Saturday, April 13, 2013

More on Avent and the IMF

In the post below, I vigorously criticize IMF staff and Ryan Avent for claiming that central banks adopted low inflation targets in the early 80s without noting that the Fed did not adopt an inflation target until January 25 2012.  I have now read Avent's post as patiently as I can (meaning I skipped ahead).

Avent wrote "That the disinflation of the 1980s has generated a flattening of the Phillips curve is precisely what the IMF demonstrates:"  

This claim is illustrated by a figure which does not show that.  Even if a curve hasn't changed at all, the slope depends on where you are (that is the curve is not a straight line).  The figure does suggest that  the IMF staff are willing to assume that the Phillips curve is a sraight line, or rather that they are willing to support their argument by presenting a graph which tends to convince people willing to make that assumption

The graph does not demonstrate any change at all in the Phillips curve (I'm not saying it didn't change just that the question can't be answered with the graph).  You can't see if different points lie on the same curve by plotting changes on changes, because the slope of a curve isn't constant.  

In particular, inflation is much lower now than it was in the early 80s.  It is possible that the slope of the Phillips curve is lower now, because the Phillips curve is a curve.  The pattern from 2007 on is clearly different from the pattern in the 1930s.   It is not clear that it is different from what would have happened from 1980 to 1994 if inflation had been around 2% in 1980.  

Oh and the 30s were different. In most developed countries, the unemployed don't risk starvation any more.  The welfare state was quite different back when high unemployment caused sharp deflation.

I swear that this post has been edited to make it less rude.  You don't want to read the first draft.
Also I deleted a draft conclusion to the update to the post below, becuse it was too inflamatory.  I am trying to be as polite as I possibly can without actually lying.

update:  Now I am going to make some graphs.  They are totally unlabeled only partly because I am lazy but also because I want the reader/graph eyeballer to try to guess what they are.  They are US analogs of the IMF graph with the change in core inflation on the y axis and the change the civilian unemployment rate on the x axis.  All graph 17 data points (as in the green series from the IMF).  Two  show data from after the Fed flattened the Phillips curve in the "early 80s".  Which two show the new flat Phillips curve ?

Figure 1 (chosen from three figures at random by the eyes closed point and click method)

Figure 2 (not chosen at random)

Figure 3

Don't peek 

do you really want to peek ?

Come on it's more fun if you guess then look ?

Did you peek ?

OK the answer is that figures 1 and 3 show the new flat post early 80s Phillips curve which is due to inflation targetting.  

Did you guess without peeking ?

Firgure 1 shows 17 quarterly inflation changes from 1985q1 minus 1984q4 on.  They are the first data which came undeniably after the early 80s.  Figure 3 shows the most recent 17 quarterly changes.   It is not markedly different from figure 1 because of autoscaling (not "not *just* because I am lazy" does not imply "I am not lazy") but it is much flatter (the range of unemployment changes is 4 times as large and the range of core inflation changes is about the same).

Figure 2 is the first 17 available quarters from Fred from 1958q2 - 1958q1 (when the core CPI series starts).  The first of those data were collected before Phillips published his famous scatter (with labled axises even) .  The last in the first quarter of 1962 rather before the modern advances in monetary theory.  It is very flat indeed.  If Phillips had relied on FRED, he wouldn't have gotten published at all.  Inflation bounced around way back then, but there is almost no relationship between the change in inflation and the change in unemployment.

This is what Phillips saw for extremely low inflation rates.  The rediscovery of the fact that the Phillips curve has a low slope at inflation rates near zero is not pathbreaking progress.

No comments: