| Summary/Abstract |
The increasing availability of event data has led to a surge in micro-level research investigating civil war violence. Event data, however, introduces bias due to systemic underreporting of events. This study introduces a method that accounts for this bias by making a spatial prediction of unreported events based on the known locations of rebels and their preferences for local-level factors estimated using habitat analysis. I then combine this prediction with known locations of rebel activity and use it to predict rebel territory and relocation patterns. I next perform a regression analysis on 119 African rebel groups that links preferences to conflict behavior. Preferences for gold, petroleum, gemstones, and cobalt are associated with civilian victimization, whereas rebels who prefer steeper slopes are more successful in battles with government forces. Further, government forces tend to be successful when battling rebels with preferences for petroleum and international borders.
|