Abstract

A newer strand of research in historical political economy applies a "design-based inference" approach to history in order to approximate a randomized controlled trial. But can this exacting approach work given the messy nature of historical data? Using the example of research on the long-term effects of British colonialism in India, I evaluate six recent articles that use techniques like natural experiments of history, instrumental variable analyses, and matching designs to overcome the fact that colonization was not random. I find that despite generating important methodological conversations about causation, the use of these techniques in these studies depends on thin or sometimes inaccurate historical evidence. It is therefore unclear that "randomized controlled history" can make more credible causal inferences than a selection on observables approach. This article suggests best practices for future research that aims to study history in an experimental format.