Uneven development and insurgency in Turkey: a computational approach

Thesis

This thesis develops computational techniques to gather, process, and analyze fine- grained data on the war between the Turkish State and the Partiya Karkeren Kurdistan (PKK), a Kurdish insurgent group. In three chapters, I seek to better understand the reasons that lead people join the PKK, assess the impact of development policies aimed at dissuading them from doing so, and explain the group’s structural resilience to military force. The first chapter explores the recruitment of young urban recruits using web scraping, fuzzy matching, and other computational techniques. Leveraging an unprecedentedly detailed research design, militants are compared to random samples of Turkish citizens. I find evidence linking insurgent recruitment and a range factors including birth order and family size, peer effects, and conflict-induced migration. The second chapter explores economic motivations in Turkey’s agrarian Southeast using remote sensing and spatial econometrics. I find that clashes are more frequent following poor harvests, irrigation decouples agricultural income from rainfall, and that conflict was reduced in areas benefiting from a state-sponsored agricultural development program. The third chapter examines the PKK as a whole, focusing on its structural characteristics. I develop a new methodology that leverages deep learning to create a social network graph based on co-appearance in photographs which retains many of the broad structural features of the PKK. Analytical results indicate that the densely interconnected nature of the PKK makes it highly robust to a range of counterinsurgency strategies. Together, substantive findings suggests that development policy is a far more promising avenue for the resolution of the conflict than military policy, while the methodological contributions include the development of forward looking analytical techniques and open source software that enable highly detailed quantitative analysis of civil conflict.

Authors: Ballinger, O

• DOI: 10.5287/ora-nzqvrpeqo
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: deep learning; network science; econometrics
• Subjects: International Development; Political Economy; Conflict