Forecasting the Risk of Extreme Massacres in Syria

Abstract

An abundance of large data sets and improved estimation methods have enabledconflict researchers to estimate the risk of war or terrorist incidents quite precisely. However, as in the case of the prediction of particularly violent earthquakes, forecasting extremely bloody events in continuing conflicts has been difficult until now. This article reports how power laws can be used to predict extreme massacres ex post and ex ante. The power-law distribution that we use is based on the observation that standard probability distributions like the normal distribution typically underestimate the risk of such escalations. Using fatality data until the end of February 2013, we calculate the probability of at least one single event with 250 or more dead civilians at 80% (59%–94%) and between March and May 2013 of up to 48%. We discuss the ethical and practical implications of these findings and argue that the forecasts could provide a transparent risk assessment tool to decision makers.