Beyond mill: Why cross-case qualitative causal inference is weak, and why we should still compare

Qualitative cross-case comparisons were once widespread and respected enough to be described as “the comparative method.” However, the current wave of research on qualitative methods has seen cross-case controlled comparisons fall substantially in esteem. Early criticisms based on selection bias by Geddes (1990) and King, Keohane, and Verba (1994) have been disputed and no longer receive sustained attention in the qualitative methods literature. A more recent argument is that qualitative comparison fails for purposes of causal inference because the required assumptions are simply implausible and because statistical methods are superior tools for the same purpose. Sekhon (2004) argues that comparisons based on Mill-type methods will always be susceptible to probabilistic alternative hypotheses, which generally cannot be reasonably evaluated using qualitative crosscase comparisons. George and Bennett (2004, 151–79) argue at length that “practically all efforts to make use of the controlled comparison method fail to achieve its strict requirements,” and that various within-case qualitative methods are simply more usable than qualitative rosscase comparisons. Collier, Mahoney, and Seawright (2004) characterize many forms of qualitative cross-case comparisons as a form of “intuitive regression” that acts inferentially as a weaker and problem-laden equivalent of statistical analysis. Seawright (2016, 107–9) argues briefly that a potential-outcomes formulation makes evident that qualitative comparisons are exceptionally weak tools for causal inference.