War and Excess Mortality

Abstract

Hagopian et al. (2013) published a headline-grabbing estimate of half a million excess deaths in the Iraq war based on data from the University Collaborative Iraq Mortality Study (UCIMS). We analyse these data and find that the Hagopian et al. (2013) claim is wrong. When we account for the stratified sampling design of the UCIMS the bottom of the 95% uncertainty interval for excess deaths drops below zero. These ranges shift further left to the extent we discount reported deaths not backed by death certificates. The evidence in favour of even a positive number of non-violent excess deaths is extremely thin, even ignoring the issues of stratification and death certificates. This finding is reinforced by a Differences-in-Differences (DID) analysis that finds no local spillovers running from violence to elevated non-violent death rates. We argue that future research should supplement traditional excess death estimates with estimates on non-violent deaths only and with DiD estimates.