“Econometrics” was little more to me than a word I occasionally misused when a copy of The Practice of Econometrics: Classic and Contemporary, by Ernst Berndt, arrived in the mail in 1993. I had read, I suppose, Ragnar Frisch’s forward to the inaugural issue of Econometrica, in 1933. After all, Frisch, a Norwegian, and Jan Tinbergen, a Hollander, had been the first economists recognized by the new Nobel Memorial Prize in Economic Sciences, in 1969.
Experience has shown that each of three viewpoint, statistics, economic theory, and mathematics, is a necessary, but not by itself sufficient, condition for a real understanding of the quantitative relations in modern economic life. It is the unification of all three that is powerful. And it is this unification that is econometrics.
I knew a little, too, of the snooty Econometric Society, with its unusual distinction between priestly fellows, often designated early in their careers, and more numerous ordinary members. But it was only when I skimmed the beginning of Berndt’s unorthodox text that the magnitude of the change began to come clear.
For one thing, the book offered an informal history of developments in the first fifty years of the field, in short profiles of major contributors scattered throughout its 700 pages. Several of these figures have been recognized since by the Nobel authorities, others have died, and a few are still waiting in the wings.
For another, the book seemed manageable, at least by those who planned to work their way through. It explored just ten particular forecasting tools of interest to economists and business students, each of them illustrated by application to a particular problem, all amplified by problem sets. First came bivariate regressions (the Capital Asset Pricing Model); followed by multivariate regressions and omitted variables (learning curves and wage determination); dummy variables and specification issues (quality change and wage discrimination); generalized least squares and distributed lags in time series analysis (explaining aggregate investment decisions, forecasting electricity demand); causality and simultaneity (relationships among advertising and sales); systems of equations (the case of small macro models); and discrete dependent variable models (whether and how much women work for pay). I was a newspaperman. I never tried any of it, goodness knows, but, for those who undertook the hard work, the book seemed likely to deliver on Robert Solow’s promise:
Studying this book and working through the exercises could do for economics students what interning in a big city hospital does for medical students – give them a tasted of what practice is really like and get them used to the sight of blood. This is a wonderful education.
Most of all, The Practice of Econometrics brought home to me the centrality of computers to the changes that were sweeping economics in the early 1990s. The back cover of the book contained a floppy disk, practically the first I had ever seen. (I had grown up in the era of punch-cards trays.) In the very first paragraph, Berndt assured students that they could still do their homework on familiar mainframe computers (if they has access to such expensive machines), but that his fundamental purpose was to introduce them to the possibilities of inexpensive “microcomputers,” such as the IBM-PC or the McIntosh, and to the coded instructions that made them perform, known increasingly as “software.” During World War II, a supervisor and two deputy human “computers,” working with adding machines, had required more than to two months to solve an eight-equation model; the new digital machines could perform far more complicated tasks in seconds.
Thirty years ago, that was news to me. Since then I have followed developments, hit and miss, through various histories: in statistics, with Theodore Porter, Stephen Stigler, Gregory Zuckerman, and Sharon Bertsch McGrayne; in computers and software, with Tracy Kidder, Paul Carroll, and G. Pascal Zachary; in the Internet, with Jennet Conant, Mitchell Waldrop, and Shane Greenstein; and in econometrics, with Mary Morgan, Charles Manski. Richard Murnane & John Willett, and Joshua Angrist & Jörn-Steffen Pischke. For the reasons that I mention here, Berndt’s text is still telling a few hundred copies a year to beginning students.
There have been remarkable developments in econometrics since 1993. Mostly they have had to do with thinking seriously about research strategies made possible by the staggering advances in computing that have taken place. The year’s Nobel Memorial Prize in Economics, announced in October, calls attention to role that natural experiments can play in reaching conclusions about various causes and effects. When Berndt published his text, this year’s laureates – David Card, Joshua Angrist and Guido Imbens – had barely begun the work that would ultimately earn them their recognition. You can attend their Wednesday lectures in Stockholm here on the Web to get some idea of what’s happened since, or, at your leisure, read “The Credibility Revolution in Empirical Economics,” from the Journal of Economic Perspectives, by Angrist and Pishke.
Applied economics – not just econometrics, but that which is described as “empirical” or of “practical relevance” – is sometimes said to have begun to produce more consensus, as opposed to an earlier “age of theory,” in which theorists blithely talked past each other and gave short shrift to data and testing. I’m not convinced. For the moment these claims seem more like a marketing device than significant value creation. I don’t doubt for a minute that the recent advances in econometrics are real. I do doubt that they add up yet to an “age.”