Contestation, Governance, and the Production of Violence: Evidence from Colombia

• In Event: Criminal Governance in South America

Abstract

During civil war, which configurations of territorial control exhibit more or less violence against civilians? This is a deceptively tricky descriptive question, because armed group territorial control is so hard to measure. Most theories assume that territorial competition increases violence against civilians, while full territorial control decreases violence. We challenge this assumption through three empirical strategies with data from Colombia. First, we measure territorial control through qualitative codings based on interviews with civilians and former combatants and secondary sources. These codings are a major improvement on violence-based measures of control because they can differentiate between controlled areas with little violence and areas where armed groups are not present, and do not rely on theoretical inferences about what violence patterns should indicate particular arrangements of territorial control. Second, we use fine-grained violence event data from CINEP’s Noche y Niebla dataset that indicates the location, the perpetrator, the victim, and the type of violence that occurred. Crucially, these data offer information both on violence between armed groups and violence against civilians. Third, we use a combination of hand coding and machine learning to stratify the event data between urban and rural violence. We do this because urban areas in Colombia almost always have at least nominal state presence, whereas rural areas have a wider variety of territorial control dynamics. We find that the link between territorial competition and violence against civilians is far from straight-forward, and that armed groups commit significant “governing violence” in areas they control. Local histories of civilian organization, civilian-armed group relations, and repertoires of violence specific to certain armed groups are necessary to explain this variation, which territorial competition alone cannot. This study has two major theoretical implications. First, while territorial control is difficult to measure, future studies should refrain from assuming that control can be coded from violence data alone. Second, just as states regulate violence in their territory, armed groups commit violence against civilians not only due to competition with adversaries, but to create local political orders that accord with the groups’ visions of their desired societies.

Authors

• Andres Uribe, University of Chicago

• Daniel Raoul Hirschel-Burns, Yale University

• Andrés Felipe Aponte, Universidad del Rosario