Continuing the chapter-by-chapter review of Seawright’s book on Multi-Method Social Science took me longer than I imagined and it should have, but here we go again. The second chapter discusses the fundamentals of multimethod research (MMR) and identifies “Causation as a Shared Standard”. One might think that the fundamentals can be quickly addressed because of the now established distinction between causal effects, studied with a cross-case method, and causal mechanisms that are studied with process tracing. However, this is not the route the chapter is taking because it is mostly concerned with discussing the potential outcomes (PO) framework as the basis for MMR.

In short, the PO framework couches causal analysis in a counterfactual framework by asking what the outcome value of a unit would be if it had been exposed to a different treatment. The first section of chapter 2 introduces the reader to the PO framework and explains how it relates to different views on causation in the social sciences. Brady’s seminal discussion of causation is the point of reference here and leads Seawright to the conclusion that the PO framework is suitable for integrating the existing perspectives – regularity, manipulation, mechanisms – more or less.

If MMR involves a regression analysis, it is straightforward to rely on the PO framework as the foundation of modern quantitative research. However, my reading is that the first section treats other perspectives on causation too lightly and tends to misrepresent their complementarity. If one conceives of process tracing as a method that allows one to make causal inferences with a single case (I don’t, but others do), it cannot be integrated into a PO framework requiring counterfactual reasoning. Seawright’s argument that process tracing and the PO framework allow one to target heterogeneity seems to force the issue a bit because these are different types of heterogeneity.

At least of equal importance, I am hesitant to put the PO framework on par with theories of causation because it is, above all, a framework for the estimation of treatment effects of all sorts (so I read). I would have loved to see a discussion of Pearl’s directed acyclic graphs and his methodological theory of causation using structural equations to translate theoretical expectations into causal models. Apart of putting theorizing center stage, it might create more opportunities for linking regression analysis and process tracing in MMR because there are also attempts to formalize causal inference for single cases with structural equations. This is not to say that the integration of both methods would be easy because, after all, they are operating on different levels. Yet, if the goal is to formulate a unified framework for MMR using similar terminology, this seems more promising to me than the PO framework.

Chapter 2 then moves on with a short discussion of using regression analysis for causal inference and a longer elaboration on regression analysis as seen through the lenses of the PO framework. The sections are written in more or less technical terms and might be new to qualitative researchers and read as familiar for quantitative researchers. In total, this means that regression analysis is central throughout the entire chapter and foreshadows what becomes clearer when reading the following chapters: For Seawright, the causal work is entirely done by the quantitative part and the qualitative analysis supports the quantitative inference in one way or the other (see the forthcoming blog post of chapter 3).

Seawright’s conception of MMR therefore differs considerably from what one might believe to be the widely held idea that assigns equal weight to the quantitative analysis (or QCA) and the process tracing part. There is nothing wrong with this *per se*, but it might have been appropriate to flag this upfront in the introduction. As I explained, the key argument in the introduction should hardly be contested in the political science community on nested analysis, whereas the dominance of the quantitative part in MMR might be. In this context, it is noteworthy that ‘process tracing’ is not mentioned once in chapter 2 and causal mechanisms are barely mentioned. It would have been interesting to learn in more detail why Seawright’s conception of MMR departs from nested analysis as introduced by Lieberman, but this question remains unanswered.