Testing Richardson's Law: A (Cautionary) Note on Power Laws in Violence Data

Abstract

Richardson’s Law says that the relationship between the size of violent wartime events and their frequency is characterized by a simple probability distribution called a power law. Power law distributions have been found in both civil war violence and terrorist attacks and have recently served as the foundation for novel theories of conflict, solutions to missing data problems, and prediction models. This note revisits Richardson’s Law in light of recent data collection efforts, looking at all relevant micro-level conflict event data publicly available in the world today (685,000 events across 16 data sets). I find substantially less support for Richardson’s Law than past research, suggesting serious caveats to claims about its universal nature. By identifying new stylized facts about the heavy-tailed nature of violence, this note lays the groundwork for more nuanced analyses of conflict severity.