He wrote about what he sees as four missing ingredients in mainstream macroeconomics. Let me focus on his last two:

There is plenty of evidence that price rigidities are important and they help us understand some of the features of the business cycle. But there must be more than that. There are other frictions in the real economy that produce a slow adjustment and are responsible for the persistence of business cycles.

… The notion that co-ordination across economic agents matters to explain the dynamics of business cycles receives very limited attention in academic research.

Robert Waldmann adds,

I think the problem is that with coordination failures, multiple equilibria are possible. Not just two or two hundred either, but a large infinite number. There is no theory which tells us how likely different equilibria are. Worse, policy shifts can cause the economy to jump from one equilibrium to another in unpredictable ways.

This does not strike me as an argument against the validity of models of coordination failure. In fact recessions are hard to predict and, well, look like panics. The problem is that models which say that macroeconomists will not be able to predict well are not popular.

Pointer from Mark Thoma.

When I wrote one of my PSST papers, Peter Howitt’s response put it in the coordination failure literature. In fact, I think there is a (possibly slight) difference. “Coordination failure” sounds to me as if there is a great equilibrium sitting there, and people just cannot find it. I think that the patterns of sustainable specialization and trade need to be discovered, through trial and error.

I believe that even if there were no frictions impeding coordination, as long as entrepreneurs have to test business models without knowing whether they will work, there will be business models that fail and unemployment can result. However, this is a weird hypothetical, since there is obviously no such thing as an economy with frictionless coordination.

So my view of what macro needs would include coordination failure of a variety of types, as well as trial-and-error learning. In my opinion, the resistance to this comes from various sources:

1. As Waldmann says, you have multiple equilibria. In fact, you never get to any one equilibrium, so that the very notion of equilibrium as a core modeling element loses salience. That creates a ton of discomfort.

2. The system is no longer hydraulic, in which more X (fiscal stimulus, monetary growth) leads to proportional increases in Y (GDP, inflation, employment). For many macroeconomists, the whole point of modeling is to come up with implications for fiscal and monetary policy. The idea that your model may not lead to a cure for business cycle makes the effort seem pointless, at least in compared with Keynesian can-do thinking.

3, It is inconsistent with popular modeling simplifications, such as the “representative agent.” For the purpose of publishing papers in journals, economists like to gravitate toward standard models. It is much easier to get published if you do a variation on a standard model than if all the people working on an idea are just groping around in different ways. I think that the coordination-failure approach to macro involves this disparate groping, and thus it suffers from….a co-ordination failure, if you will. The DSGE folks can co-ordinate. The rest of us can’t. So even though the DSGE stuff is a blind alley we have better ideas, when it comes to the journal process, we lose.