In the comments to my previous post on the Nobel announcement yesterday, David Tufte asks for some extended comments on the Hurwicz, Maskin, Myerson Nobel decision.
My perspective shares quite a bit with those of Alex Tabarrok in Reason yesterday and Pete Boettke in the Wall Street Journal today (subs. required). Both Hurwicz and Myerson were on the NU faculty when I was a graduate student, although I did not take classes with either one.
I knew about the origins of the literature in Hurwicz’s early work, and I knew that it had been inspired by trying to “kick the tires”, if you will, of the Lange-Hayek socialist calculation debate. As Pete notes in the Journal today:
Leonid Hurwicz, in his classic papers “On the Concept and Possibility of Informational Decentralization” (1969), “On Informationally Decentralized Systems” (1972), and “The Design of Mechanisms for Resource Allocation” (1973), embraced Hayek’s challenge. He developed mechanism-design theory to test the logic of the Mises-Hayek contention that socialism could not possibly mobilize the dispersed knowledge in society in a way that would permit rational economic calculation for the alternative uses of scarce resources. Mises and Hayek argued that replacing the invisible hand of the market with the guided one of government would not work. Mr. Hurwicz wanted to see if they were right, and under what conditions one could say they were wrong.
For me, the value of mechanism design starts there: what are the conditions under which decentralized coordination is possible through some mechanism to aggregate private knowledge? Basic mechanism design models (such as, for example, a contract between a principal and an agent who will exert some effort on behalf of the principal) start from the idea that each individual will maximize his/her value function (utility, profit, whatever), subject to two important constraints — individual rationality (or participation) and incentive compatibility. The form those constraints take will depend on the specific nature of the relationship being modeled. Given that basis, is there a way to arrange the relationship such that both parties are made better off, and neither party has an incentive to choose a different action? See also Alex’s mechanism design post from yesterday.
Put another way, mechanism design is in a lot of ways about institutions. Can we devise an institution in which both parties do things that create the most net benefit? In that sense institutional design is a lot about figuring out what those individual rationality and incentive compatibility constraints are, and what institutions will satisfy them and lead to optimal net benefit.
That said, it’s also very abstract, mathematical theory. I think Tyler’s probably right that pure mechanism design is going out of favor (although not here at Northwestern!), and in part I think that’s because of the success that mechanism design has had in tackling important questions. It’s been fruitful for a few decades, but it’s been so fruitful that diminishing returns to the technique are probably kicking in. But it’s also true that, like Austrian economics, mechanism design has become part of many other fields and types of modeling approaches, even those that don’t call themselves mechanism design.
In short, I think mechanism design has the myriad pros and cons associated with lots of theoretical mathematical modeling done in economics. But by focusing on the conditions under which individual agents can achieve decentralized coordination, mechanism design has produced an extensive body of theoretical knowledge that has been applied in both empirical analyses and in public policy.