Hearts, Minds, and Pockets: the Political Economy of Civilian Support for Kurdish Insurgency in Turkey
Total Page:16
File Type:pdf, Size:1020Kb
Hearts, Minds, and Pockets: The Political Economy of Civilian Support for Kurdish Insurgency in Turkey Tolga Sinmazdemir∗ March 28, 2016 ∗Assistant Professor, Department of Political Science and International Relations, Bogazici University, Istan- bul; Research Officer, Department of Government, University of Essex Abstract Does poverty make ethnic insurgencies more likely? I argue that if (a) civilian support for insurgents plays a decisive role in their success, and if (b) civilians experience a reduction in public goods provision after supporting insurgents, then insurgent attacks are less likely to occur in poorer regions. This is because lower income makes poorer individuals more vul- nerable to losing public goods, and therefore makes them less willing to support insurgents. I test this argument using data on attacks by the Kurdish insurgency between 1988 and 1999 in 20 Turkish provinces that are poorer than the national average, and in which Kurds form a significant share of the population. I find that in provinces with initially high levels of public goods provision, attacks are less likely when per-capita income of the province is a smaller percentage of the national average. Moreover, higher unemployment is associated with fewer attacks. Keywords: income inequality, unemployment, public goods provision, insurgency, ethnic conflict Word Count: 9971 Do economic grievances play a role in insurgency outbreaks? While there is an ongo- ing scholarly debate on this question,1 recent counterinsurgency efforts presume that economic grievances do play a critical role in insurgent violence, and economic development assistance has been used extensively as a tool for combating insurgencies in countries such as Iraq and Afghanistan (Beath et al. 2011). The logic underlying these efforts is straightforward: Civilian support is critical for successful insurgency, and economically better-off civilians are less suscep- tible to the appeal of armed groups. Therefore, economic development should make insurgent violence less likely.2 The literature on the political economy of insurgencies does not provide a clear empirical support for this logic: On the one hand, recent studies document that public goods provision lowers violence in countries involved in civil wars and insurgency (Beath et al. 2011; Berman et al. 2011b), and individuals with lower income are more likely to become involved in violence (Humphreys and Weinstein 2008; Justino 2009). On the other hand, there are subnational studies that find no relationship between unemployment and insurgent attacks (Berman et al. 2011a), and between poverty and support for political violence (Blair et al. 2012). In this paper, I contribute to this literature, and propose an explanation of insurgent violence that links together the effects of income and of public goods provision on civilians' willingness to assist militants. My explanation builds on an additional body of research that finds that populations associated with militants suffer negative economic consequences in the aftermath of violence (Abadie and Gardeazabal 2003; Benmelech et al. 2010). Following these findings in the literature, I hypothesize that if violence leads to a reduction in public goods provision, then the effect of this reduction on civilian support for insurgents depends on the initial income of the local population, and on the initial level of public goods provision by the government. Specifically, I argue that provision of public goods prior to the outbreak of violence makes poorer populations less likely to assist militants because the prospect of losing public goods, partially or even entirely, increases their opportunity cost for involvement in an insurgency. Since assistance from non-combatants, for example in the form of material support, shelter, or information, is vital for the insurgents' ability to fight (Justino 1See e.g. Cederman et al. (2011) for a recent review. 2See e.g. U.S. Army's Counterinsurgency Field Manual (2007). See also Blair et al. (2012) for other examples of the same logic motivating economic assistance from U.S., U.K. and international organizations to Pakistan. 3 2009; Kalyvas 2007), we should observe less violence where the local population is poorer, and where violence can lead to a reduction in public goods provision. To test this argument, I collect province-level data on per capita income, unemployment, public goods provision, and attacks by the Kurdish insurgency (Partiya Karkeren Kurdistan (PKK) which in Kurdish means Kurdistan Workers' Party) between 1988 and 1999 in 20 provinces of Turkey. PKK was founded in late 1970s as a Marxist-Leninist group, aiming at a separate state for the minority Kurds living in Turkey, in response to several assimilationist measures by the Turkish state such as bans on the use of Kurdish language in schools, in publi- cations and in names of places that are predominantly Kurdish.3 The only official language of the country has been (and still is) Turkish. The first major attacks of PKK took place in 1984. Following the capture of its leader Abdullah Ocalan in 1999, the demands of the insurgents have changed into political and cultural autonomy for Kurds living in the southeastern parts of Turkey (Kapmaz 2011). Studies of insurgencies have demonstrated that an active and prolonged insurgency requires public support (Kalyvas 2006; Petersen 2001; Weinstein 2007). I interpret the presence of PKK attacks as an indicator of civilian support for insurgents in areas where these attacks took place. I follow this interpretation for four reasons: First, as I show in my empirical analysis, insurgent attacks are more likely in provinces with Kurdish majorities, and support for Kurdish insurgents should be higher in these Kurdish majority provinces relative to ethnically more mixed provinces. Second, my empirical analysis shows that attacks are more likely in provinces with lower turnout in elections. Since turnout can be seen as an indicator of support for the government and the Turkish political system, this negative relationship between attacks and turnout suggests that attacks are more likely to happen in provinces with lower support for the government, and hence higher support for the insurgents. Third, and perhaps most importantly, alternative interpretations of insurgent attacks (e.g. attacks proxying for support for the government) are not consistent with patterns in the data. More specifically, my analysis shows that insurgent attacks are more likely in provinces with lower public goods provision. If attacks are a proxy for support for the government, this pattern implies that lower public goods 3For a history of the Kurds, see McDowall (1996) and Van Bruinessen (1992). For in-depth analyses of the Kurdish question in Turkey, see Barkey and Fuller (1998); Kirisci and Winrow (1997); Van Bruinessen (2000). For a study of PKK insurgency, see Marcus (2007). 4 provision is associated with higher support for the government. This is a highly implausible scenario given the fact that these provinces are poor and benefit from public goods extensively. Fourth, prior research on PKK violence supports my interpretation of attacks as a proxy for support for the insurgents. For instance, using election data from 1999, 2004 and 2007, Gergin (2012) finds that more popular support for the insurgents (measured as the vote share of the political party affiliated with PKK) is associated with more insurgent attacks. My econometric analysis of insurgent attacks supports the hypothesis that the prospect of adverse economic consequences due to insurgent violence affects civilian collaboration with the militants, and the ultimate level of insurgent attacks, conditional on the initial level of public goods provision and income of the local residents: In provinces with high levels of investment in public goods, insurgent attacks are less likely where the income per capita is lower. However, when investment in public goods is low, insurgent attacks are more likely in provinces where income per capita is lower. Moreover, in provinces with high levels of investment, an increase in unemployment is associated with a lower rate of attacks. My argument hinges on the premise that insurgent violence leads to a subsequent decline in public goods provision in the affected areas. To demonstrate that this premise is empirically supported, I also provide qualitative and quantitative evidence on the adverse effects of insurgent violence on different types of public goods provision such as health services and education in these provinces. In what follows, first I review the literature that my paper relates to and highlight the contribution of this paper. Then, I outline the logic of my hypotheses in greater detail and also provide qualitative and quantitative evidence that supports the logic of my hypotheses. Having specified my hypotheses, I describe the data collected for the empirical tests, and present regression results of the spatial and temporal distribution of civilian killings by PKK both at the province and the lower district level. I conclude with a summary and discussion of the implications of my results. All tables and figures are in the Appendix. Poverty, Public Goods Provision, and Conflict Motivated by the robust positive correlation between low per-capita income and civil war risk at the cross-national level (Fearon and Laitin 2003; Hegre and Sambanis 2006), there is a growing 5 body of work that uses sub-national data to understand further the links between local economic conditions and the likelihood