Not All Intrastate Violence is Civil War: Varieties of Violence and the Perils of Model Dependence in Civil War Research

When is intrastate violence a civil war? Selecting criteria to determine the inclusion and exclusion of cases is one of the most challenging---and contentious---aspects of quantitatively-oriented civil war research. While debate has centered on concerns over selection bias arising from the exclusion of lower magnitude conflicts, we argue that an equally pressing methodological concern---model dependence---can result when conflicts of variable magnitudes are studied uniformly. Major and minor conflicts have different origins and they manifest in different ways. Rather than search for hard and fast thresholds, scholars should instead evaluate their empirical findings across a range of battle death thresholds. Such an approach can provide new leverage when drawing inferences about civil wars, identifying interesting trends in the data that might otherwise be obscured by the strict application of a threshold that is either too high or too low. We illustrate this point with an example drawn from the civil war literature which studies the relationship between drug cultivation and civil war severity.