The Bodies of Evidence: An Exploration of Multiple Systems Estimation for Empirical Research with Violent Deaths. Jule Krüger
Some political phenomena are difficult to observe. Model-based approaches offer a measurement solution when available observational data are not complete or representative of the research population. I illustrate this challenge using the example of political violence that, in its many forms, is typically measured by counting the bodies of the dead. While an accurate account of lethality is a prerequisite for testing key hypotheses empirically, limitations on what can be observed inhibit the collection of representative data required for valid and reliable statistical inference. Examining the case of armed conflict in Kosovo (1999), I demonstrate that the statistical method of multiple systems estimation (MSE) can be used to estimate the number of unobserved deaths from three incomplete lists of observed victims. I compare estimates of the number of dead to the observed data and a recently completed war victim census. Only the estimates and the census reveal a theoretically relevant relationship between ethnicity and lethality. This provides-to my knowledge-the first test of MSE in the context of a human population experiencing violent deaths. It emphasizes that political violence scholars need to account for underregistration and bias in their empirical work. I further show how MSE allows to model data-generating processes, account for list dependence, and identify research contexts in which more data needs to be collected. I conclude with recommendations for future research.