What's happening in monetary policy and macroeconomics.
Friday, August 27, 2010
What Happened to Coordination Failure Models?
In the 1980s, there was a Keynesian research program using coordination failure models. Early on there was John Bryant's model, and Peter Diamond has a neat search model with multiple equilibria. Russ Cooper worked on this too. People took those ideas, and what was known about sunspot equilibria, and developed dynamic models with intrinsic aggregate uncertainty, that in some forms (I'm thinking of work by Farmer and Guo, for example) could replicate features of the data. Woodford did some work in this literature, as did Benhabib, and others. Those were coherent market-clearing competitive general equilibrium models that, one could argue, captured ideas in Keynes's General Theory. One could not only fit the models to data, but extract features that could in principle be used to argue for or against their empirical plausibility. Why did that stuff get shoved aside in favor of the New Keynesian sticky-price model?
Posted by Stephen Williamson at 11:26 AM
Subscribe to: Post Comments (Atom)
Well, people are uncomfortable with multiple equilibrium models unless they also tell a convincing story of why we should happen to be in any particular equilibrium at any particular point in time. This is basically the standard objection to sunspots: it is not at all obvious how expectations would end up coordinated around a particular non-MSV solution.ReplyDelete
That said, there's plenty of good work still being done with sunspots, e.g. Jaimovich (2007) or Wang and Wen (2007).
I agree with Tom's explanation, but I disagree with "people's" opinion. Multiple equilibria is the only sensible explanation of keynesian's lack-of-demand explanation of a recession. Why economies between equilibria is another matter, why the government should know how to make people switch (using public investment) is yet another matter, etc... etc... but it is, indeed, a coherent story.ReplyDelete
Farmer's still working on these things, though based more on his work with Beyer rather than Guo(I'm still working through his "Expectations, Employment and Prices" book. Definitely GE models with Keynesian features though.ReplyDelete
Speaking only personally, I gave up building coordination failure models because the only models I could build were very "fragile". Fragile in the sense that the tiniest change in the assumptions meant very big changes in the conclusions. I needed "knife edge" parameter values to get the desired results. So my models just didn't seem very plausible.
But I haven't given up on the idea. Just can't think of any plausible and robust way to make it work.
Plus, the labour supply curve is just too inelastic to make most of them work. Exactly the same problem that RBC theorists face.ReplyDelete
I think that Howitt and McAfee (1992) is a place you might start. Before you do that though, here is a nice summary by Peter:
As one of the commenters already mentioned, Roger Farmer is still doing this type of work.ReplyDelete
Here is his recent JME paper:
Thanks for the links, David and Josh. Ah, the UWO connections! (Roger, Peter, Preston, as well as Steve, David, and me.)ReplyDelete
I still keep thinking though, just as Samuelson asked "Have you ever seen an egg standing on end?", how often do you see a ball resting on a perfectly flat surface? It's more usual to invoke friction, to explain why things don't move.
Virtually the entire General Theory was devoted to the question of "why things might not move on their own." Are you now suggesting that we abandon that enterprise? ;)
John Duffy and I published something in the AER five years ago that was an experiment based on correlated equilibrium. It was not quite a Cooper-John framework because it lacked the super-modularity of Keynesian games. But it was picked up by the New Scientist as a possible explanation for the financial crisis afte the way the Paulson bailout plan was announced.ReplyDelete