Evidence for the slope of the supply curve
This is a companion post to the previous one. In that post I argued that NGDP was superior when the supply curve is nearly flat, in this one I argue that the empirical evidence strongly favours a flat supply curve at the following times.
What follows are a series of plots of RGDP vs NGDP with associated trend lines.
The gradient of this graph is 0.55, it shows a reasonable correlation of . In other words, over the time frame, in general a one percentage point rise in NGDP was associated with a 0.55% rise change in real output, and a 0.45% rise in inflation.
Now lets look at 1990-Present, i.e. the era of low inflation. Now we see:
Note that now, in this era, a 1% point rise in NGDP is associated with a 0.85% rise in RGDP. I.e. The supply curve is very flat. This is exactly the danger area identified in my previous post, where in order to reduce inflation by a small amount, one must sacrifice a huge amount of output. A 1% reduction in inflation would correspond to a 6% reduction in output. Note also that the correlation is a lot better in this era, than in the whole data set. This should increase our confidence that this is a meaningful subset – i.e. that there is something qualitative different about this selection of points from a random sampling. In this case, the adoption of a low inflation target.
We can also estimate “expected inflation” from this analysis. Expected inflation is the level of NGDP growth needed for RGDP to be flat, so in this case, 1.44/0.849 = 1.7%. Since this is close to the inflation target, that is further evidence that this analysis is meaningful.
Lets look at the 1970’s. We might expect that this is a time when we got onto the vertical part of the supply curve, :
This time we have an expected inflation reading of 6.5%, along side an almost identical gradient. We can say here with some confidence that inflation expectations were higher in the 1970’s, but what to make of the identical gradient? I expect what we are seeing here is expectation adjustment. Every time the expectations adjust you get moved back onto the flatter part of the supply curve. This is the theory of the inflation adjusted Philips curve. This causes the correlation to break down, as a change in inflation expectation moves the whole line up and down.
Lets look at this era in more detail:
Here are two periods, Oct 68 to Jan 72, and Oct 73-79/01. In both of these periods we get pretty good correlations, but inflation expectations have jumped, and the supply curve becomes even flatter. Perhaps what is happening here is that as uncertainty about inflation becomes chronic, output falls, and we get onto the flatter part of the supply curve. Here inflation expectations jumped 5 percentage points in a few years. This is indeed puzzling. Perhaps the real lesson here is that the supply/demand framework does not work well without a nominal anchor around which to have “unexpected” stimulus.
Nevertheless, the broad point stands: Throughout the 1990’s, the trade off between inflation and output was terrible. Thus, attempts to keep inflation on target by tightening policy in response to a 1% supply side shock to prices, would result in around a 6% loss of output. Given the small costs of small deviations from the inflation target, central banks should be prepared to tolerate deviations from the inflation target, provided that they are doing so in order to keep aggregate demand more stable, because the costs of a loss of demand are huge, and the costs of missing your inflation target are tiny.
Nowhere is this more clear than in the 2011 rate tightening of the European Central bank, which in order to prevent a small overrun in its inflation target, was forced to induce a catastrophic loss of demand.