The Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel 2000
James J. Heckman, Daniel L. McFadden
Transcript of an interview with the 2000 Laureates in Economics, James J. Heckman and Daniel L. McFadden on 13 December 2000. Interviewer is Professor Bertil Holmlund, Uppsala University.
My name is Bertil Holmlund. I am interviewing the two Laureates in Economic Sciences this year: Professor Daniel McFadden from University of California at the Berkley and Professor James Heckman from University of Chicago. They have received the Prize for their contributions to micro econometrics, which is a field on the boundary between economics and statistics. This interview takes place at the Nobel Foundation in Stockholm on December 13, a couple of days after the Laureates have received the Prize. I think I should like to start by asking you how it all began. How come that you started to study economics seriously? Was it more or less by accident or was it something that you had wanted to do for a long time?
James Heckman: In my case it probably is an accident. I went to a liberal arts college and as part of my background I was majoring in mathematics and physics. But a part of the liberal arts college there's an option to take readings courses. Working one and one or in small groups. And out of curiosity I chose a readings class in economics where the classics were taught. Allegedly the theme was economic development. but it was reading people like Marshall, Ricardo, Smith and some of the more modern people, Arthur Lewis. I had always had a deep interest in social science, history. So even when I was in high school I was debating, and in college debating, and interested in contemporary events. And I had some interest in mathematics. And I was amazed. And then probably the most single event, the important event, was in the same class. The instructor gave me a copy of Samuelson's Foundations which I found to be an amazingly nice synthesis of things. It was an accident.
Also a social interest?
James Heckman: Yes. I found it interesting because I could then pursue several of my interests simultaneously. I was interested in quantitative analysis, but I also thought economists were asking interesting questions.
What about you?
Daniel McFadden: Well, I came from the natural sciences, from physics. I was a graduate student in physics and found psychology very interesting, and the psychology of behaviour. I had an opportunity to enter this disciplinary programme in the behavioural sciences. And I pursued that for my PhD and got into economics pretty much by accident, because the formal modelling, the axiomatic work that I was interested in was being done primarily by two economists, John Chipman and Leo Hurwicz. So I made economics my speciality in order to work with them.
Economists typically analyse occupational choice affected by income prospects. Did your choice of occupation to some extent reflect comparisons of incomes across occupations?
Daniel McFadden: For myself personally I would say no. I expected to be a poor academician all my life. And that was my choice. Non pecuniary returns.
What about you Jim?
James Heckman: I was going to say non pecuniary returns are what drove it. It was a lifestyle choice more than a choice of money. I never expected or particularly sought after financial resources and it was an accident. I always felt I earned a huge rent. I shouldn't say that on the television, but in the sense that I enjoy very much what I do. And the occupational choice is probably driven more by the psychic than ...
You have both a background in mathematics and physics and perhaps physics is often regarded as an ideal for research. Also economics. Some critics may say that perhaps at the same time economics become more and more like a branch of applied mathematics, how do you react to that criticism of economics?
Daniel McFadden: I think it's natural in the development of economics to quantify it. That is to say to move from general theories about how economic systems respond to numerical predictions on what will happen if you change some economic variable. And so mathematics is a natural language for developing a quantitative version of economic theory. Mathematics for its own sake sometimes may have insufficient contact with the facts because in the end a science has to be an interplay between the logical development of the theory and the reality of the facts.
What about you Jim?
James Heckman: I think the example, the contrast between physics and mathematics is a good one, because in some sense physics is driven by mathematics. Mathematics is very useful there. But it always orients itself towards data. Some sort of empirical phenomena or maybe not string theory but in the traditional physics. And in some sense I think that is a major model. It certainly was a major model for me. And I think it's not a bad model for economics generally to think of using mathematical models but to try to explain some aspect of reality, in this case social reality rather than a physical reality.
What about econometrics and its role to help us to choose between different models or theories? Ideally statistical testing should be a way of discriminating between alternative theories to weed out the weak theories. Do you think econometrics has been reasonably successful in this regard?
Daniel McFadden: I certainly think there has been great progress in the last few decades in doing real windowing out of hypothesis about behaviour. I think you see it in the kinds of things that Jim has looked at. Job, the effectiveness of various alternatives for job training. You see it a great deal in labour market and public finance applications. And there I think econometrics has been very successful. If you ask is it successful in modifying the deeper theory of economics I think the answer is less so. That's partly because that theory is more closely held by economists and they're less willing to change. Perhaps also because the deepest theory of economics is often viewed as a kind of a parable for how people should behave rather than something that is predictive act by act. So when evidence appears that seems to violate one particular act people will say it's not a failure of theory. It's a failure of interpretation or a failure of approximation.
James Heckman: I think econometrics has had a very large role, but I think probably one of the biggest developments, it's been around since the beginning of econometrics. But one of the most important developments has been understanding when we can use data to sort out hypotheses. You know identification questions. And when we cannot. When essentially the matters given available data do not allow a decisive resolution of the issue. I think economists have made a lot of progress on this question in the last 20 or 30 years. So we understand that some matters can't be settled with the available data and that stimulates the collection of new data where that might be settled. So I think econometrics plays a huge role in thinking through the issues very clearly. So testing is like an event, an activity which is very useful, but I think considering identification and considering what we can in principle separate from what we can't is extremely useful and is productive in many areas. Observational equivalents is what the macro economists call it. And identification is what the micro economists call it. But it's the same idea. Same pressures.
Do you think we should have much more of replication of existing empirical studies to check how robust the results are? Replications of econometrics studies to a much more systematic extent than we have seen so far.
James Heckman: I'm thinking in particular of a book that you may know of edited by Mary Morgan and Magnus and others, where there was an attempt to replicate Tobin's study of the demand for food. You know this study? No. Ed Lamar and others participated in this. And it was a little disturbing. Certainly the act of replication was very important but it seemed there were many other judgements that were brought in. And it led people, who were participating in this study of replication, to become aware of something. They started talking about the art of econometrics or the practise. The term, I'm forgetting right now, but there was a term about all the implicit assumptions, tacit econometrics I think was the term. So I think replication is extremely important precisely because in the past to any number there are a lot of other, quote enabling auxiliary assumptions.
Do you want to comment on this?
Daniel McFadden: I certainly think that economics will progress as replication becomes more important. And I would criticise the way our journals currently operate. They tend to always look for things which involve some new technique or some completely new data or some completely new idea. There's probably insufficient value placed on good work which verifies and checks things that have been done before.
On the same topic I think of Wassily Leontief. He's actually a former Nobel Laureate in economics and he has drawn the attention to the dominance of theory in economics journals. It seems as if the fraction of pure theory papers without any data is much higher in economics than in say physics or chemistry. Do you think that this is a problem? That it seems as if theory is given much more emphasis in economics than perhaps it is in other sciences.
James Heckman: I would make the remark that his remark was written in 1972. And it may have been more of a problem in 1972 or 74. I think it was his presidential address for the American Economic Association. It was certainly much earlier. I think there's been a huge growth in the last 20 years of applied economics. And empirical economics. And if anything one might criticise the other way. That there's been a huge amount of empiricism without any theory which I think may in fact be equally harmful and possibly more harmful for reasons we talked about earlier. So I'd be less worried now than I think I would have been, or less sympathetic with that comment, than I might have been 20 years ago.
Daniel McFadden: I've actually encouraged that traditionally empirical econometric work is hard to present within the bounds of a journal article simply because so much background documentation is required to fully explain what you do empirically. And I think with the development of electronic journals and the possibility of referencing things in a way which is available to readers, to users. Empirical presentation of results should get better and should begin to have more sway I think because you can present the results and provide at the same time the adequate background for those who want to fill in the gaps.
Empirical research requires data and good quality data and to compile data takes time. Do you think that the collection in improving data is valued as much as it should in the profession? Jim.
James Heckman: In some quarters yes. If you could point for example study done by Truman Bewley a few years ago. Truman Bewley is a very first rate mathematical economist who suddenly developed a mid career, I wouldn't want to call it a mid career crisis but it was certainly a mid career development, where he thought it would be very useful to interview firms about their wage setting policies. I think the profession has actually become more and more appreciative of data. I mean we've had large efforts across the board. Many fields. I think it's much more common now for individuals to collect their own data and to encourage the collection of data. I think 20 or 30 years ago it used to be the case that economists and sociologists were completely opposite ends of the spectrum. That only very few economists were collecting data. And now I think there's a very, very active group of many economists in many fields collecting their own data whether through experiments or through the secondary collection of data. So I think it's a major activity.
Daniel McFadden: I agree with Jim. My recommendation to young people though is get tenure first and then develop large data sessions ...
There is another way of testing, at least a complimentary way, and that is experiments and economics is becoming more and more an experimental science. Do you think that this development is all for the good or are there some drawbacks here? Dan.
Daniel McFadden: I'm very enthusiastic about the opportunities that experiments offer for understanding economic behaviour. And, and for that matter understanding more about economics including the organisation of markets and alternative market forms. So I'm very encouraged. The experimentalists and the econometricians could benefit from talking more to each other because right now the experimentalists have even more than the econometricians difficulty presenting their results in a form which is as concise and informative as a theorem that the theorists can do. And so I think a great opportunity is for the possibility of using econometrics to distil and extract the essence of experimental results.
Jim, you have voiced some scepticism towards the use of social experiments or at least argued that these experiments have their limitations as a source of knowledge for example to evaluate the effects of programmes.
James Heckman: Yes, and I think the kind of experiment that Dan's talking about and the social experiment I think maybe different. And I think one has greater control of the laboratory experiment of the Vernon Smith variety for example. In the social experimental context I've been worried, and have written some papers on this, where I've seen misuse of experiments. The danger in a lot of empirical work is that sloganeering takes over. Like it's true everywhere, I suppose. But in the context of social experiment the very name experiment seems to bring up the image of science, Bunsen burners and truth. Whereas in fact there are serious compliance problems, attrition problems, and those can substantially degrade the inference from an experiment. So that the kind of careful analysis from an experiment that has to be done at the end of the day starts to resemble very much the kind of analysis that comes from a non experimental study. It's an additional source of variation.
I had a paper published in the May issue of the Quarterly Journal of Economics, the May 2000 issue. Where we saw that a social experiment which I was engaged had on a naïve basis led to the notion that job training, all types of job training, were extremely poor. However when you looked at crossover and attrition problems in that study there was a major reworking of the evidence. So very simple corrections. Non-experimental corrections to the experiment caused a rethinking and I think a reshaping conclusion. So it's a delicate tool. It's a valuable source of information. But unfortunately in policy circles it's viewed as somehow a panacea. You know again any slogan whether it's multiple regression or [INAUDIBLE] or a selection bias correction, any un-critical notion econometric tool is taken out of context can be a very dangerous tool and misapply.
Your works have been highly relevant for policy making. Do you see cases where your studies have had some impact on actual policy decisions? Dan.
Daniel McFadden: I think certainly in the area of labour force participation both our works have had major impacts. The question of whether people enter the labour force or not. If they become unemployed how long they stay unemployed. I think our ability to understand the incentives that influence that and how people respond to those incentives have greatly improved.
Can you point to some specific instances? Where do you believe that your studies made a difference in terms of policy decisions?
James Heckman: I would give some examples. Just for the work in job training. Both here and in other countries has had a direct impact in the sense of advice for governments. I was actually involved in the job training partnership act experiment and non experimental study. And as a result of that study, this was a large scale manpower training programme in the US, the programme was fundamentally altered. Now unfortunately I wouldn't say it was my particular study alone that cost the body of work. There was a group of associates. But there are specific examples. There is now an examination under way about what's called the GED programme. Exam Certification Programmes. Partly based on this work. So I think there has been. Dan's being modest but I would say the area of rapid transit study has had huge impact on policy and that's a prototype for analysis of other policies.
This policy of economics also creates a greater demand for policy advice. Have you been personally involved in policy advice informal or formal?
Daniel McFadden: Personally I'm rather heavily involved in research related to health outcomes and the economics of health. And economic planning particularly for the elderly. And yes, I get requested quite regularly to try to understand what the impacts are going to be of changes say for example in the healthcare delivery system and how that will influence in turn people's behaviour both in terms of how they stay healthy and what they do and in terms of their economic planning.
What about you Jim?
James Heckman: Yes I've been involved even at a very micro scale. Some of my students and the Cook County Chicago area we've been involved in the design and analysis of very small scale job training programmes for concentrated poverty. But also national policy advice. Although I wouldn't consider myself a policy pundit in the sense of travelling and doing this frequently, more actually in Latin America recently. In the last four or five years through a variety of circumstances I have been giving advice. But only on broad scale issues. Trying to look at education training.
I would like to conclude by economics as a science with imperialistic ambitions towards the other sciences. Economics invade political science, sociology, demography and so on. Presumably economists find that natural or perhaps a good thing. What about influences in the other direction? Do you believe that we can learn something from say psychology or sociology? Dan?
Daniel McFadden: I believe we could and we must learn a great deal about behaviour from the people who are studying the individual more as an observational unit itself. That is to say economists tend to concentrate on the outer man. The person who goes to the store and buys products. But you can understand a lot more about behaviour by looking a little bit inside. The processes by which decisions are made. And I think we have a great deal to learn. Right now from the psychologists. Perhaps eventually from the people who are closer to physiology and medicine. The operation of the brain. How perceptions are formed physiologically.
What about you Jim?
James Heckman: I think certainly in my own case I would argue in the 70s a major stimulus to my own thinking was the work of the sociologist on panel data. I think people like Paul Lazarsfeld and his students like Jim Coleman actually were way ahead of economists at one time in analysing panel data. So in the 1970s I taught at a course with Jim Coleman for several years on panel data analysis and the flow was one way for a while. I think we caught up and went ahead of the sociologists. The demographers for years have been dealing with things that are called distributions of fecundicity and mortality, frailty it's now called. But there's been a huge amount of stimulus the other way. At least in technical statistical tools and also questions of social science. Issues of social interactions and social processes. I think the sociologists have been ahead of economists and have been stimulating to us and political science as well. Voting theory has actually been quite a productive source of ideas for economists.
Right. By this we should conclude. I thank you very much indeed. Professor McFadden. Thank you.
Daniel McFadden: Thank you.
Did you find any typos in this text? We would appreciate your assistance in identifying errors in the text.
E-mail to email@example.com. Thank you for taking the time to report the errors.