The Changing Face of Economics: Conversations with Cutting Edge Economists

10. Kenneth Arrow

chapter 10 Kenneth Arrow Professor (Emeritus) of Economics, Stanford University  The interview was conducted at Kenneth Arrow’s home in Stanford on August 10, 2003. Could you comment brie›y on your overall reaction to our project? Economics as practiced is enormously different in orientation and detail today than it was when I was a graduate student. I got into economics because I was interested in statistics, not because I was interested in economics. At the time Columbia was one of the few places you could study statistics. When I went to graduate school at Columbia University, the dominant views in economics there were totally antitheoretical and critical of received wisdom; the direction was highly empirical. The dominant ‹gure at Columbia was Wesley Mitchell, although few economists today would know of him. At Columbia in 1942 there was no course in price theory at all. Hotelling had a course in mathematical economics, but that was for an extremely small number of students. Instead the department had a course in the history of theory. Columbia wasn’t unusual; at the time institutionalists of one kind or another dominated a number of economics departments. My point in raising this history is that there is nothing new in the idea that theory is controversial.  291 The absence of a theory course led to one funny aspect. The faculty felt the need to examine the Ph.D. students in theory, but since there was no course in theory, doing so was dif‹cult. They had an ingenious solution: you went to a professor and negotiated a topic, and then you were examined on that topic. John Maurice Clark, the nearest thing to a theorist that Columbia had, examined me in theory. The hot topic of the day was imperfect competition. So something like game theory was already around back in the 1940s. It’s funny how things come around. Mitchell was on leave when I was there, and Arthur Burns replaced him. Burns was one of the smartest men I have ever known. Keynes wasn’t mentioned in Burns’s course, and this was 1942. At Harvard and Chicago students would have gotten more theory, but it wouldn’t have been a dogmatic competitive theory; for example, Chamberlin was teaching at Harvard, and he even had some experiments in his class. I had to learn neoclassical economics by my own reading. As far as my formal education in economics, I learned exactly zero about neoclassical economics in my courses. One of the arguments we make in the introduction is that it isn’t helpful to call the economics done today neoclassical. Would you agree with that? The ideas that people behave somewhat rationally and somewhat foresightedly are central to economics. But there are major differences in the knowledge attributed to agents. I think one of the biggest differences between 1950 and 2000 is the much greater role given now to the role of knowledge and information, and the theorizing about knowledge and information. Nobody would have denied the importance of these in 1950, but the tools to handle them formally did not exist then. Probability thinking was pretty much unknown to the average economists in 1950. The idea of some type of rational action is still at the core of economics . It can be expanded in many ways. Similarly, the idea of competition remains important. The difference is that the kind of person who would have been a rigid competitive equilibrium person now writes game theory papers. I think there is still a lot on the agenda of applying game theory to economics. The big issue that needs to be cleared up in game theory is summed up in the word general. The question is how to generalize all the noncompetitive elements in the world into a theory. the changing face of economics 292 Game theory tends to be partial equilibrium, and somehow it must be generalized since the effects in the general model may be quite different than in the partial model. A game in one part of the economy—say a patent race—likely in›uences the rest of the economy . In general equilibrium you can trace out the links, and we aren’t close to being able to do that in game theory yet. These links are important because sometimes the main effect of a change can be quite separate from the particular area you are looking at. I...