From “science as measurement” to “measurement and theory”: the Cowles Commission and contrasting empirical methodologies at the University of Chicago, 1943 to 1955

Abstract

While located at the University of Chicago, the Cowles Commission for Research in Economics advanced a distinctive “Cowles Commission approach” to macroeconomic modelling, using maximum likelihood methods to estimate structurally-identified Keynesian simultaneous-equations models, with the methods presented in two Cowles monographs edited or co-edited by Tjalling Koopmans, a pioneering implementation in another Cowles monograph by Lawrence Klein, and a polemical contrast with the earlier NBER approach in Koopmans’s “Measurement Without Theory” critique of Arthur Burns and Wesley Mitchell. The Cowles approach to empirical methodology was vigorously contested at the University of Chicago by Milton Friedman (a former student of Burns and Mitchell), notably in Cowles seminars, in a defence of “Wesley Mitchell as an Economic Theorist” and at a 1949 NBER conference on business cycles. This paper examines the two contrasting approaches to empirical economics at the University of Chicago in the late 1940s and early 1950s, at the confrontations and exchanges between the two approaches, which contributed to the 1955 move of Cowles from Chicago to Yale.

Keywords:

• Cowles Commission approach to econometrics
• Chicago school
• “measurement without theory” debate

JEL classifications:

• B23 History of economic thought since 1925: Econometrics and Quantitative and Mathematical Studies
• B41 Economic Methodology

Acknowledgements

We are indebted to three anonymous referees for comments and suggestions. Nonetheless, the usual disclaimers apply.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Friedman’s criticism in Cowles Commission seminars of the Cowles approach to modelling began earlier, within months of his and Koopmans’s faculty appointments. Roy Epstein (1987, 107) cited minutes of the Cowles staff meeting of 20 September 1946 when stating that “Friedman constantly intimated in the [Cowles] seminars that the estimation results discussed there merely reflected the prejudices of the investigator, prompting 1987 Koopmans to exclaim at one point, ‘But what if the investigator is honest?’”

2 As reminded by Christ (1994), “Leonid Hurwicz (1950) defined a model as a set of structures that are admissible according to this a priori information. Thus the problem in the Cowles approach is first to construct the model, and second to estimate the structure” (Christ 1994, 36)

3 Christ (1994) repeats that this was not a point of contention between Friedman and the Cowles group, for Friedman (1950) argued that “a theory should be judged by its results, rather than by whether its assumptions are realistic or not” (Christ 1994, 47).

4 “Marschak (1950) characterized a model by means of a set of a priori information (true or false) about the structure, information obtained from economic theory, not from observed values of the variables.” (Christ 1994, 36)

5 The wording recalls the title of the first draft of his essay, “Descriptive validity vs analytical relevance”.

6 The following footnote was deleted in the published version: “I have greatly benefited from discussions with Milton Friedman of some of the problems involved. These are my views, however, and Friedman should not be held accountable for any errors or confusions that may be involved in these views” (Milton Friedman papers, box 34–33, Hoover Institution archives).

7 In his “Probability Approach in Econometrics” (1944), which was published as a Cowles Commission paper, “Haavelmo gives a rough fourfold classification of the main problems encountered in quantitative research: first, the construction of tentative theoretical models ; second, the testing of theories ; third, the problem of estimation ; and fourth, the problem of prediction” (Vining 1949b, 82).