I’ve continued mulling over the question of what a course in Market Design should cover and decided to take a stab at writing out at least what the beginnings of such a course would cover. Suggestions, additions, deletions and brickbats are welcome. First, a pretentious title: Principles of Market Design.
What should one assume in the way of prior knowledge? Some convexity, Linear Programming, Mas-Collel-Whinston & Green and a course in Information Economics.
Begin with Coase’s theorem. Once property rights are defined, assigned and transaction costs of bargaining are eliminated, resources should be allocated efficiently. The theorem prompts three questions:
1) Does it matter how property rights are defined?
Yes, spectrum is an example. Current property rights for spectrum were shaped by the nature of technologies present half a century ago. This in turn has influenced how we think about allocating spectrum. With current technologies we might have shaped these rights differently and the market for spectrum might be very different. Another example might revolve around the duration for permits for emissions of various noxious gases.
2) Does it matter how the property rights are assigned?
In Coase, no. But, the initial assignment does have distributional consequences for the parties concerned.
3) What are the transaction costs that hinder efficient bargaining?
Presumably the market designer has a role to play precisely when these frictions are an obstacle.
Where next? The Myerson-Satterthwiate bargaining set up. One can use it (and variations) to make a couple of points:
a) Private information as a `transaction’ cost that impedes efficiency.
b) The Cramton-Gibbons-Klemperer variation reveals that efficiency could be recovered for a different initial allocation of property rights, i.e., the initial allocation of property rights does matter.
c) The Rustichini, Satterthwiate and Williams variation shows that as we increase the size of the market, we can get closer to efficiency. McAfee’s trick of dropping the last trade in Vickrey is also worth a mention.
The set up in (c) is, in a sense, what one might mean by a thick market. So, it is worth exploring the assumptions embedded in it. They are
1) Private values rather than common values. This would feed into adverse selection and lemons. Large markets with interdependent values…..Swinkels? Perry & Reny?
2) Co-location of buyers and sellers. Buyers and sellers know where to show up in the model. Relax that, and one has a coordination problem. A central exchange solves that problem.
3) Synchronicity. Buyers and sellers are all present at the same time. Suppose buyers and sellers arrive dynamically and can `wait’ for trading opportunities? A basic search model (which one?) would allow one to show how efficiency of trade is diminished.
4) Trade is only permitted via the exchange. What if agents can peel off and engage in transactions outside the exchange? An opportunity to talk about the core. Related to synchrony above, a desire to move at a different clock speed, i.e., unraveling. This would be related to common values above. I suppose one should connect to (2) and perhaps say something about competition between platforms.
5) Unit demand. What would happen if sellers had multiple units to offer and buyers wanted to consume more than one unit? This could lead to a discussion of the role of uniform pricing and demand/supply reduction.
6) Homogenous goods. The model assumes that sellers sell the same good. What about heterogenous goods? This could tee off a discussion of the Shaply-Shubik assignment model, substitutes preferences and discrete convexity.
7) Quasi-linear preferences. But, keep the money. What should one say here? General equilibrium? The generalizations of the assignment model due to Kaneko and Quinzii?
8) Money. One might speculate on where it comes from. Examine what can be achieved without the use of transfers.
Pausing for breath, I see I haven’t even got to distributional concerns yet or applications!