December 1989

Thomas J. Sargent

Over the past 18 years, the Minneapolis Fed has been privileged to count Thomas J. Sargent among its staff of economic advisers. With six books and nearly 100 articles to his credit, Sargent is well-known and highly regarded by his colleagues in the economics community. He has received many honors for his work including election to both the National Academy of Sciences and the American Academy of Art and Sciences in 1983 at the age of 39.

Currently a senior fellow at the Hoover Institution, Stanford University, Sargent began his relationship with the Minneapolis Fed when he joined the University of Minnesota faculty in 1971. He arrived with a B.A. degree from the University of California at Berkley and a Ph.D. from Harvard University.

During his tenure at the University of Minnesota, Sargent also lent his expertise to several universities and think tanks. Most notable, perhaps, is his stint as a visiting research professor at the University of Chicago, the home of Robert E. Lucas, a leading proponent of the rational expectations theory.

In the early '80s Sargent, with University of Minnesota associate and advisor to the Minneapolis Fed Neil Wallace, produced groundbreaking research and writing on rational expectations. Their collaboration created a stir in the economic community.

Sargent once likened the rational expectations theory to the strategy pursued by the Minnesota Vikings; that is, a game plan is designed to counter the predicted strategies of the opposing team. In the following interview, Sargent talks about the theory that made him famous and his latest projects.

Region: In the 1970s, a controversial new theory called rational expectations swept through the economics profession. You, Robert Lucas and Neil Wallace were the primary architects of this new school of thought. What was the essence of that thinking?

Sargent: The basic idea is this. In lots of economic situations, we want to study people who have the problem of making decisions now based on things that they expect will happen in the future. An example would be a business firm that has to make investment plans now, say purchase and design a new plant now which isn't going to be producing output until the future. So the firm has to forecast the prices it's going to sell at in the future. It also has to forecast what competitive products are going to be around. It has to forecast the prices of the inputs it uses that are complementary with their capital, and so on. Thus, one of the big jobs of business is to forecast the future. Economists want to study what business does and to have some principle for understanding how business or consumers or any economic agent forecasts. Rational expectations assumes that people use all the information they have as well as they can. This is just an application of the economist's general perspective in analyzing people's choices. We assume that people do the best that they can, that they behave in their own interests.

Region: In the decade since the beginnings of these ideas, I'm sure there's been an evolution. Is that correct? How would you describe that evolution of rational expectations?

Sargent: In the '70s when we were doing work here at the bank, it was very controversial. At that time there were still plenty of defenders of the old Keynesian models that rational expectations really threatened in a very profound way. There was a lot of resistance to the notion that expectations are rational. Critics would say people can't possibly be as rational as you're saying they are. Nevertheless, what's happened in the '80s is that most younger economists use the rational expectations assumption. It's no longer controversial. It's used as a tool throughout macrocconomics and microeconomics.

Region: Part of mainstream thinking?

Sargent: It's part of mainstream thinking--but not as much fun. It was more fun when it was controversial.

Region: I found an old article where you were quoted as saying Keynesianism and monetarism, they're both gone, they're dead.

Sargent: They're dead in the following sense. There's a schism between research being done on the frontiers of macrocconomics and thought in policy circles in Washington, and it's actually somewhat alarming. This schism has never been so great. Most economic policymakers in Washington show little acquaintance with macroeconomic theory in general. It's very hard to tell if their thinking is guided by any kind of model, so those guys wouldn't be able to tell you that Keynesianism and monetarism are dead. But among the young economists doing research, which is really what's influential in the long run, Keynesian models in the style of the '60s are just not used. There are some people who view themselves as using rational expectations to resuscitate some ideas that they attribute to Keynes, but their ideas are very different than those of 1960s-style Keynesian models.

Region: What would rational expectations say about current monetary policy pursued by the Fed?

Sargent: One of the things that rational expectations says is that you have to look at budget constraints. Another thing that rational expectations says is that you should consider the behavior of both private agents and government. Private agents will, and therefore government policymakers should, be looking forward and consider the present ramifications of future actions. I'll give you an example of what I mean in the form of a problem that has been written on repeatedly by people at this bank since the very beginning of the '80s.

In the 1980s the United States has been running a very large fiscal deficit. Depending upon how you treat the Social Security system, the fiscal deficit can be viewed as unprecedentedly high or at least troublesome, and there are two consequences. If you look in the '80s, real government debt per capita has grown, whereas it shrunk during most of the postwar era. Furthermore, and even more significant, real interest payments on the government debt have grown steadily during the '80s. And not all of the debt is on the books.

The government has incurred implicit liabilities in the '80s. The savings and loan crisis and the bailout is one. And there are other bailouts that threaten to be on the horizon. A bailout is an implicit debt obligation of the government. That's all on the fiscal side, but it has an implication for monetary policy. Why? Because the government has a budget constraint. The essential job of the Fed from a macroeconomic point of view is to manage the government's portfolio of debts. That's all it does. It doesn't have the power to tax. The Fed is like a portfolio manager who manages a portfolio made up wholly of debts--it determines how much of its portfolio is in the form of money, which doesn't cost the government any interest, how much is in the form of T-bills and how much is in 30-year bonds. The Fed continually manages this portfolio. But it doesn't determine the size. A stream of deficits hands the Fed a larger and larger portfolio. As that portfolio of debts gets bigger, it constrains the ability of the Fed to fight inflation by creating pressures for it to create more money.

What you've seen throughout the world in the 20th century is that governments that have run very big deficits and given their monetary authorities big portfolios of debts to manage have sooner or later given way to the temptation to inflate a lot. One of the things that's remarkable about the '80s is the way that temptation has been building up in the United States. The fiscal deficit and the accumulation of debt creates a big pressure on the Fed to print a lot of money, to reduce the interest costs and to reduce the real value of the debt. A one percent increase in the price level helps the federal government's debt problem by writing down the real value of those debts. There's increasing pressure to do that.

Region: The Fed releases the policy actions of the Federal Open Market Committee (FOMC) a few days after the next meeting. Would you say that it would be more efficient if the FOMC publicly announced those policy actions right after its meeting instead of holding that information temporarily confidential?

Sargent: My prejudice is that it should. I've never seen a compelling reason that it should keep those things secret. In the absence of a compelling reason not to keep secret, it shouldn't--it's a public agency.

Region: Let me turn to the Hoover Institution. You're working there but you still maintain a relationship with the Federal Reserve Bank. Could you explain your relationship with the Minneapolis Fed while you're working at Hoover?

Sargent: When I was a professor at Minnesota I was a consultant and adviser at the Fed, and much of my research was based here. The Research Department here is one of the most distinguished in terms of macroeconomic theory and macro monetary policy. It's one of the two or three best places in the world. In terms of economists under 40, it's the very best in the world. My interest in coming here is to interact and do work with these people. A lot of the most important ideas in the monetary economics and macroeconomics have appeared in discussion papers at this bank.

Region: Art Rolnick, the bank's director of research, has been especially impressed with what he described to me as micro modeling. Could you explain what micro modeling is? It is something that you're working on at Hoover, as understand it. Is that correct?

Sargent: Yes. I'll tell you basic ideas. If you look in course catalogs in the 1920s or earlier, a subject called macroeconomics is not there. There was one economics, basically; it was micro, which means that you analyze individuals as doing the best they can, being motivated by purposes and subject to whatever constraints the market imposes on them. In the 1930s Keynes thought that this whole body of reasoning, in which agents are serving their own purposes as well as possible, probably couldn't explain things like recessions and depressions. So he wrote a book called The General Theory of Employment, Interest and Money. By "general theory," he meant one not based on the view that agents were doing the best they could and that markets worked in a way that reflected the coordination of those efforts. And in place of people doing the best they could, he substituted some other things that weren't based on optimization. And that was the start of macroeconomics. One of the first reviews of Keynes' book was written by an economist named Wassily Leontief. He deplored the fact that Keynes was departing from the older tradition of attributing rationality and optimality to people. Ever since then there's been a persistent call to put microeconomic foundations underneath macroeconomics. And if you try to build micro foundations, you're going to get back to the elements of classical economics.

These elements give us a set of methods for trying to study how people behave basically using Optimization Theory. Optimization Theory was something that was created by mathematicians, and that constitutes our main tool. The simplest example of a micro model is the supply and demand model for determining the price and quantity of a commodity produced, where demand is a reflection of preferences of people for various goods in their income level and supply is a reflection of the costs of production and the technology.

Region: What are you working on personally these days?

Sargent: A couple of things. I'm working on artificial intelligence, taking some models of macroeconomics and instead of saying that people are rational, we assume that they're artificially intelligent. I'm doing this with some colleagues, Ramon Marimon and Ellen McGrattan. What we've done is gone to the literature on artificial intelligence. We met a man named John Holland who is a distinguished computer scientist at the University of Michigan. He's developed models of the way people learn from the environment. So people start out not being rational, not understanding how the world works. They start out being very stupid. One of the hardest problems in artificial intelligence is to figure out how to make computers learn. People seem to learn much faster than computers can and much better. Holland has, from our point of view, discovered some of the most exciting algorithms for creating computer programs that will learn. Instead of assuming rationality, we endow our people with these little computer programs and then see what happens. And what's happened in a lot of situations is that they end up learning and being rational.

Region: Is this applied to the marketplace or is it more of a cybernetic exploration?

Sargent: Well, I've been having conversations with Robert Litterman, a former Minneapolis Fed economist who now works in Wall Street. He's using some of these artificial intelligence methods to solve practical problems, market problems. But we're using them basically to understand how systems of individuals can come to learn how to cope in a market. One of the criticisms that's been made of rational expectations is attributing way too much knowledge and information to people, and the point of this work is for us to learn how people learn. One of the things that literature on learning and artificial intelligence does is to buttress rational expectations. I'm also working with Lars Hansen of the University of Chicago on a book that makes it very easy to use a whole class of models that this bank has been instrumental in developing, namely linear rational expectations models. When you receive our book, you also get a computer disk, and what you can do is essentially make up your own model economy. The computer solves it and allows you to analyze it very quickly. One of the criticisms of some of the work that's been done at this bank is it's very technically hard, and it's been difficult for people to use. What we're trying to do is to make it much easier to use, to demystify it.

Region: We look forward to a less mystifying explanation of these new findings in economics. Thank you, Mr. Sargent.

--David Levy

CÚdric / QueST