The extensive coverage of household surveys in conflict regions in recent
decades has fueled a growing literature on the microeconomic effects of war. Most
researchers identify these effects using econometric methods, with difference-indifferences
– which exploits variation across birth cohorts and war intensity – being the
most popular. This paper highlights problems of endogenous war intensity and selfselection
due to non-random displacement when using common empirical methods on
cross-sectional data, and explains how they can be overcome with panel data. We draw on
a unique set of cross-sectional and panel data from Nepal to demonstrate our proposition.
Both unobserved locality factors and individual heterogeneity lead to huge swings in the
estimates of war intensity effects. Our results imply that researchers ought to think
carefully about empirical methods and explain possible statistical biases, especially when
their results are used to inform policy decisions. For researchers who use panel data, we
propose augmentations to existing methods.