results are stronger using MAR data across measures. While this could suggest that more direct discrimination, such as MAR captures, is more influential, the low overlap between the two data-sets’ coverage over time makes it hard to draw substantive conclusions from the difference. Forms of discrimination are often related: political power makes cultural repression and economic dominance more likely, economic control facilitates political dominance and cultural policies have socioeconomic effects. But while these dimensions of exclusion co-occur in many cases, it is far from certain. The correlation between the policies ranges from 0.354 for cultural and political rights to 0.289 for political and economic policies and 0.179 for economic and cultural rights. Low VIF scores confirm that multicollinearity does not significantly affect the results. It would be particularly interesting to explore further how the mechanisms may vary between the different forms of restrictions; for instance, political and economic limits can influence group capabilities and organisational capacity in ways that cultural restrictions are unlikely to impact.

24. See Daly (2012) on how organisational legacies influence violence. These variables are all from EPR.

25. Data from Fearon and Laitin (2003) and Kirschner (2015), with the exception of the autonomy measure, which is from the EPR data.