Predicting the severity of civil wars: An actor-centric approach

We introduce an actor-centric approach to predict the severity of conflict one and six months into the future. We argue that the prediction of conflict severity needs to focus on the actors that are responsible for conducting armed violence. Hence, we predict the severity of conflict in government-rebel organization dyads. Our predictors focus especially on rebel organization characteristics, behavior, and the conflict networks actors are embedded in. Our statistical learning approach relies on random forests to predict the severity of conflict. We demonstrate that our model performs especially well distinguishing high levels of severity from very low levels.