Fri May 21st, 2010 at 12:21:05 PM EST
Paul Krugman: Core Logic
Now, the measurement issue: we'd like to keep track of this sort of inflation inertia, both on the upside and on the downside -- because just as embedded inflation is hard to get rid of, so is embedded deflation (ask the Japanese). But in the real world, while some (many) goods behave like this, some don't: their prices rise quickly with supply and demand changes, and don't display inertia. So we need a measure that extracts the signal from the noise, getting at the inertial part of the story.
The standard measure tries to do this by excluding the obviously non-inertial prices: food and energy. But are they the whole story? Of course not -- and standard core measures have been behaving a bit erratically lately. Hence the growing preference among many economists for measures like medians and trimmed means, which exclude prices that move by a lot in any given month, presumably therefore isolating the prices that move sluggishly, which is what we want.
So let me get this right: The concept of "core inflation" is justified by the need to keep track of "embedded," or systemic inflation caused by oligopolistic price-setting.
Now, the best way to do that would be to study the actual economics of the issue and try to figure out which parts of the economy are dominated by oligopolies and monopolies, and which parts of the economy are dominated by competitive firms. Of course, that may or may not be practical. So the second best way to do it would be to develop some kind of heuristic for detecting oligopolistic pricing in the raw price data.1
But that's not what our dear econometricians are doing.
[editor's note, by Migeru] Part of the Reasons for Despair occasional series.
What they are doing is simply arbitrarily stripping out sectors that "everybody knows" "obviously" do not contain systemic inflation. And now, after 20-odd years of talking about "core inflation," they figure out that gee, maybe that isn't as meaningfully defined as we thought it was. Oh R'lyeh? I am shocked, shocked I tell you, to see gambling going on in my casino!
So what do our dear econometricians do to remedy this defect? According to Krugman, they go and make a heuristic and apply it to the raw data. Which is good. About 20 years late, but good. What's bad is that they're only looking at this iteration's data, which contains no information about inertia,2 which is supposedly the effect that they want the heuristic to look for. As far as I can tell from Krugman's description, they're killing the signal with the noise.
But the upshot of all this is that, since they are now essentially cherry-picking data that behaves like they want it to behave, they'll never have to be worried about the index not displaying inertia again. The resulting index is still nonsense, but now it's nonsense that matches the models of how it's supposed to behave.
1Just off the top of my head, and based on Krugman's description of the phenomenon that they want to model, you could decompose the price changes of a given good by magnitude and frequency and call oligopolistic price-setting by looking at the shape of the resulting spectrum.
2That 10 % price hike you saw this month? Could be adjustment to changing market conditions, or it could be adjustment according to the seller's price adjustment schedule - without looking at the pattern this price shows over several previous iterations, you have precious little way of telling.