John Stuart Mill on empirical economics and causal inference

Written by me, here is a passage from GOAT: Who is the Greatest Economist of All Time, and Why Should We Care?

A System of Logic covers many different topics, but for our purposes the most important discussion is Mill’s treatment “Of the Four Methods of Experimental Inquiry,” sometimes called “Mill’s Methods” and indeed receiving their own Wikipedia page. Mill outlines different manners in which causes and effects might be correlated, or not, and what we can infer from such patterns, and how difficult it can be to sort out actual cause and effect from the data. He refers to the “direct method of agreement,” the “method of difference,” “joint method of agreement difference,” the “method of residue,” and the “method of concomitant variations,” all as ways of trying to make correct or at least better inferences from the data.

I’ll spare you the details on the full argument, but in essence Mill was trying to figure out how to do causal inference econometrics, but with words only. That enterprise was doomed to fail, but it gives us insight into what Mill thought was by far the most important question in social science, namely causal inference when faced with complex underlying chains of cause and effect. For Mill, everything is what we would now call “an identification problem,” and this understanding is clearest in Mill’s chapters “Fallacies of Generalization” and “Fallacies of Ratiocination.” Mill also serves up a remarkably on-target discussion of how the different nature of social science problems, and their possibly greater complexity, can lead to identification problems that are not necessarily present in the natural sciences – see his chapter “Of the Chemical, or Experimental, Method in the Social Science.” That entire approach is remarkably 2020s in orientation, and you won’t find earlier history of thought books giving Mill much if any credit for this.

In a funny way, Mill was ahead of Milton Friedman in his understanding here. Friedman knew much more statistics, but in his economics he often presented causal inference as fairly straightforward. In his Monetary History of the United States, co-authored with Anna Schwartz, the reader does get the impression that the historical correlations, and ordinary least squares techniques, do in fact show that the money supply is a central driver of nominal income, given the relative stability of money demand. Later, the real business cycle theorists were to challenge that inference, and suggest that often it was income that was causing the money supply. That is a kind of complex challenge Mill seemed quite comfortable with in A System of Logic, whereas Friedman and Schwartz assigned higher power to common sense approaches to cause and effect.

Mill remains in my eyes one of the most underrated thinkers.