Weekly Philo Economics: On Econometrics as Policy Science

[I am giving four talks this week (here, here, here, here), so I have been distracted from the blog. Come say hi if you are in the Baltimore/DC area!--ES]

In 1949, (the future Nobel-laureate) Tjalling Koopmans published a blistering review, "Measurement Without Theory," of the then most detailed (and sophisticated) empirical effort at understanding business cycles. The review prompted a celebrated methodological exchange (about which more some other time) with Rutledge Vining. (Hat-tip to Roger Backhouse for calling my attention to it a few years ago.) Koopmans is widely thought to have won the debate (and, thus, inaugurated the age of econometrics). Koopman's argument draws an inordinate amount on an interpretation of the history of physics (Tycho, Kepler, and Newton), but here I focus on the following claim by Koopmans:

"..a still further objective is included, which extrapolates the idea of explanation: the prediction, within the narrowest attainable limits of error, of the effects of stated hypothetical measures of economic policy on the level and movements of economic variables. However, I feel that such prediction is actually the most important objective of the analysis of economic fluctuations. The criterion of social usefulness of scientific analysis gives us the right to discuss the merits of any particular approach to the problem of economic fluctuation on he basis of the guidance it gives to economic policy, even if such guidance were not claimed by the authors" [A.F. Burns and W. C. Mitchell--ES]. (T. Koopmans, 1949, “Measurement Without Theory,” 167)

In context, Koopmans is making a number of points:

1. He is attacking an empirical approach that investigates aggregate, economic phenomena without reducing these to (what we would now call) micro-foundations. This is non-trivial because it not-so-subtly pushes economic research away from understanding complex systemic phenomena on their own terms.
2. Koopmans is defending a notion of economic explanation that incorporates policy impact as an essential feature of one's evaluation of the merits of explanation. This is non-trivial because it subtly pushes the notion of economic explanation away from topics that are unrelated to policy interventions.
3. This second point leads us to the most significant aspect of Koopmans' position: the elevation of the criterion of social usefulness.

The mathematical econometric techniques and tools – and more generally inferential technologies that produce univocal and stable figures in calculating the implications of policy alternatives – were promoted within economics and, of course, to policymakers, in part because they would make economists attractive as policy experts (as opposed to say, sociologists, lawyers, anthropologists, and historians). (Recall my treatment on the Alchian move.) To do so Koopmans had to displace a very different vision for economics, one that focused primarily on understanding long-range economic phenomena. Economics has been hugely successful in becoming the indispensable policy science, but it is by no means clear that now, more than a half century later, we really understand long-run economic phenomena much better than the great economists of the eighteenth and nineteenth centuries.