Wednesday, August 05, 2015

What Wouldn't be a Dynamic-Stochastic General-Equilibrium model?

In Which I Try to Answer a Question asked by Larry Summers as a Joke

The LSE held a discussion on Reconstructing Macroeconomics. Brad DeLong posted a transcript , the video is here.

Larry Summers opened with a joke

Larry Summers: You know I was tempted to blast off at Dynamic-Stochastic General-Equilibrium models. That is, actually, my inclination. But on the other hand it occurred to me to ask the question: "What wouldn't be a Dynamic-Stochastic General-Equilibrium model?" That would be a Static-Certain Partial-Equilibrium model. It is hard to see how that represent any kind of an improvement. So I can't be against DSGE on principle.

I have a sense of humor, but I am going to suppress it and pretend to take the joking answer literally. I note that the diametric opposite of a Dynamic-Stochastic General-Equilibrium model would be a Static-Certain Partial-Disequilibrium model. Even in jest, even Summers has trouble separating the concepts of model and equilibrium -- which in context means Nash equilibrium. Also the joke is a joke, a Dynamic-stochastic-partial equilibrium model is not a Dynamic-Stochastic General-Equilibrium model. It is also easy to answer the question, because there are models older than any DSGE model -- Summers can propose we go back to using those models. For one thing, he clearly does use those models (as do DeLong and Krugman). They weren't equilibrium models. Bernanke and Blanchard (who have made huge contributions to Reconstructing Macroeconomics) assume in their answers that they are required to start with a standard new Keynesian DSGE model and modify it to reconcile it with reality. Blanchard said

Suppose you are writing two textbooks, one undergrad, one grad. In the undergraduate textbook, it seems to me that when teaching the IS-LM, [skip]

At the graduate level, we now have this explosion of DSGE models which put one friction and another into the model. Again, targeting pedagogy, it seems to me that there are two mechanisms which are central. The first is leverage, which starting with Ben [Bernanke's] work and earlier work we have, I think we know how to deal with it. The second is liquidity. And I think there we are much less far along the way. Again, I am hoping that someday we will put it together and have a simple way of thinking about leverage and a simple way of thinking about liquidity. These two things will come into our New Keynesian model, and we will be able to tell a simple story. We are at the stage at which the DSGE models have much too much in them to be fully understood.

Blanchard is not joking. He takes it as a given that the IS-LM model is for undergraduates and that graduate teaching and research should be based on new Keynesian DSGE models. He also notes a problem -- current DSGE models do not clarify thought, because we don't understand what is going on in the computer as it simulates them. He neglects another problem -- DSGE models are based on extremely strong assumptions (including rational expectations but also including say the assumption that there is no housing sector or inventories) which we are all sure aren't literally true. The only defense of the approach is that we should think about simple things which we understand which might give us insights into the much more complex real world. I find it hard to accept the assumption that macroeconmics must be based on incomprehensible models which fundamentally rely on assumptions we are sure are false in ways which seem to have been critically important and which yield, at best, mediocre forecasts.

I'd like to see a debate where Summers (or Krugman or DeLong) argues for the resolution "Old Keynesian models from the 60s and 70s are a more promising starting point for macroeconomic research than New Keynesian DSGE models". Someone would have to argue contra. Oddly, I find it extremely difficult to think of (and impossible to find) anyone willing to do this. I can't recall hearing or reading a defense of the NK DSGE approach. It is just assumed that this is what macroeconomics is and must be, but I honestly can't recall an argument for why it should be (hmmm am I too young to be senile?).

Starting this post, I had planned to argue for the resolution, but this is getting long and I want to type about how we got where we are. Senile or not, I am too young to remember, Old-Keynes had been abandoned already when I arrived in economics in 1985. But this is a blog.

Clearly the new Keynesian school is a response to the rational expectations revolution. Roughly (and remembering vaguely) Lucas argued that old Keynesian conclusions were based on strong simple assumptions about how people formed expectations which were not rational expectations. He argued that economists should assume rational expectations and presented a model in which there is no output-inflation tradeoff. New Keynesians (Akerloff, Yellen, Blanchard, Mankiw, Fischer, Taylor etc) argued that with price stickiness there was an (actually very minor) tradeoff and optimal monetary policy could prevent second order costs. But somehow the assumption that Lucas was right on all topics other than maybe price stickiness ceased to be something assumed for the sake of argument and became something which must be assumed.

There are two popular explanations for this. One is the Lucas Critique "Econometric policy evaluation: A Critique" (which Summers once told me was considered obvious, well known and uninteresting at the conference where it was presented). A problem with this historical hypothesis is that, as noted by Lucas, the Lucas critique was not very original at all -- he cited the same arguments made by the leading paleo Keynesians. A deeper problem is that the methodology macroeconomists adopted combine an appeal to the Lucas critique to reject all models not based on optimizing agents and an appeal to Friedman's Methodology of Positive Economics to argue that it is fine to make false assumptions which are inconsistent with available data, so long as the model fits past data on the variables of interest. The two claims about methodology on which modern macro rests are not just contradictory, each is exactly the assertion that the other is wrong. Now people clearly can hold contradictory beliefs in their mind at the same time, but it is hard to argue that two contradictory claims were very widely accepted because logical inconsistency is so attractive.

I think the claim that the shift was a result of the power of the Lucas critique utterly fails as a hypothesis in the history of thought.

The much more popular explanation is the argument that the stagflation of the 70s convinced economists that the old Keynesians had missed something fundamental. The problem with this argument is that Keynes, Hicks, Samuelson and Solow all demonstrably didn't miss that possibility. I think this is a matter of historical fact established by James Forder. I suspected it for months and found confirmation after 5 minutes of googling.

I am not aware of an active debate on this topic. I don't know of anyone who argues that old Keynesians actually made the mistake ascribed to them by Friedman and Lucas. I also don't know of many people willing to argue for the old approach.

No comments: