Lines of Demarcation: Causation, Design-Based Inference, and Historical Research

Abstract

Qualitative historical knowledge is essential for validating natural experiments. Specifically, the validity of a natural experiment depends on the historical processes of treatment assignment and administration, including broader macro-historical dynamics. But if validating a natural experiment requires trust in the ability of qualitative evidence to establish the causal processes through which the data were generated, there is no good reason for natural experiments to be considered epistemically superior to historical research. To the contrary, the epistemic status of natural experiments is on a par with that of the historical research on which their validation depends. They are two modes of social-scientific explanation, each with its own pros and cons; neither is privileged. We illustrate this argument by re-examining an important recent contribution to the literature on violent conflict: Ferwerda and Miller’s 2014 natural experiment estimating the causal effect of the German decision to devolve authority to the Vichy French government on violent resistance during World War II.