Monday, September 15, 2014

History of econometrics: special conference session at the EEA-ESEM meeting

One of my secret manias (and one that I am less ashamed to admit) is an interest for the history of econometrics. With such an inclination, the two successive sessions of the EEA-ESEM meeting on the history of econometric thought were obviously an absolute delight to me. Here, I am going to try to convey some of the excitement that I got out of this sessions by giving a quick summary of the papers presented. This post will also appear in the next issue of the TSE Mag.

In the first talk of the first session, John Aldrich detailed how the father of modern econometrics  Trygve Haavelmo, contributed to the understanding and formalization of the concept of causality. John especially emphasized the modernity of Haavelmo's views on this topic and how they have resurfaced in more recent literature. Marcel Boumans described the interaction between Milton Friedman and the members of the Cowles Commission, the founders of modern econometrics. Both interacted in the late forties in Chicago. Marcel gave a fascinating account of how much Friedman was involved into methodological debates at the time, that eventually yielded to the writing of his essays in positive economics. Friedman was at the time extremely critical of the strong theorisation of econometrics promoted but the then Commission president Tjalling Koopmans (Marcel quoted from an unpublished assessment by Friedman of the works of the Cowles commission). At some point, Friedman defied Koopmans and one of his boys, Lawrence Klein, to check whether the complex macro simultaneous equations model estimated on pre-war data could predict macro conditions in 1948. This was a dismal failure and yielded Klein to reassess his model. Marcel put this debate in perspective by recalling how it related to the views of early economists Alfred Marshall on one side and Léon Walras on the other. Till Dueppe presented a thorough account of a trip taken by Tjalling Koopmans to the Soviet Union in 1976. Koopmans was really eager to discover if the linear programming methods developed by Leonid Kantorovich were applied in practice by the planning bureau of the Soviet Union. What he found was a very dispirited Kantorovitch, saying that they might have read about it, at least it was somewhere in published form. Koopmans was also somewhat disappointed by the quality of economics research that he found there. He was on the other hand extremely impressed to meet with some of the most outstanding mathematicians of the world.

The next session started with a minutely accurate detail of the first meeting of the econometrics society in 1931 in Lausanne by Olav Bjerkholt. What striked me the most was the impressive imprint that Ragnar Frish left on this meeting and on the early years of the econometric society. While François Divisia, the Vice President officially in charge of organizing the meeting was preoccupied by such urgent matters as choosing the suitable French translation of "econometrician," Frish used his spare time while recovering from a skiing accident to write a sketch of a programme and one of the three papers he presented at the conference. Frish also insisted that the conference opened with a presentation of some of the works of six of the founding fathers of economics (Marshall was not one of them). This echoes evidence given by Marcel Boumans that early Econometrica issues published handwritten notes by Walras, along with his correspondence with other economists. Duo Qin advocated the return of the use of the terms autonomy and confluence in modern econometrics textbooks. These terms have been coined by Frish to separate what we now call structural and reduced form relationships. These late notions were in fact derived from the earlier notions by Tjallling Koopmans. The great advantage of the term "autonomy" is that it puts emphasis on the fact that we are looking for invariant relationships that remain true when the other relations in the economy change. The Lucas critique is in a sense a mere application of the notion of autonomy, which Lucas acknowledged explicitly in his paper (see the reference to Marschak in footnote 3). Cléo Chassonery-Zaïgouche presented the evolution over the years of the empirical analysis of discrimination. She especially showed that the traditional econometric techniques developed in the 70s in the Cowles tradition are progressively replaced by experiments. This trend is not apparent in the number of papers published in top journals though, but is clear in the influence of the published papers: experiments gain much more citations. Finally Jan Höffler presented an exciting collaborative wiki project on replication in economics. With his colleagues, he has set up a list of the published papers in top journals with a link to the data sets and code. With his students, they are now trying to replicate the authors' findings. Being a collaborative project, anyone can report the results of his own replication exercise on the web page. An exciting exercise for students in econometrics.

No comments:

Post a Comment