Monday, August 13, 2007

Koopmans On Friedman's Claimed Methodology

Over on Crooked Timber, "reason" says he likes a James Galbraith quotation I refer to. Galbraith mentions Koopmans. I suppose Galbraith is talking about this:
"Here the 'direct' implications of the postulates, their accuracy in describing directly observed individual behavior, are placed [by Friedman] in a category with which we need to be less concerned.

There are several objections to such a concept of theory construction. In the first place, in order that we shall have a refutable theory at all, the postulates then need to be supplemented by a clear description of the class of implications by which the theory stands or falls. Otherwise, every contradiction between an implication and an observation could be met by reclassifying the implication as a 'direct' one.

This objection is met by Friedman's suggestion that there should in each case be 'rules for using the model,' that is, a specification of the 'class of phenomena the hypothesis is designed to explain.' But a second objection arises out of this answer to the first. To state a set of postulates, and then to exempt a subclass of their implications from verification is a curiously roundabout way of specifying the content of a theory that is regarded as open to empirical refutation. It leaves one without an understanding of the reasons for the exemptions. The impression of ingeniousness that this procedure gives is reinforced by the fact that in each of Professor Friedman's examples he knows more about the phenomenon in question than he lets on in his suggested postulates. He is willing to predict the expert billard player's shots from the hypothesis that the player knows the mathematical formulae of mechanics and computes their application to each situation with lightning speed, even though he (Friedman) knows that most experts at billards do not have these abilities. He is willing to predict the distribution of leaves on a tree from the hypothesis that each leaf seeks a position of maximum exposure to sunlight (given the position of all other leaves), although no one has reported observing a leaf change its location on a tree.

One cannot help but feel uneasy in the face of so much ingenuity. Truth, like peace, is indivisible. It cannot be compartmentalized. Before we can accept the view that obvious discrepancies between behavior postulates and directly observed behavior do not affect the predictive power of specified implications of the postulates, we need to understand the reason why these discrepancies do not matter. This is all the more important in a field such as economics where, as Friedman also emphasizes, the opportunities for verification of the predictions and implications derived from the postulates are scarce and the outcome of such verification often remains somewhat uncertain..." -- Tjalling C. Koopmans (1957). Three Essays on the State of Economic Science, McGraw-Hill: 139-140
Koopmans also discusses methodology, in an attack on American institutionalism, in his "Measurement without Theory" (Review of Economic Statistics, V. 29, N. 3 (August 1947)). I have already pointed out some more recent criticisms of Friedman's methodology.

No comments: