Weekend Reading: New Keynesian Orthdoxy and Hysteresis: "The debate between Gerald Friedman and the economics profession seems to be settled...:
...Friedman agrees that he may have made a mistake. It is really unfair to him that preliminary unpublished analysis received so much attention (mostly from the Sanders campaign). Oh and it is unfair to them to expect them to ignore favorable forecasts made by a Clinton supporter of the effects of Sanders’s proposals. The discussion did lead to an interesting discussion.
Uh on twitter:
@Noahpinion: "The idea that temporary bursts of govt spending will raise output forever and ever is well beyond the bounds of Keynesian thinking."
@robertwaldmann: "@Noahpinion hysteresis is outside the bounds of new Keynesian thinking by assumption not evidence or even argument."
@fl0pson: "@robertwaldmann @Noahpinion didn’t Blanchard and Summers write the famous hysteresis paper? and wrote a new one months ago?"
@robertwaldmann: "@fl0pson Summers is not a new Keynesian. Blanchard assumed no hysteresis soon after arguing there is hsyteresis (sic and link — I think this is the most influential paper with the name ‘Blanchard’ on it)."
@fl0pson: "@robertwaldmann @Noahpinion i hate NK models (mostly bc solving them sucks) but i figured you could kludge hysteresis in there somehow."
@Noahpinion: "@fl0pson @robertwaldmann sure." @robertwaldmann: "@fl0pson
@Noahpinion economic orthodoxy is odd — too complicated for twitter."
Hence this post:
A large number of prominent progressive economists argued that it is generally agreed and probably true that demand stimulus has only temporary effects and in particular that the level of GDP depends on the level of stimulus. I note in passing that this is already a big step from New-Keynesian back to paleo-Keynesian. In 2008 the mainstream New-Keynesian view was that expansionary monetary and fiscal policy both cause higher output (so new Keynesian not new classical) but a permanent shift towards policy which causes higher output in the short run would have no effect on output in the long run.
Yet leading New Keynesian Olivier Blanchard and Larry Summers recently argued that depressed demand can have permanent a effects on output and quite possibly the rate of growth of output (exactly Friedman’s derided assumption). What is going on here?
I think there is a common pattern (not entirely fitting this particular case). There is a standard model in which demand stimulus has only a temporary effect on output and certainly no long term effect on the rate of growth of output. The model does not fit the data. In the academic discussion, macro economists note this and discuss alternative models. But if an outsiders (the Sanders campaign) or other than top status academics say something inconsistent with the standard model, economists say they are wrong and appeal to the standard model.
This can make the orthodoxy invulnerable to data. It can be noted that it is rejected by data and alternatives discussed when confronting the problematic data, but this discussion isn’t shared with non-economists.
So far I have been unfair to the orthodox progressive economists. Their actual position is that the long run effect of temporary stimulus is smaller than the short run effect. Friedman assumed they were equal. The CEA chairs’ critique of Friedman is consistent with Summers’s view expressed here, here and here (all pdfs). When discussing the discussion Krugman asserted:
Finally, there’s hysteresis: the proposition that demand-side weakness now breeds supply-side weakness later, so that there are big payoffs to boosting the economy through public spending. There’s now a lot of evidence for that proposition...
But there are still odd things going on: One is the appeal to models which are known to be inadequate when explaining economics to non-economists. This is very odd. It makes sense to teach simple models before moving on to realistic models (Newtonian physics is still taught). But it makes no sense to discuss policy using old models which don’t fit the data — that would be like assuming no relativity when discussion public spending on particle accelerators (why spend so much on colliding beams when according to the simple model the energy released by the collision is only twice as high as with a beam hitting a stationary target ?) or assuming no tunneling when designing transistors.
The other is the use of failed models to make forecasts. It is known that the standard New Keynesian model (Smets-Wouters) doesn’t fit the data post-2008 well at all. But similar models are still used by central banks and such. The problem as Sam G (and Paul K and many others note) is that it is hard to program up DSGE models so the choice is sticking to the old one or working without a DSGE model. I think the actual practice is to have a DSGE model, ignore it when its implications are silly but treat it as authoritative when this is useful.
The recent G. Friedman vs Keynesian consensus debate has been civil and intellectually honest...