Networks of Resource Extraction: Firm Alignments and Rebel Violence

Research Proposal (PhD Thesis) Research on armed conflict has long documented that natural resources correlate with insurgent violence, yet the dominant literature explains these patterns using macro-level country or resource characteristics such as resource dependence, lootability, or state capacity (Collier & Hoeffler, 2004; Le Billon, 2001; Ross, 2003; Snyder & Bhavnani, 2005). These approaches overlook the political relationships that shape on-the-ground extraction. A growing body of qualitative research from cases such as Colombia, Nigeria, and Syria shows that extractive firms cultivate ties with state actors, local political brokers, and community organizations, and that rebels interpret these ties as signals of political allegiance. Therefore, these firms, geographically bound to a location, operate in a competitive political environment in which states and rebels both seek their allegiance through competitive protection offers. Rebels have, in many documented cases, entered into informal agreements with businesses to secure revenue. However, the field lacks systematic data to capture firms’ positions within these political networks or to test whether network embeddedness, rather than mere resource presence or characteristics, explains why some firm assets are targeted more than others. My project proposes that firm positioning within political networks predicts variation in rebel violence. Companies form relational profiles through their connections to state actors and community organizations. I expect rebels to evaluate these ties when deciding whether to attack or tolerate a firm, given the instrumental and costly nature of violence. I theorize two core alignment axes—state alignment and community alignment—that jointly determine firms’ political embeddedness. I expect that firms strongly connected to state elites and security forces but weakly embedded in local communities will face higher levels of rebel targeting, while firms that cultivate community ties and avoid reliance on state coercive actors will face lower levels of violence. This network-based perspective moves beyond static firm attributes by treating corporate actors as profit-oriented interest groups embedded within contested wartime governance structures. As a pilot study, I am building a project–region–year panel dataset of extractive firms operating in Colombia merging geocoded project locations with geocoded rebel violence data from ACLED and UCDP GED. To infer political networks, I plan to use a structured LLM workflow to extract entities, relational ties, and sentiment from Spanish- and English-language media. With this project, I aim to show how LLM-based network extraction can portray the political embeddedness of firms and explain variation in rebel targeting.