One of the questions I like asking students is whether inflation following the Great Recession has tended to favour the New Keynesian (NK) Phillips curve or its more traditional counterpart (TK). I like it because it allows me to draw a nice diagram, and also because it shows students how difficult it is to discriminate between theories in macro.
So first the theory. The two competing models are
- NK: Inflation at t = expected inflation at t+1 together with a term in the output gap
- TK: inflation at t = inflation at t-1 together with a term in the output gap
I’m ignoring discounting in the NK Phillips curve for simplicity. Assume expectations about inflation are rational, and suppose the economy is hit by an unexpected recession of known size and duration. The two models predict the following:
With the traditional model, inflation gradually falls as the recession continues, and once it comes to an end, inflation remains lower. In the New Keynesian model, assuming that the inflation target is credible, inflation jumps down when the unexpected recession occurs, and then inflation gradually rises towards its target as the recession progresses. (We assume here that the output gap is constant while the recession lasts, again for simplicity.) For the NK model, it is critical in drawing this diagram that the extent of the recession is known – more on this below. The patterns implied by the two models are distinct, and this difference is likely to persist even if each curve becomes flatter as we approach zero inflation because of nominal wage rigidities.
To see what has actually happened, see this nice post from Gavyn Davies. The immediate aftermath of the recession looked more like the NK model: a sharp fall followed by a gradual rise. Furthermore I would argue that – once the recession hit – most people expected it to be large and persistent, so my diagram is not totally unrealistic. But if we look at what has been happening in the last two years, it looks much more like the TK model, with inflation gradually falling below target.
That is probably as far as we should go without doing some econometrics, and also taking account of some of the complexities discussed here. We could probably get any pattern to fit the NK model by imagining a suitable sequence of expectations errors. In addition if we are looking at consumer price inflation we should account for commodity price changes, which neither model does. (If we look at GDP deflators, you could tell a story where agents were initially expecting a recession lasting three or four years, and have been surprised that the recession has persisted ever since.) That is why some proper econometrics is required, preferably looking at both price and wage inflation together with expectations data. (If such studies have been done, please let me know.)
However perhaps I can suggest two possible conclusions that such studies could test more rigorously. First, the traditional Phillips curve, where expectations are implicitly naive and backward looking, does not look like a promising basis for explaining inflation following the recession. Either the New Keynesian model, or some combination of the two models, looks more like providing an adequate foundation for a reasonable explanation. Second, an explanation based on the NK model that treats the size and extent of the recession (whatever that turns out to be) as one initially unexpected but then completely anticipated shock is also going to struggle to fit the data.
Could the theory that rates are stuck at the ZLB be used to patch the New Keynesian model? I thought that was the purpose of the analysis around the ZLB.ReplyDelete
One concern I have is that we do not have a measure of the output gap. Is there a risk is that the output gap measure willl be revised to give the right inflation prediction?
Nice post. I largely agree.ReplyDelete
A rule of thumb I find useful:
A PC based on forward looking expectations (for which RE may be a useful approximation) works better when there is a particular type of regime change. Not your typical NK regime change where a change in the Taylor rule would be a regime change.
Rather, a regime change such that at least one of the regimes involve a period of significant, widespread and shared socio-economic anxiety, which plays a role in coordinating the behaviour of the aggregate.
So a forward looking expectations based PC is likely a better guide in the following scenarios:
1. The start of a major depression, esp if preceded by the bursting of a bubble
2. The beginning or end of major wars
3. The end of a hyperinflation.
There are so many dogmatic believes here. For instance, I am not sure why the discount term in the NKPC should be close to unity. How do you know the average John devalues tomorrow by only 5%? Why should you in the first place assume that we are all identical when you build your DSGE model? People are different in so many respects and admiring this thing would destroy the macroeconomics as we know it.ReplyDelete
I am astonished that Philips curves, in whatever guise, still feature in discussions of economic behaviour. It seems to me that the theoretical underpinnings of Philips curves are deeply flawed. For example see the analysis that is presented on Page 6 of website www.economyuk.co.ukReplyDelete