Can Statistics and Experimental Designs Help Process Tracing?

• In Event: Process Tracing. New Directions in Qualitative Research

Abstract

Most existing research on multi-method designs concentrates on frameworks in which qualitative elements contribute to a fundamentally quantitative measurement, description, or causal inference. This essay argues that things can productively work in the opposite direction. Cutting-edge research on qualitative process tracing reveals important assumptions: that the set of alternative explanations considered in the study includes most of the important potential hypotheses, that links in process-tracing chains that involve political psychology and/or mass behavior are credibly tested, and that background clusters of moderators are configured in a way that fits with a common pattern within the relevant universe of cases. This essay proposes research designs using random forests, LASSO regression, and experiments that can test and refine these assumptions. These designs, together with examples of applications, demonstrate that fundamentally qualitative research designs can benefit from multi-method approaches just as much as fundamentally statistical designs.

Authors

• Jason Seawright, Northwestern University

• Kendra L. Koivu, University of New Mexico