The maximization of such an expected value may also be regarded as a maxim for behavior. As such, it promises to have important implications for statistical theory. Cf. L. J. Savage, "The Theory of Statistical Decision," Journal of the American Statistical Association, XLVI (March, 1951), 55-67. Success of the maxim in this domain, as in any other area of decisions involving uncertainty, depends, not on its empirical verification for the economic behavior of men at large, but on its acceptability, to individuals who are particularly concerned with such decisions, as a rule guiding "wise" behavior in the face of uncertainty.--Milton Friedman & L. J. Savage (1952) The Expected-Utility Hypothesis and the Measurability of Utility (463).
Yesterday, at a workshop I co-hosted with Rogier de Langhe in Ghent, David Teira called attention to the quoted passage in a paper (co-authored with Rasmus Winther) he presented. The passage occurs in the first substantial footnote of the Friedman/Savage paper (which by building on work by Von Neaumann & Morgernstern, together with their earlier 1948 paper, generated the canonical treatment of the axioms of expected utility theory).* Hereafter I will call their joint 1952 paper: 'F&S1952.' Teira astutely noted that the quote puts a normative ("wise") interpretation on expected utility theory. This is surprising because standard historiography suggests that the normative interpretation was a much later consequence, the result of criticism (see, for example, this fine paper by Hands.) In fact, the standard story can base itself on F&S1952:
Two quite different classes of objections are made to this particular convention for assigning a "measure" to utility:... (2) that any convention that would treat utility as measurable is undesirable or unnecessary... The second objection, on the other hand, cannot be accepted. It would deny economics a mode of speaking and expressing theories that has been found useful if not indispensable in other sciences. It expresses a viewpoint that its proponents would find impossible to apply consistently; they would have to deny themselves the privilege of speaking of money income, or stock of money, or size of population as measurable. If this viewpoint seems persuasive for utility, it is, we think, partly because some hypotheses that would, if verified, have implied convenient measures of utility were actually flagrantly contradicted by experience, and partly because of a widespread confusion between positive economics and normative economics, a widespread tendency to use the same word---utility- to stand for two quite different things: on the one hand, a quantity that it is useful to regard an individual as maximizing in interpreting his behavior and predicting his reactions to changed circumstances, and, on the other hand, a quantity that he "should" maximize or that society "should" maximize or help him maximize. The identification of these two conceptually different magnitudes is a maxim for wise behavior or an ethical precept, not a scientific proposition, and is in no way required by the adoption of a particular convention for "measuring" utility. (474)
This is the closing paragraph of F&S1952. Here wise behavior is in the domain of normative not positive (or empirical/scientific) economics. As is well known, for Friedman the distinction between normative and positive economics is central to his methodology of economics (which he was working on at the time and which he published in 1953). And, his ''as if" methodoloy is fully present in the first sentence of F&S1952: "individuals choose in such circumstances as if they were seeking to maximize the expected value of some quantity." Now, Teira offered an interesting explanation for the presence of a normative conception in F&S1952, and here I don't want to steal his thunder.
Before I get to what I take to be the significance of the footnote, I should note that, I think the footnote reflects the influence of Savage. For three reasons: (i) it fits uneasily with Friedman's larger methodology at the time; (ii) in the note they explicitly appeal to Savage's work, and (iii) Savage thought that one could treat uncertainty as a species of risk: one treated the probabilities as subjective (see his Foundations of Satistics), although Savage was quick to note that “we must be prepared to find reasoning inadequate to bring about complete agreement.” (1954 [1972]:7; see also 3, 67ff.) So, Savage did not endorse the ideal that there could be full consensus about wisdom. By claiming that this line of thought reflects Savage's influence, I do not mean to deny the significance of it to Friedman's thinking.
For, the idea that wisdom in the face of genuine uncertainty could still be a species of (mathematical) rule-following is an important rupture with earlier modes of thought which understood wisdom in terms of judgment. For example, at the hight of the financial crisis, Keynes had argued, reflecting widely held belief, that in the face of genuine (computational) complexity, that is uncertainty, central bankers should be given considerable "discretionary power" (to manage its reserves): he denied that “the rules of wise behavior by a central bank could be conveniently laid down -- having regard to the immense complexity of its problems and their varying character and varying circumstances." (Quoted from Bibow.) Of course, Keynes insisted that the discretionary power should not be absolute; as Bibow notes, Keynes thought that the economic and moral aims of the central bankers should be set by government.
Friedman, by contrast, in reflecting on the experience of monetary policy during the financial crisis in his great work with Anna Schwartz came to think that much of monetary policy ought to be rule-based (see, for example, here, and here). Whether we think of these rules as robust heuristics (to be ignored in some future emergencies) or as a species of optimizing rationality is a further question that lay in the future. Sadly, whether it's wise to think of wise behavior in conditions of uncertainty as an instance of utility maximization was not debated in full when introduced into the literature, perhaps because it was treated as pertaining to 'individuals' (who are free to do what they wish) rather than social institutions; or, perhaps because sometimes footnotes fly under the radar screen.
*In Friedman's collected papers, the original first footnote has been removed, and this has altered the numbering of all the notes.
Comments