I began the first class with a discussion of the definition of market design. Here, as best I can, I recall my remarks.
The flippant answer is its whatever Alvin Roth says it is. A serious answer is the same, witness the following from Econometrica 2002:
…..design involves a responsibility for detail; this creates a need to deal with complications. Dealing with complications requires not only careful attention to the institutional details of a particular market, it also requires new tools, to supplement the traditional analytical toolbox of the theorist.
Two items to highlight: institutional details and and new tools. So, it is not `generic’ theorizing and one might wish to (should?) use experiments and data to support an analysis.
The NBER offers the following:
“Market design” examines the reasons why markets institutions fail and considers the properties of alternative mechanisms, in terms of efficiency, fairness, incentives, and complexity. Research on market design is influenced by ideas from industrial organization and microeconomic theory; it brings together theoretical, empirical, and experimental methods, with an aim of studying policy-relevant tradeoffs with practical consequences.
Notice the concern with market failure, the clearly normative perspective and the intent to influence policy. Finally, there is the Journal of Economic Literature which now recognizes it as a subfield (JEL classification D47) and offers this definition:
Covers studies concerning the design and evolution of economic institutions, the design of mechanisms for economic transactions (such as price determination and dispute resolution), and the interplay between a market’s size and complexity and the information available to agents in such an environment.
In this definition no methodological stand is taken nor is there an explicit concern for practical consequences. Two additional lines, that I do not reproduce, exclude straight mechanism design and straight empirical work.
Where does that leave us? Clearly, decision theory, repeated games, refinements are not market design. What about the study of tax policy to encourage investments in human capital? What about labor economics, in particular the thread related to search and matching? Regulating oligopoly and merger policy? By my reading they count. Are the labor economists all going to label their papers D47? Unlikely. We generally know we are writing for, so, perhaps the flippant answer I gave first is the correct one!
Do I have a proposal? Yes, Richard Whatley’s 1831 suggestion for what to call what we now dub economics.
Next, what principles, if any, are there to guide the market designer? Roth’s 2007 Hahn lecture suggests three:
By which he means encouraging coordination on a single venue for trade. Why? Presumably beause it reduces search costs and increases the speed of execution. One way (not the only) of formalizing the notion of thickness is stability or the core.
2) Reduce Congestion
Roth argues that as a market thickens, it produces congestion. This runs counter to the benefits of thickness: increasing speed of execution. What he has in mind are markets where transactions are heterogenous and offers are personalized. During the time in which an offer is being evaluated, other potential opportunities may evaporate. I’m not yet convinced that thickness produces congestion in this sense. I cannot see why the time to evaluate an offer and conclude a transaction should depend on the thickness of the market. However, in order to benefit from the increased opportunities that thickness provides, it makes sense that one would want to increase the speed at which offers are screened. I think the correct way to phrase this might be in terms of bottlenecks. As the market thickens steps in the trading process that were not bottlenecks might become so. Not removing them, defeats the gains to be had from thickness.
3) Discourage Welfare Reducing Strategic Behavior
Equivalently, make participation simple.
While these 3 items are useful rules of thumbs in thinking about the design goals, they do not cover the details the designer should pay attention to in achieving them. I’ve tried to list what I think these are below.
a) Carefully decide on the asset to be traded.
I’m channeling Coase. To illustrate, recall the use of auctions to allocate spectrum rights. A frequently heralded an example of successful market design. The use of auctions takes as given that the asset to be traded is rights to a range of frequencies. Those rights protect the holder from `harmful interference’. However, as Coase himself observed, an entirely different asset should be the object of analysis:
What does not seem to have been understood is that what is being allocated by the Federal Communications Commission, or, if there were a market, what would be sold, is the right to use a piece of equipment to transmit signals in particular way. Once the question is looked at in this way, it is unnecessary to think in terms of ownership of frequencies of the ether.
As one can imagine this might lead to a very different way of organizing the market for wireless communication. For an illustration of how different views on spectrum property rights affect market outcomes see the spirited piece on the Lightsquared debacle by Hazlett and Skorup. An important take away from this piece is the (constructive/destructive) role that government plays in markets.
b) Nature of contracts.
What kinds of contracts are feasible? Must they stipulate a uniform price? Can they be perpetual (indentured servitude is typically outlawed)? We know, for example, that Walrasian prices can implement certain outcomes but not others.
c) What is the medium of exchange?
In some settings we have money, in others it is ruled out. That money is ruled out does not eliminate other mediums of exchange. Prisoners have been known to use cigarettes. One can trade years of service for preferential postings. One can exchange kidneys for kidneys. What about livers for kidneys, or health insurance for livers?
d) What is the measure of performance and what is the status quo?
Typically one is concerned with proposing a set of changes to an existing institution. What is the measure by which we decide that a proposed change is a good one? I rope into this question the decision about which agents preferences matter in the design. In the literature on resident matching the focus is on stability (thickness). However, one might also focus on the effect on wages. Does the use of a stable mechanism depress wages for interns? If one focuses on wages, then one must specify what the status quo is. For example, is it one where wages are set by a perfectly competitive centralized market? Or is it an imperfectly competitive one where wages are set by bilateral contracting? How would one model this (see Bulow and Levin for an example)? Even if the status quo were a monopoly, is it regulated or unregulated?
In choosing a measure of performance one will bump up against the `universal interconnectedness of all things’. In ancient times this challenge was called `regulating the second best’. Imagine a polluting monopolist charging a uniform price. If we replace our monopolist by a perfectly competitive market we reduce the distortion caused by high prices but increase pollution (assuming it increases with output). To make headway one must be prepared to draw a boundary and ignore what happens outside of it.
e) Why does a market need to be designed?
Many markets are the product of evolution rather than intelligent design. So, why is it necessary to design a market? One answer is that there is an externality that is difficult to price without a high degree of coordination. Electricity markets are offered as an example. In the second lecture we will examine this in more detail.