I find his critique of DSGE entirely convincing. However, he doesn't. He wrote " I see the current DSGE models as seriously flawed, but they are eminently improvable and central to the future of macroeconomics. "
He notes the flaws that (in my words)
1) the core assumptions are absurd and have yielded false implications (the models can be fiddled to eliminate those implications as one would expect if the approach were "a dangerous dead end").
2) After the fiddling, the models are too richly parametrised to estimate so the implications are the result of conventions and not of any interaction with reality.
3) the models have normative implications which are clearly nonsense.
and 4) only the economists who write the literature can understand it.
Sure sounds like a "dangerous dead end" to me.
Blanchard responds to the straw person who says DSGE is a dead end as follows (I mean that literally -- the is all of the defense DSGE models in the article)
The pursuit of a widely accepted analytical macroeconomic core, in which to locate discussions and extensions, may be a pipe dream, but it is a dream surely worth pursuing. If so, the three main modeling choices of DSGEs are the right ones. Starting from explicit microfoundations is clearly essential; where else to start from? Ad hoc equations will not do for that purpose. Thinking in terms of a set of distortions to a competitive economy implies a long slog from the competitive model to a reasonably plausible description of the economy. But, again, it is hard to see where else to start from. Turning to estimation, calibrating/estimating the model as a system rather than equation by equation also seems essential. Experience from past equation-by-equation models has shown that their dynamic properties can be very much at odds with the actual dynamics of the system.
So he argues that DSGE is a good approach, because he can't think of another one. So, just for example, it is necessary to start with plainly false assumptions about individual behavior, because one obviously can't start with observations of individual behavior and estimate behavioral patterns (note there are more than 240 data points). Also one must start with a competitive model, because OJ Blanchard said you must.
He argues that one must start from DSGE because he can't think of an alternative *and* dismisses other approaches without considering them. I think that Blanchard genuinely doesn't think there is any macro but DSGE macro. I note in passing that he has friendly relations with Paul Krugman, Larry Summers and Brad DeLong.
OK his conclusion (includes) "DSGE models can fulfill an important need in macroeconomics, that of offering a core structure around which to build and organize discussions. " Krugman notes that this sounds just like the defense of Marxist theory he used to hear. I ask why the core structure couldn't be national income and product account identities and the definitions of wage inflation, price inflation and interest rates ? update: Pulled back from comments. I got a bit uh enthusiastic down in the comment thread and decided to pull things up here.
First the comments
Anonymous [I] said...
Hi Robert, I'm an interested graduate student (who used a DSGE model in his dissertation). Would you please elaborate on your suggestion about a core structure using national income and product account identities etc.? I'm interested in what we would use instead of the elaborations on the 3-equation NK DSGE model. Cheers!Anonymous [II] said...
Robert, this is perhaps a criticism you have heard too often, but if there is a better alternative out there why isn't in the journals and at conferences? There is currently a consensus why not convince the other side that there is a better model out there? To me, and I do not do macro, DSGE seems like a sensible and reasonable approach. I have read about alternatives but their proponents seem rather ideological and rather unprofessional in their behavior. Their arguments seem to boil down to ad hominems and the models are incidental to it.
Anonymous I: The alternative framework I had in mind is the Paleo Keynesian framework with an IS curve, a Taylor rule and some sort of Phillips curve. Oh my, those are the same trhee equations. The IS curve and the Phillips curve are old equations from the 30s and 60s respectively. The Taylor rule is the odd equation out in the 3 eqn NK model, because it is not derive from an optimization problem and is rather an estimated empirical relationship.
My point (if any) is that the New Keynesian project is to reconcile the old Keynesian equations with optimization. To the extent that the mathematical and theoretical effort is completely successful and the behavior of the New Keynesian model is identical to that of the old Keynesian model, nothing has bee accomplished (if this were so, Keynesians would have learned nothing and forgotten nothing).
In actual history, the NK models have implications different from the old Keynesian equations. Also consistently without one exception that comes to mind, the new implications are also rejected by the data. Blanchard mentions this when he discusses how the PIH had to be patched to fit the excess sensitivity of consumption to income and the NK Phillips curve had to have an unmotivated assumption about indexation somehow so that lagged inflation matters separately from the conditional mean of future inflation.
He has also discussed how the new improved micro founded Phillips curve doesn't work as well as the old primitive one https://piie.com/system/files/documents/pb16-1.pdf
and in the post under discussion he explains the approach of figuring out what a market information of policy using an old ad hoc model, then noting that the same result still holds in a DSGE model which is designed to correspond to the old ad hoc model, then explaining what is going on so people can understand it by presenting the old ad hoc model. This is what he does. However, he also asserted that the DSGE step is useful in some way. He gave no hint as to what way exactly.
I stress that the DSGE step is *not* harmless. It can be very difficult and time consuming to translate to DSGE (and worse to translate from theoretical DSGE to code a computer can run). So it is standard to leave out little details such as the housing sector when discussing the great recession. To get to DSGE sacrifices are made. They include any consideration of the fact that not all investment is business fixed investment -- the models don't include houses and don't include inventories. Banks were added in the past 8 years. The current models have no place for oil shocks (which are considered to be sort of like technology shocks which is fine except for the fact that standard theory says their effects are different).
Now Blanchard could say, if you want to get a job or, having a job, want to get tenure, you have to do DSGE too even if you also present an ad hoc model as a heuristic explanation of the behavior of the DSGE model. In any case I do. Don't ruing your career by refusing to dress up your actual thought in DSGE disguise.
There is another approach. It is what Blanchard himself does. Also Krugman, Summers, and Stiglitz. These are not marginal figures.
Anonymous II I refer you to the linked Blanchard post. He explains that the other way, the old way, is presented at conferences by Blanchard who is allowed to do so because he is the discussant. It isn't hidden at all. It's the way New Keynesian macroeconomists talk. It isn't the way they write (except on blogs).
Another approach is available. It has been available and used for decades. Somewhat more than half of macroeconomists are saltwater economists who sure seem to think the old way (as old Keynesians) then disguise their thoughts with DSGE models. A now somewhat smaller number are fresh water RBC economists who assert that monetary policy doesn't affect GDP, that recessions are efficient, and that nothing much unusual was going on in January 2009 (I could look up the exact Prescott quotes -- he said all those things -- but I won't bother). In contrast, I can't think of a single economist who is demonstrably a sincere New Keynesian economist in that he or she accepts the new Keynesian model but demonstrably doesn't secretly also accept the old Keynesian models.