Scott Sumner makes a very good point (though my interest here is somewhat peripheral to the main thrust of his post):
Government price indices don’t measure the prices that are of macroeconomic interest. For instance in the 6 years after the housing bubble peaked the US, BLS data shows housing prices rising by about 10%, while Case-Shiller showed a 35% decline. Housing is 39% of the core CPI. That’s a big deal.
Here’s what that looks like:
Yow. (The important CPI sub-components of housing, i.e. owner equivalent rent, look similar.)
Contra the sky-is-falling inflationistas at ShadowStats, this suggests that CPI has greatly overstated inflation since the (Shiller) housing peak in April 2006. Just for illumination, here’s a rough-and-ready shot at replacing the 40% of housing movement in the CPI with the movement we see in Case-Shiller:
If this has any merit, we’re looking back at three to six years of deflation. It also suggest that inflation has been shooting up in the last year or so. Do with that what you will.
Paul Krugman often defends CPI against ShadowStats-style attacks by pointing to the Billion Price Index, which tracks closely with CPI over time. But the BPI is an index of retail prices. It doesn’t include housing, health care, education, and many other components that make up more than 50% of the Consumer Price Index. (Which makes you wonder why CPI and BPI track so closely…)
I notice that this Sumner item caught Karl Smith’s eye as well, and he points out rightly that constructing indexes is always a problematic venture:
basically anyone with MS Excel and a rudimentary knowledge of the subject matter in question can create a workable index
But, still, based on this quick look, at least since the housing peak in 2006, Scott’s right that CPI has been looking like an especially dicey measure.
Cross-posted at Asymptosis.