Nobel Prize 2000
The statistical techniques of measurement, for which Heckman and McFadden have been awarded the prize, may seem esoteric to a non-specialist. Yet, they are an indispensable part of the tool-kit of an empirical economist.
Economic Times, October 18 2000.
Heckman and McFadden: Revolutionaries in Empirical Microeconomics
If you wanted to predict the Nobel Prize winners, your best bet would be to look among the winners of the John Bates Clark Medal. Starting with Paul Samuelson in 1947, this medal has been awarded every other year (except 1953) to the most outstanding American economist under forty. Of the first seven winners of the medal, six—Samuelson, Milton Friedman, James Tobin, Kenneth Arrow, Lawrence Klein and Robert Solow—became early recipients of the Nobel Prize.
James Heckman of the University of Chicago and Daniel McFadden of the University of California, Berkeley, the winners of this year’s Nobel Prize, are both recipients of the Clark Medal: Heckman in 1983 and McFadden in 1975. Thus, there is little surprise in the fact that they have been awarded the coveted prize. The surprise lies, instead, in the fact that by opting for these younger applied econometricians, the Royal Swedish Academy of Sciences has passed on two other econometricians who have been regarded worthy claimants of the prize: Marc Nerlov of the University of Maryland and Dale Jorgenson of Harvard University, who won the Clark Medal in 1969 and 1971, respectively.
The statistical techniques of measurement, for which Heckman and McFadden have been awarded the prize, may seem esoteric to a non-specialist. Yet, they are an indispensable part of the tool-kit of an empirical economist. The two economists have concentrated their talents on studying the behavior of micro units such as individual, households and firms, using large micro data sets. The common thread running through their work is the use of microeconomic theory to develop econometric techniques that allow researchers to measure the effect of one or more variables on individual behavior when data sets may be biased in certain respects or are missing information on some of the relevant variables.
Thus, consider a researcher who wants to quantify the effect of education on wages and collects data on the wages received by workers and the years of education received. Assume, however, that there is a threshold level of wage below which workers do not accept employment. Then, among the workers with low levels of education, those offered low wages will not accept employment. Since data collection is limited to the workers who are employed, these workers will not be represented in the sample.
Put differently, among the less educated, only those workers lucky enough to be offered high-wage jobs will be represented in the sample. This “selection bias” will lead the researcher to underestimate the effect of education on wages: both less educated workers and the more educated ones would appear to receive high wages.
Using microeconomic theory, Heckman devised a statistical technique in the mid-seventies, called the Heckman correction, which allows the investigator to correct for this bias. Selection-bias problems are endemic to applied microeconomic problems, which make Heckman’s original technique and its subsequent refinements by himself and others indispensable to applied econometricians.
McFadden tackled similar problems but with a major difference: he devised methods for estimating relationships when choices are discrete. Thus, the traditional microeconomic theory is confined to the problems that involve continuous choices: how much of a good to consume or how much of a good to produce. But many choices are discrete. For example, individuals may be faced with the problem of choosing among a limited number of modes of transportation: car, bus and subway. Their choices may depend on the travel time and cost associated with each mode and individual characteristics such as age, sex, income and education.
Often a researcher has the information on only a subset of the relevant characteristics. The issue then is how best to predict the individual’s choice. McFadden developed the so-called conditional logit model in 1974 that enables the researcher to accomplish this task. The model and its refinements have been routinely applied in studies of urban travel demand.
Professor Orley Ashenfelter of Princeton University, where Heckman received his Ph.D., has an interesting story to tell about the new Nobel Laureate. The late Albert Rees, one of Heckman's teachers at Princeton, once remarked that Heckman’s was the best graduate class in labor economics he had ever taught or ever would teach. “When Al was asked how he could know that it was the best class ever, he explained, with a twinkle in his eye,” recalls Ashenfelter, “that Heckman was the only student in the class in that year!”