Measuring Inflation: The CPI and Houses

By Adam Ozimek

Semi-retired blogger Scott Sumner commented on Karl’s post about how the CPI calcuates inflation of owner-occupied housing. So long as he keeps commenting around the econ blogosphere, we can effectively keep him from retirement by hoisting these comments. Here is Scott on Angus, to whom Karl was responding:

I read his argument differently. Reading between the lines, here’s what I think he meant:

1. The Fed doesn’t care about the “cost of living” per se, they care about the price level because supposedly a stable price level produces macroeconomic stability.

2. The price of new homes is an important part of the overall price of goods and services produced in the US.

3. If the Fed stabilizes a price index, that index should include the price of new homes.

4. It’s fine if we have a cost of living index that excludes the price of new homes, just don’t have the Fed target that index.

I don’t see your argument as being inconsistent with what he wrote.

I think Scott’s point here is that there’s a distinction between what should go in a cost-of-living here and what the Fed should target. I’m pretty skeptical of this idea, so I’ll run through some reasons why.

One conceptual problem is that house prices reflect value of the flow of housing services consumed and the investment value of the home. As Karl points out, why should housing investments be counted in a cost-of-living index while other investments are excluded?

A 2005 survey of OECD countries (pdf) found that 13 out of 31 statistical agencies used the owners-equivalent method for their CPI, and the next most popular method, used by 9, was to just leave it out. Only two countries, Australia and New Zealand, utilize the acquisition approach, which includes house prices in the same way you would with non-durable goods. Below is a graph from The Economist of the ratio of house prices to rents from Q1 2000 to Q4 2010 for Australia, the U.S., and New Zealand (the United States is the lowest one at the end, it’s a little hard to tell with the colors).

(Click chart to enlarge)

While the other two countries have thus far avoided a massive crash, they didn’t avoid massive appreciation. You could argue that this just means the changes in prices in these two countries was real and driven by fundamentals. But given these massive differences in the real paths of price-to-rent ratios, what level should the Fed target? How do they distinguish price appreciation drive by real changes in cap rates from pure nominal or speculative inflation? I’m not sure how Australia and New Zealand handle this.

Another problem with the acquisitions approach is that the sale of a home is frequently an exchange between households, and so a sold house is both a cost and a revenue for the household sector. According to the BLS, when statistical agencies use the acquisitions approach they control for this by only looking at home sales to the household sector from other sectors. In practice, this usually amounts to new housing units. This may reflect the cost of goods and services produced in the U.S., but clearly doesn’t represent a measure of the cost of goods and services being consumed by households. How would the Fed weigh the trade off between these two price measures? Say demographics or something else permanently shifts average cap rates down and thus real house prices up while the cost of living is unchanged. If the Fed tries to tamp the house price inclusive price index down won’t inflation and thus nominal GDP be too low? Of all people I’d expect Scott to worry about this.

Let me end by saying that I am pretty skeptical about the inclusion of house prices as a component of a cost-of-living index. I’m also skeptical about the idea that the Fed should target a non-COLI price level that includes house prices, but less so than in the former, and I’m not strictly closed to the idea.