The Rise And Fall Of Extraneous Estimation: Lessons From Econometric History?

It is possible to view econometric history as the retrospective assessment of intellectual resource allocations. This paper uses such a perspective to examine the history of extraneous estimation. Independently proposed by Marschak (1939, Econometrica 7, 332–335) and Wold (1940, Efterfrågan på jordbruksprodukter och dess känslighet för prisoch inkomstförändringar (The demand for agricultural products and its sensitivity to price and income changes)), extraneous estimation was widely used as a collinearity solution in demand analysis from 1950 to 1980. We give a detailed account of the origins of this procedure and the theoretical work that clarified the statistical structure that underpinned it. The subsequent literature is then assessed in terms of the optimal use of available econometric tools and the interaction between theory and applications. We find that optimal econometric procedures were rarely used and that theoretical refinements continued to be published well after extraneous estimation had ceased to be a useful tool for demand analysis.An earlier version of this paper titled “To Pool or Not to Pool? Why Is That Still a Question?” was presented at the Thirteenth Annual Meeting of the Canadian Econometrics Study Group, University of Waterloo, September 21–22, 1996. I thank Robin Carter for his comments at that presentation. Another version titled “The Optimum Amount of Econometrics” was presented on October 31, 1998, at Concordia University on the occasion of Gordon Fisher's seventieth birthday. I am indebted to Lady Giovanna Stone and Peter Jones of King's College Library for arranging access to Sir Richard Stone's papers and to the staff of the UCLA Research Library for access to the Jacob Marschak Archive. James Tobin was kind enough to share some of his recollections of the “extraneous estimation” period. George Szava-Kovats and Chris Minns provided able research assistance, and Charlene Hill typed the manuscript with her customary efficiency. Colin and Sally Ash, Reinhold Bergstrom, Jim Hamilton, Balder von Hohenbalken, Joe Ostroy, and Martin Weale provided help of the sort that does not lend itself to detailed elaboration. An earlier version of the paper had the benefit of a close and perceptive reading by John Aldrich. The referees' and editor's comments have also helped shape the paper into its present form, and I thank them for their contributions.