Monday, October 15, 2007

It’s a Small World and the Nobel Prize in Economics

This morning they announced the Nobel Prize winners in Economics. One of the winners, Roger Myerson, was a professor of mine at Northwestern (Winter quarter, 1977) for a course in non-linear programming. That was my first year in grad school and I believe Roger’s first quarter teaching. Before that, Roger was Ken Arrow’s student at Harvard. Arrow was another Nobel winner. During that first year we learned about all the giants of economic theory. Arrow and Gerard Debreu were at the very top for their formulation of the general equilibrium model. (Lionel McKenzie was also a pioneer in this area but didn’t get quite the same notice.) The third great theorist we became aware of at the time was Leo Hurwicz, for mechanism design. It’s remarkable that he didn’t win the prize until now. He was a giant even then. The funny thing is that Hurwicz was once a faculty member at Illinois, before I was born. So was Franco Modigliani, another Nobel winner. Back then we let some very big fish get away.

Let me get back to Roger. Reading the New York Times piece and this piece in the Chicago Tribune, which is a little more informative, it seems these papers don’t think they can educate their readers as to what “mechanism design” really means so they talk about it but don’t explain it. I’m going to try to provide some explanation with an example.

Before the Internet and Websites like Travelocity or Expedia, one had to book flights either through the airlines themselves or through a travel agent, who would do the lookup on your behalf to find the appropriate flights at the appropriate fares. In turn, the airlines set the fares in kind of an arcane way but with a particular goal in mind – they wanted to separate out the buyers by their willingness to pay, and get those willing to pay more to do that, in exchange for some perq that these buyers would value. Equivalently, they wanted to give discounts to those buyers who were looking for a bargain, but they needed to provide some inconvenience in addition or else all buyers would want to buy at the lower price.

One mechanism that they opted for, a mechanism that persisted for quite some time, was the Saturday stay over. “Business travelers”, who had their travel paid by their employer, wanted to get home once their business concluded. But “tourists”, looking for bargains, would be willing to accept the Saturday stay over to get the lower fairs and accommodate their vacations to that.

In 1979 Roger wrote a paper where he introduced an idea called the “Revelation Principle” (I believe the same idea was later invented independently by Robert Townsend) where Roger showed that for every mechanism (Saturday stay over is one such mechanism) there is an equivalent “direct mechanism” that leads to exactly the same outcomes. In this case in the direct mechanism the traveler announces either “I’m a business traveler” or “I’m a tourist” and then the mechanism allocates both the travel time and the payments in a way that makes them “incentive compatible,” meaning it is in the interest of each type to report their type truthfully given the incentives in place.

The value of this finding was huge for economics research. For an arbitrary mechanism it is sometimes quite hard to work through the full implications of the incentive scheme. But for a direct mechanism where it is known that each type will report truthfully, it is much easier to design an optimal mechanism. During the 1980s and early 1990s, many economics papers (including a few I wrote) looked for the optimal direct mechanism within a certain particular environment that was deemed worthy of study, as a way to explain behavior in that particular environment. The approach was adopted in Industrial Organization, Labor Economics, International Trade, and other fields as well. Roger’s work paved the way for all this other work.

I’m glad he won the award. It makes me feel proud of the graduate education I received and that the whole thing really is close to home.

No comments: