Testing Richardson's Law: A (cautionary) note on power laws in violence data

Abstract

Richardson’s Law states that the frequency and severity of armed conflicts exhibit a power law relationship. This idea has shaped theories of violence and found applications in forecasting and missing data estimation. However, it is based on examining only subsets of violence data. This paper re-examines Richardson’s Law using recent data collection efforts. Looking at micro-level violence from 16 datasets containing 700,000 conflict events across the world, the study finds that power laws sometimes provide a good fit and sometimes they do not, indicating that Richardson’s Law as a law-like regularity needs qualification. Nonetheless, the results shed new light on the heavy-tailed nature of conflicts and highlight new stylized facts for future research to build on. – AI-generated abstract.