"Is There More Violence in the Middle?" by Zachary M. Jones and Yonatan Lupu.
Is there more violence in the middle? A series of studies over the last twenty years has suggested that violent outcomes such as civil war and human rights abuses are more common in regimes that are neither full autocracies nor full democracies. Much of the literature takes this notion as a given, despite the fact that it has not been subjected to sufficient testing. Existing work contains references to multiple types of violence, the outbreak of which is likely related to regime characteristics but also likely related to other types of violence. In addition, existing work largely uses arbitrary operationalizations of "the middle" of the regime type range. zMost importantly, all of the existing work testing this hypothesis, which is ultimately about functional form, does so using models in which a particular functional form is assumed. This project aims to describe which types of regime are predictive of which types of conflict. We aim to understand better the extent to which conflict is more likely in regimes that are not on either end of the spectrum. We use a random forest-like ensemble of multivariate regression and classification trees to simultaneously predict multiple forms of conflict using a large set of explanatory variables. We compute the marginal relationship between regime type and political violence for multiple measures of regime type to explore the possibly nonlinear and interactive relationship between conflict, regime type, and explanatory variables that are related to regime type and political violence. Our methodology allows us to estimate the conditions (both in terms of regime type and type of violence) under which there is or is not more violence in the middle.