James Heckman is the Henry
Schultz Distinguished Service Professor of Economics at the
Heckman's major contributions have been in the field of micro-econometrics,
where he has developed theory and methods widely used in the statistical
analysis of individual and household behavior. Heckman has developed methods to
deal with the "selective samples" that characterize most micro data.
Because samples of individuals (say, students attending college, married women
in the labor force, those participating in job training programs) are not determined
randomly and there are differences across individuals not observed by the
researcher, it is difficult to obtain unbiased measures of statistical
relationships. Heckman's methods address such concerns and permit one to better
infer causality in the social sciences.
The Nobel Committee recognized Professor Heckman, jointly with Daniel McFadden,
for their work in the field of microeconometrics, where each of the laureates
has developed theory and methods that are widely used in the statistical analysis
of individual and household behavior, within economics as well as other social
sciences. Professor Heckman was cited for "his development of theory and
methods for analyzing selective samples."
Quotes from James Heckman’s March 2003 lecture at
As a young child, I fervently
embraced the faith of my parents and by age eight, I was a child minister
giving sermons on Sunday evenings introduced by the motto taken from the New
Testament ‘A little child shall lead them.’ It was expected that when I became
an adult, I would be a minister…My deep religious belief led me to read and
study the Bible. I committed vast passages to memory…By my early teenage
years…I began having doubts...I began to notice inconsistencies in the Bible,
and as I read more broadly in science and philosophy, I came to question the
literal interpretations of creation and aspects of the fundamentalist
Christianity with which I was raised...I found that I was unable to accept
authority qua authority, a trait that has characterized me ever since…At the
same time, I learned early to live off my own intellectual and emotional
resources…Anchoring my own research in serious empirical analysis has protected
me from the fads that plague the profession and has given me the autonomy and
independence I have always craved…This is my way of following my father’s
advice of being independent and developing an independent reputation, and my
way of avoiding appeals to authority in any form – whether mathematical or
divine – that motivated my early break with organized religion.
That [high school physics]
class and my exposure to [Frank Oppenheimer] were truly life-transforming
experiences. He was the first truly brilliant mind I had ever encountered. He
introduced me to the life of the mind. Our class read everything he told us to
read in science and other subjects. We rushed out to read the books and
articles he casually mentioned while he conducted physics experiments with us.
We learned to love the late quartets of Beethoven at his house, and we
discussed the philosophy of science into the night…The profound simplicity of
the physical world and the ability of mathematics to predict it overwhelmed me.
If I had any religion at that time, it was Deism…The possibilities of science
seemed limitless.
My choice of
I have closely
followed the social experiment movement ever since those days. I am impressed
by the blind enthusiasm the term “experiment” evokes and by the serious
limitations of social experiments. The more carefully one looks at the product
of social experiments, the less enthusiastic one can be about them. Human beings
are purposeful in a way seed corn and laboratory rats are not. Economic policy
design requires that analysts know more than the dose-response relationships
produced by experiments. Years later I wrote on social experiments. What
impressed me then was how little had been learned from the first wave of social
experiments and how powerful and evocative the term “experiment” is to both
popular and professional minds.
In the mid ‘80s, in the course of my
research with Richard Robb on the nonparametric identification of selection
models and causal models, I was invited to an interdisciplinary conference on
inference from self-selected samples. Robb and I presented a paper before an
audience that consisted largely of statisticians…We discussed the fundamental
identification problem that arises from self selection when evaluating the
impact of “treatment” on outcomes…We presented the fundamental causal model
that economists have been using since the work of Haavelmo in the 1940s, and we
showed how alternative assumptions about agent program participation rules,
outcome equations, distributions of unobservables and access to different types
of data solved the problem of causal inference. We were stunned by the
overwhelming negative reaction to our paper and especially to our main point
about the conditional nature of causal knowledge. We made what we thought was
the obvious point that in the absence of an idealized experiment, answers to
causal questions required assumptions…We were not happy with this intrinsic ambiguity
in causal models, but we felt it was necessary to be intellectually honest
about it.
Although I have had
many brilliant colleagues, and I have benefited greatly from them, it is to the
students of
They were and are a
major stimulus to my own work.
The larger lesson I
took from this research [on race] was that one had to use all of the available
information to analyze causal questions. The available information does not
necessarily come in the form of sampling distributions from government surveys
or bounds on the ranges of random variables. Knowing the problem being studied
and its context, which in this case meant reading the relevant history and
newspapers, reading court and legislative records, using the relevant theory
and looking at all of the available data were essential in reaching any causal
conclusion. There is no algorithm to crank out convincing evidence, although
econometrics textbooks are often written as if there is. This lesson forever
impressed on me the limitations of standard econometrics, the even greater limitations
of the universal context-free nostrums for finding causal relationships offered
up by statisticians and the need to understand the relevant economic models and
contextual information in establishing causal relationships.
Additional
resources on James Heckman are available at the Nobel web
site.