John Sutton, Marshall’s Tendencies: What Can Economists Know?
If you’re into economics, find it a little disappointing, and would like a more-philosophical (while still firmly mathematical and rigorous) take on the discipline, this book is for you.
In two earlier, exceedingly hefty and fascinating books — Sunk Costs and Market Structure and Technology and Market Structure — Sutton has put forth a particular, humble vision of economic modeling. Most economic models involve specifying a set of parameters quite precisely, very carefully laying out how actors (that is, people or companies or whatnot) will behave, then solving for their behavior in “equilibrium.” That equilibrium can evolve over time, so another class of economic model — those based on evolutionary game theory maybe being the most famous — carefully lays out the rules by which people change over time. The models might include some process of learning, for instance.
Sometimes this precision works — matches up with the data — and sometimes it doesn’t. When it doesn’t match up, quite often it’s because our models are missing important variables. Models need to be simple in order to be usable, though, so we can’t very well add in every conceivable variable that might affect an economic outcome. (Though here is the place to note that a realistic psychology would go a long way to building microfoundations that might eventually bubble up into accurate macroscopic models of, e.g., stock-market pricing.)
Sutton’s response is refreshing, and is unique at least among the bits of economics that I’ve read: abandon altogether the search for One True Model. Instead, pick a few axioms that any credible model must satisfy, then use those axioms to derive a class of models in which the truth is likely to lie. Specifically, his models of industrial organization rest on two principles:
- Viability: In equilibrium, every company in a particular industry will be making nonnegative profits.
- Stability: No new company could enter and make a certain profit.
The latter condition is essentially an arbitrage principle: don’t assume that all economic actors are rational; only assume that if there were an obvious opportunity, someone would eventually take it. An equilibrium industry configuration is then one in which both viability and stability are satisfied. (I found a paper of Sutton’s entitled “One Smart Agent” that bears on this subject and may be interesting to some of my readers.)
Sutton’s approach here is really elegant, really simple, and promises to be really productive. Being an eminently fair man, his next step is to ask under what conditions the classic economic approach — one model to rule them all — is likely to bear fruit, and under what conditions his class-of-models approach will work better. In the process of answering this, he sketches some really beautiful game theory on the design of auctions, specifically auctions of petroleum-bearing lands. I can’t do any better than Sutton in laying out the theory here, so I’ll just point you to page 47. The upshot is that in the case of an auction, we know very precisely how participants will behave, because we know exactly what the rules of the auction are. Sutton’s own field of industrial organization is much less well-formed, hence much more usefully treated with a class-of-models approach. (Full disclosure: I never finished Technology and Market Structure or Sunk Costs and Market Structure; that mostly had nothing to do with their mathematical content — which is substantial — and had more to do with my available time.)
His writing is dense but not difficult; one just needs to read a bit more slowly than usual. Without ever having met the man, I can only imagine that he’s a fun, amiable, brilliant sort. On the way to telling us what sort of workable models he thinks we have any right to expect in economics, he sketches the history of modeling tides in physics — fascinatingly enough to make me want to rush out and read the appropriate citations. This is where Marshall's Tendencies gets started, in fact: it seeks to understand why modeling aggregated human behavior might be a much different task than modeling aggregated water waves.
Sutton traipses from waves to game theory to industrial organization, all with enough rigor to satisfy the most demanding reader but with enough of a light touch to never bore you. All this in just over 100 pages. Bravo to Professor Sutton.
(Thanks to Cosma Shalizi for the recommendation. Though I should note — purely to prove that I still do possess an independent will — that my earlier exposure to Sutton inclined me toward this book. That said, I wouldn’t have even looked at TAMS or SCAMS had I not seen them recommended on his blog. This quest for free will is more than likely illusory.)