Measuring Inflation: The Difficulties Posed by Housing

Should the Fed continue to raise interest rates? Should they raise them faster? Typically economists presume that the central bank cares about two things when deciding how high interest rates should be: the output gap, which tells us how far the economy is from some notion of “full employment”, and inflation. The Taylor Rule is one slightly more formal way of expressing the intuition that these two indicators should tell us whether current interest rates are too high, too low, or just right.

But measuring these two variables is not easy. Last week PGL did a nice job discussing some of the difficulties in estimating exactly how large the output gap is in the US economy right now, with a follow-up over the weekend. Even among economists who think carefully about this issue, there’s substantial disagreement about how large the output gap is right now.

But it turns out that it may be at least as difficult to estimate the other key factor affecting monetary policy decisions: the inflation rate. (Warning: the rest of this post is a bit long and wonky.)

There are several potential problems with the way inflation is measured in official statistics. Measuring average price changes of thousands of very different goods and services is an enormous task, after all, and there are no clear ways to resolve many of the difficulties involved.

But recently, the US government’s official measures of inflation in housing have drawn increasing criticism specifically for understating the impact of the rise in house prices on the cost of living. According to the BLS’s Consumer Price Inflation measure, the cost of shelter has risen by only 2-4% per year over the past couple of years. Yet we all know that the price of houses has risen by far, far more than that recently. According to the widely-cited House Price Index, house prices have risen by 12.5% over the past year across the US as a whole, and risen by 20% or more in many metropolitan areas.

The discrepancy is explained by a number of factors, but the biggest is simply this: the BLS measures the price of housing services as the rent that a household would have to pay for an equivalent living space. And inflation in rents has indeed remained low in recent years, despite the run-up in housing prices.

The following pair of charts illustrates. The first shows the price index for shelter in the four Census regions of the US over the past 20 years according to the CPI’s “rental-equivalent” measure. The second chart shows an index of the monthly payments that an individual would have incurred in any particular year when buying a house of a constant level of quality. (In this case, the benchmark is a house with all of the features of the average house sold in 1996.)


Note: Constant-quality house price index from Census. “Cost of buying a house” reflects the monthly payment needed to buy an average 1996 house using the average 30-year mortgage rate for the year. 2005 data assumes a 5% price appreciation over 2004 prices, and uses the average Jan-Jun 2005 mortgage rate.

Clearly the two measures are wildly different. Much of the variation in the second chart comes from changes in the interest rate, which changes the monthly mortgage payment entailed by the home purchase. But over the past few years the 30-year mortgage rate has remained roughly constant, and rents have risen only modestly, while the cost of buying a house has actually risen substantially. That’s the effect of the recent sharp rise in house prices.

This means that the inflation experienced by people who have bought a house in the past couple of years is probably considerably higher than the inflation measured by the CPI. It’s worth noting at this point that typically 40-50% of households in the US move to a new house during any 5-year period, so the fraction of the US that has experienced this greater-than-measured inflation is not insignificant.

Note that the BLS’s decision to use the “rental-equivalent” cost of shelter instead of the actual cost that individual’s incur in their housing makes sense in a lot of ways, for example by keeping temporary fluctuations in interest rates from showing up in the CPI. And when compared over many years (such as in the charts above), the CPI measure actually have may overstated the inflation rate in shelter.

But recent history highlights this potential drawback of the “rental-equivalent” measure of housing inflation: it assumes that the cost of renting and owning move at least roughly in step with each other. And that has clearly not been the case in recent years.

This makes the job of estimating the inflation rate in the US even tougher than usual. While official inflation measures show rates of about 2-3% over the past year, that certainly doesn’t correspond to the experience of a large fraction of the US population. Which just makes the Fed’s job of balancing inflation concerns against the output gap that much tougher.

Kash