Ever since Sir William Petty began recording (and inventing) economic data in 17th Century Britain, empirical work has always played an important role in economics for two reasons: namely, economists think that (1) economic insights can be gained by careful examination of the data (the "inductive approach") and (2) existing economic theories can be validated or falsified by comparing their claims against empirical data (the "theoretical approach"). However, not all economists agree with these two reasons, and those that agree with one of them, often disagree with the other.
The inductivist approach has a long history: from data, Jevons (1875, 1884) gleaned evidence of a sunspot-driven business cycle; Juglar (1862) saw in financial tables evidence for a credit-driven cycle. Similarly, H.L. Moore (1914, 1923) used the inductive approach to argue for a weather and astral-driven cycle.
However, the theoretical approach was also used: as far back as D'Avenant and Jenkin, economists had attempted to fit data to demand curves. This was the particular skill of the same H.L.Moore who attempted to fit data to Walras's demand equations. His students at Columbia, Henry Schultz and P.H. Douglas carried this work on into the 1930s. Similar work was carried out in Britain by A.C.Pigou and in Germany by Jacob Marschak.
Another development occurred at Columbia soon after: a return of the inductivist approach for analyzing business cycles under Wesley C. Mitchell. The inductivist approach had, naturally, been favored both by the German Historical School and the American Institutionalists, but only under W.C. Mitchell and his National Bureau of Economic Research (NBER) was much systematic research done in this regard. The measurement of business cycles were the main topic and Mitchell's NBER included such notable contributors to empirical economics such as Arthur F. Burns, John Maurice Clark, Simon Kuznets, Frederick C. Mills, Rutledge Vining, Solomon Fabricant, Leonard P. Ayres and others of the American Institutionalist School. Business cycle measurement and analysis also in vogue elsewhere - such as under Persons and Bullock at Harvard, Kondratieff in Moscow, Wagemann in Berlin, Åkerman in Lund, Morgenstern at Vienna, and at the Kiel Institute.
Naturally, the empirical recording and analysis of the business cycle spilt over into collecting and analyzing all sorts of empirical data - notably, the collection of national income accounts (output, investment, etc.), which became the primary activity of Simon Kuznets at the NBER and of Colin Clark in England. It was also in England that the massive collection of family expenditure data was undertaken by R.G.D. Allen and Arthur Bowley.
Naturally, theoretical treatises on the business cycle, such as those by J.A.Schumpeter, D.H.Robertson, A.C. Pigou and G.Haberler, had been accompanied the presentation of empirical data, but they could not have said to be adopting the second "theoretical approach" entirely because they used no proper statistical inference methods to do so.
The NBER approach was severely criticized by George Yule, Eugene Slutsky, Ragnar Frisch and, most famously, Tjalling C. Koopmans. This resulted in a rebirth of the theoretical approach and thus econometrics as we know it. The theoretical approach had been first applied usefully to business cycles by Jan Tinbergen in the late 1930s after the appearance of the General Theory of J.M. Keynes. By proposing, some simple functional relationships between some relatively accessible variables (consumption, income, investment, etc.), Keynes had help prompt Tinbergen into action. Tinbergen used statistical methods, such as linear regressions, to estimate the parameters of the Keynesian relationships. Keynes himself was not too pleased with Tinbergen's techniques, considering them little more than "black magic" (Keynes, 1939). Keynes criticism of Tinbergen was only the first salvo in a series of critical reappraisals of Tinbergen's econometrics undertaken by Frisch and Haavelmo, Allen, Marschak and Lange.
Galvanized into action, a great leap forward was achieved with the "probabilistic approach" to econometrics as famously proposed by Trygve Haavelmo (1944) and then taken up as the rallying cry of the Cowles Commission. Thus was formal textbook econometrics born. In a sense, Haavelmo's approach was what had been effectively been appealed to by older econometricians, but Haavelmo solidified it. The ensuing Methodenstreit in 1947-9 between the Cowles Commission (led by Koopmans) and the old NBER Institutionalists (represented by Rutlege Vining) that established the "probabilistic" version of the theoretical approach as the dominant one in empirical economics.
The econometric boom was on. The work of the Cowles Commission led to development of estimation techniques for systems of simultaneous equations, such as the Klein-Goldberger model of Keynesian macroeconomics. However, to do so necessitated dealing with the "identification" problem which had so bedeviled the early work of Moore and Schultz. Effective solutions to the identification problem were worked on in the early Cowles years.
The Klein-Goldberger Model of the macroeconomy complemented other large-scale structural models designed for empirical work and government planning - such as Leontief's input-output system. Modigliani and associates developed another large scale macro-model, the so-called MPS model (after the initials of the institutions which helped develop it, M.I.T., Penn and the Social Science Research Council (SSRC)). These models existed not so much to be tested as to help policy-makers.
Yet not all was fine on the Cowles front. Herman Wold criticized the simultaneous techniques of these large models, arguing instead for recursive or at least block-recursive systems, which would require more careful analysis of time-series techniques. Richard Stone, an older style of data analyzer who conducted several pioneering empirical studies of demand, was similarly critical, as was Oskar Morgenstern. A more wide ranging criticism was expressed by Lord Robbins, where he denounced the empirical verification of theories on methodological grounds - following the footsteps of the Austrians. Finally, the most famous criticism was unveiled by Robert Lucas (1976), wherein he argued that the structural parameters of large scale models, particularly when used for policy policies, were assumed constant and thus did not comply with the propositions of what was then becoming the new mainstream macroeconomic theory. In a series of other papers, Lucas then outlined a new econometric methodology: time series macroeconometrics.
Early Microeconometrics: Market-Level Measurement and Estimation
Early Macroeconometrics: Business Cycle Measurement and National Income Accounts
The Large-Scale Models
Time Series Macroeconometrics
Resources on Econometrics
All rights reserved, Gonçalo L. Fonseca