top of page
I study the nature of causation and of causal inference, especially in political science and economics. This work has been funded by a European Research Council Starting Grant which is running 2017 to 2022 Qualitative and Quantitative Social Science: Unifying the Logic of Causal Inference?
Social scientists disagree over the nature of causation as well as over the best methods for identifying causes. This is especially true of the disagreement between social scientists in the quantitative tradition versus social scientists in the qualitative tradition. Even though this disagreement is long-standing and seemingly intractable, I still think that one can develop a principled theory of causation that does justice to the concerns, insights and goals of both traditions.
To do this, I take a conceptual engineering approach to causation. This approach asks: how could we use the concept of causation so that it is maximally useful for the predictive and explanatory goals of the social sciences? This offers a principled way to adjudicate many of the disagreements over the nature of causation (within and between the quantitative and qualitative traditions). In particular, it allows one to examine the rival assumptions that social scientists make about the nature of causation. For each assumption one can ask: under what circumstances is this assumption valid?
In work that is currently in progress, I make the following contributions to the discussion:
The Causal Markov Principle and the related Principle of the Common Cause are assumed (implicitly or explicitly) by all quantitative methods of causal inference. Cartwright and Sober have claimed that these principles do not hold under all circumstances. For example, these principle do not hold for mixed populations or for time series. I use conceptual engineering to identify the precise circumstances under which these principles hold. On the one hand, these principles do not hold in all circumstances, I argue. On the other hand, they can hold for mixed populations and for time series.
Some methodologists argue that the sort of causation that quantitative political scientists study is distinct from and independent of the sort of causation that qualitative political scientists study. I deny this on conceptual engineering grounds. If the qualitative concept of causation is to be a useful one, I argue, then it has to be closely related to the quantitative concept of causation in terms of counterfactual conditionals, although it doesn't have to be identical to it. This has implications for how process tracing, as a qualitative method in political science, ought to proceed.
In published work, I have cleared some of the ground for the above project:
In Paper 8, I look at process tracing as a core qualitative method. I argue that extant accounts of process tracing fail to make clear what makes process tracing distinctive in comparison with econometric methods.
In Paper 9, I look at the concept of one variable making a causal contribution to another variable, and at the related concept of a what-if question. What account of causal contributions and what-if questions best suits economics? The interventionist answer to this question assumes that structural equations are modular. In contrast, what I call the ceteris paribus approach drops this modularity assumption. However, the result is often that causal contributions and what-if questions are indeterminate, I show.
bottom of page