Detecting hidden violence: The spatial distribution of excess mortality in Rwanda

Existing sub-national studies on the geography of armed conflict mainly focus on observed battle events. Other dimensions of the conflict cycle, including battle events that remain undetected in news reports, one-sided violence against civilians and the indirect death toll, have gone largely unstudied due to lack of systematic data. This article demonstrates how these different dimensions of a conflict cycle can be detected and how their relative importance can be assessed. The basic tools include population census data, principal component analysis and spatial analysis. When applied to the Rwandan case, the method reveals high excess mortality from the civil war, genocide, (counter)insurgency and the refugee crisis. Hidden violence is detected by presenting the first quantitative evidence to date of high excess mortality in Gisenyi, the northwestern province which was the location of (counter)insurgency and served as a corridor for more than a million refugees. The spatial regression analysis indicates that both the refugee crisis and (counter)insurgency contributed to a high death toll in Gisenyi. This latter finding is important because it provides empirical weight in support of contentious qualitative reports by various human rights groups.

► Observed battle events are only one aspect of armed conflict.
► Other aspects include hidden battles, the indirect death toll and violence against civilians.
► These different aspects can be detected relying on population census data.
► I apply the proposed method to Rwanda.
► I find high excess mortality from the civil war, genocide, (counter)insurgency and the refugee crisis.