The Atlanta Fed's macroblog provides commentary and analysis on economic topics including monetary policy, macroeconomic developments, inflation, labor economics, and financial issues.

Authors for macroblog are Dave Altig, John Robertson, and other Atlanta Fed economists and researchers.

« A Nice Cup Of Tea For The EU | Main | Measured Pace Watch »

July 13, 2005

Do Rising House Prices Distort Inflation Measurement?

Kash at Angry Bear, in a thoughtful piece posted yesterday, suggests that the answer is yes. When confronted with questions about inflation measurement, my habit is to turn to my colleague Mike Bryan, who knows a heck of a lot more about this topic than I do.  Here is what Mike had to say (lightly edited):

1) The idea of the CPI is to measure, as closely as possible, the average person's cost-of-living. Although the BLS doesn't exactly claim that this is their intent, for all practical purposes, that is indeed what they are trying to do. As such, the rental equivalence approach is the "right" approach because it measures the cost of owning a home to a homeowner.

What is the cost of owning a home to someone who already owns a home? Well, it's the opportunity cost of living in your home rather than renting it out. So, in this sense, there is no "mismeasurement" as Kash suggests. The BLS does it correctly (which is always a good assumption since these people are really quite good at what they do.)

Remember, housing is an asset, and for most, rising home prices isn't a direct cost. In the same way, would I ask the BLS to add rising stock market values in their cost of living index? No, of course not. (But rising equity values MAY be a sign of rising inflation, and I will return to this at the end.)

2) There may actually BE a mismeasurement here, but it comes from the fact that, despite the great efforts of the BLS to match the owner-occupied housing stock with a comparable sample of rental homes, rental homes tend to be in less desirable neighborhoods without the same amenities.  As a result, when the housing market "heats up" there could easily be a quality shift that makes the BLS sample less conformable to the true owner-occupied market. Work has been done in this area, and the BLS is well aware of this potential "bias."

3) I also have a problem with this bit from Kash's post: "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." The CPI weighs housing on the basis of what the average person spends on housing (implicitly via opportunity cost) and not what a new, first-time homeowner spends. Again, if you are already a homeowner (without appealing to opportunity costs), you have NO cost-of-living rise as home prices rise. You bought your home before prices went up. (And remember, you have to be a first-time buyer, or the rising gains from selling your old home will largely offset the rising costs of the new one.)

Now, I don't know exactly what the proportion of the housing market these people represent, but I would bet they are quite small. So, by the approach implied by the blog, one would weigh the CPI housing MUCH smaller than the 23 percent of the basket it currently commands. In fact, I wouldn't be surprised if the downward revision to the CPI from the reduction in the component's weight had a much larger impact than the upward revision to the CPI from plugging in actual new home costs.

4) Now, let's take on the issue I think is most important: Is a cost-of-living approximation like the CPI a good "inflation" measure? I think not and have written on this in many places -- here, for example. The sound bite is this: The Federal Reserve cannot control the cost-of-living. This is REAL, and as such, it is influenced by real factors that are outside of the Fed's control, like oil, droughts, acts of war, and such. Inflation is a monetary phenomenon that, in the end, leaves ones cost-of-living unchanged.

I think it entirely possible that assets can provide a leading signal of a generalized, monetary inflation. But admittedly, this is mostly an article of faith and not science--the statistical link between these cost-of-living measures and asset prices is very tenuous. And economists have not been able to map the link that leads from excess money growth to a generalized inflation. So until that day, I think one who ignores asset prices when looking for signs of rising inflation does so at some peril.

So I would tend to take these rising housing price movements seriously and judge them along with a host of other asset measures when thinking about Fed policy. But of course, now we are a long way from the thread of this discussion: is the CPI mismeasured because of the rental approach? I say no--well, not in the way suggested by at Angry Bear.

July 13, 2005 in Housing , Inflation | Permalink


TrackBack URL for this entry:

Listed below are links to blogs that reference Do Rising House Prices Distort Inflation Measurement? :


"Inflation is a monetary phenomenon that, in the end, leaves ones cost-of-living unchanged."

If my salary or pension lags (consumer good)inflation, it costs me more of my income to buy the same basket of goods. Seems like a stretch to define "cost of living" in a way that doesn't capture this.

Posted by: Mark Sullivan | July 14, 2005 at 02:27 PM

Is the cost of "financing services" somehow included in inflation measures? If not, it seems like it could be a significant source of "disconnect" between CPI and perceived inflation.

At one level, one could argue that CPI should measure the price of non-financial things from the perspective of an "average, unleveraged" consumer. But an "average, unleveraged" consumer would be a hypothetical, artificial beast. The real average consumer requires, on net, financing, which is a service. She pays for that service in the form of interest. Over the past few years, a financing-aware CPI might on the one hand show the price-per-unit of financing services (i.e. interest rates) to have gone down, but the weight of financing services in the average consumer's basket to have gone up, perhaps dramatically (interest on housing and consumer debt). How would including this (if it is not already) affect the reported CPI?

(I think that a CPI that included financing services would need to include gross interest paid, not interest net of investment returns. Investment returns belong in measures of average income, which if growing might take the sting out of a high CPI measure.)

Posted by: Steve | July 14, 2005 at 02:50 PM

On the issue of the relationship between monetary policy and asset prices (with housing as the focus), I refer you to a paper out of the Atlanta Fed ("Examining Contributions to Core Consumer Inflation Measures" by Bauer, Haltom and Peterman, 2004). The paper does not address the relationship, but it does a nifty job of showing that as interest rates declined after 2000, house price increases accelerated and rental rates did not. These lower/unchanged/weakly rising rental rates are what is used to calculate part of the CPI housing component. And recall that the FOMC was cutting the fed funds rate for much of that time. So even as house prices were accelerating, the housing component of the CPI was not only not tracking house prices, but was, at times, moving in the opposite direction. This may be the source of some of the data-based frustration with the CPI.

But sorting out whether a cost-of-living measure is also an inflation measure is an issue I will leave to the index number experts.

Posted by: Mary R. | July 14, 2005 at 02:56 PM

When Dave asked my opinion on this issue, I didn’t expect I would have to work so hard! Let me personalize responses to some of these posts (in order).

Stefan, I have written several papers dating back to the late 1980s that have attempted to put asset prices into an inflation statistic, including approaches very similar to that done by Reis. I think this work has great potential, but it’s a long way from being fully cooked, if you know what I mean.

Bailey, I do indeed believe that gauging cost-of-living changes for wage adjustment was a motivating factor for the establishment of the CPI many, many years ago. This is why the measure is produced by the U.S. Department of Labor, not Commerce, where you might more naturally expect to find an inflation statistic. The Labor Department does not ask people what they could rent their homes for when computing changes in the owners’ equivalent rent measure—they use rents from rental properties that have the same characteristics as the homes people own and live in.

Steve Kyle, the CPI does not attempt to measure “out of pocket” cost of living. As I said, about ¼ of the index is the implicit rent people who own homes “pay” by living in their homes rather than renting them out. You can easily subtract this component out of the CPI if you wish. For the large proportion of people who own their homes, rising home costs do not represent an out-of-pocket cost. In fact, it represents a net addition to their wealth. And I don’t follow your point about ARMs and other issues of the cost of financing a home. These costs are falling, not rising (and some would say this is the reason why home prices are rising so rapidly.)

Bailey, everything the statistics jocks at the BLS do is well documented and thoroughly debated by good academics. Yes, the issue of CPI measurement is a political football, and every year the various adjustments made using the index, including Social Security, are the subject of debate. But I think it unfair (and inaccurate) to imply that the methodologies used by the BLS are altered for reasons of political favoritism. The CPI is far from perfect, but I believe the BLS is an honest broker of data and should be given credit for such.

Kash, if I accept your figure that 50 percent of people move in a five year period, that means roughly 10 percent move in any year. That drops the 23 percent weight afforded owner-occupied housing by more than half. Now, of that amount, what percentage are NEW homeowners? Remember, those people who own a home, sell it, and move into another home, get both the benefit and the cost of the higher home costs (the added gains of the one they just sold must be netted against the added costs of the one they buy.) So, it seems clear to me that, by this approach (which I do not recommend in any event,) the weight of homeownership costs in the CPI would be cut to a very, very small number.

Dave Iverson, the search for a good inflation forecasting model is one of the grails of macroeconomics. No economist, on either side of the Atlantic, knows the transmission of central bank policy to retail price measures. The common refrain is that the lags between the two are “long and variable” which is to say, “we don’t know.” Milton Friedman suggested that the transmission might go from money, to assets, to resource costs, to retail prices, and others have hypothesized other paths that lead from a central bank policy to your grocery bill, but NONE of these has, as yet, proven very reliable. I would not say that the U.S. is disinterested in asset prices and central bank policy. The Federal Reserve Bank of Chicago held a major conference on this topic and published a volume from it just a few years ago. Also, I would note that the harmonized price measures that the ECB uses to gauge inflation don’t include ANY housing costs, so they too have work yet to do in this important area.

Mark, of course, you are right. If your wages are fixed, and prices are rising, you get hurt. But the key to your statement is the fixed wage. My point is that a monetary inflation should ultimately affect ALL values, including incomes, the same.

Steve, everything about an “average” is a hypothetical, artificial beast! The average home has 2.5 kids, after all. That’s the curious nature of any average.

And Mary, if I live in a home, and its value is rising, in what sense are my costs rising? I bought it before the prices went up. The only sense is that I could rent it out for income, but I choose to live in it myself. I am my own renter. This is what economists call an “opportunity cost” and this is the logic of the CPI rental equivalence measure. And it’s good logic too. Now, what you are pointing out is that the rental market is relatively soft and therefore, that my opportunity cost has fallen. Yep, it has. But I fail to see any “mismeasurement” here!

I hope I have added more light than smoke here, and thanks for the space Dave.

Posted by: Mike Bryan | July 15, 2005 at 12:34 AM

in answer to your answer, my point is exactly that (as you say) the CPI doesnt measure out-of-pocket expenses. Where we disagree is that I think it should do that particularly since that is how it is often used and also because that is how most people (granted most of them arent economists) understand it. My point is that opportunity cost is something we all learn in econ 101 as the true cost of things that dont have a market price, but that in the case of housing there actually IS a market price for the out-pocket-cost of home ownership. It is extremely messy compared to simply using the rental equivalent but it does exist. And it is exactly because it DOES exist that people feel the pain not measured by the CPI. There are many issues involved in measuring it exactly but all of these issues are typical index and weighting issues such as we have all dealt with before. A footnote: My point about ARMS was that the out-of-pocket costs are variable, not that they are necessarily always higher - That interest rates are low is one reason for my statement that quality-adjusted cost of home ownership might not give us an answer as different as some might think

Another note: These index number problems are tame compared to anyone with experience playing with Brazilian time series data from the past few decades. After doing this for a while I have come to believe that I am not enough of a masochist to do it again. Maybe that's why I am not as impressed with the messiness of doing our CPI differently as some are.

Posted by: steve kyle | July 15, 2005 at 04:32 AM

Mike Bryan, THANKS for your time & obviously well-considered thoughts. I will give them the serious reflection they deserve before responding. Have a great day.

Posted by: bailey | July 15, 2005 at 09:43 AM

Mike, thanks for taking the time to do the work. Your occasional guest forays on macroblog are always informative.

This is probably a dead comment thread by now, but if you stop by, I'd still be interested in your take on whether or how financing as a service cost ought to figure into inflation measures.

Thanks again!

Posted by: Steve | July 16, 2005 at 02:44 AM

Hi Steve,

There are several ways statisticians might try to measure the cost of using an asset (which is what we are talking about here.) The two most popular approaches are rental equivalence and user cost. In theory, and assuming that markets work competitively, the two approaches should yield the same number…well, in theory. OK, you obviously know the BLS uses the rental equivalence approach when computing housing costs in the CPI, but a few years back they used the user cost approach. The user cost method requires them to add up all bits and pieces that go into using a home, including depreciation, interest payments, taxes, the whole shebang. The BLS frequently got criticized for this approach and what bothered economists most was the treatment of interest rates. See, the CPI is (and MUST be) a fixed marketbasket. So, in this case, they have to assume that the amount people finance is constant. But, of course, the amount people finance is anything BUT constant. As interest rates move up and down, the number of people taking out credit fluctuates widely. So when rates would spike, and the public cut back on debt acquisition, the BLS had to assume that people continued to take out mortgages in the same proportion as when interest rates were low. This is an age-old problem in price index measurement called “substitution bias” and it was an especially big problem with mortgages. The problem was magnified by the large weight of housing in the CPI and the BLS was constantly under fire that they were grossly overstating inflation by using this approach. If I remember correctly, this is why they changed to the rental approach. Most nations who bother to put asset usage costs into cost-of-living statistics use the user cost approach, but I suspect they do so not so much because they think it a superior measure, but because they don’t have a deep enough rental market on which to produce a rental equivalent measure. My opinion is this: The rental equivalence measure isn’t perfect, but the problems of the approach pale in comparison to user costs. If you ask me (and you did) the area most in need of the BLS’s energies is the measurement of service quality, not asset usage costs. How does a price statistician separate out cost increases from quality increases for things like medical care, or financial services, or education, etc.? This is a major statistical problem for the BLS and will become an even greater problem as our economy continues to generate more of its income from services. It will be a problem in inflation measurement long after the rental/housing price controversy is long forgotten. Hope this helps.

Posted by: Mike Bryan | July 17, 2005 at 10:55 PM

The BLS web site confuses its own rental equivalency methodology. It includes statements: "... as of 2001 ... both renters & homeowners are included in CE sample ... In fact, homeowners constitute roughly 63% of the CE sample in Urban areas ... Expenditure weight in CPI for renter equivalence is obtained by directly asking sampled owner households: "If someone were to rent your home today, how much do you think it would rent for ...?"" It then shares "background" of CPI changes in rental calculations, ending in 1997 with a statement that directly refutes the above method. So,
I'll accept that Mike Bryan's explanation & assume the BLS web page
needs some updating.
I don't accept that elimination of asset price method for primary residences in the early '80s was non-political. It seems ludicrous to compare a resident's purchase of his primary residence to an investment in the stock market. One obvious difference is that we live in our primary residence, & only the home owner is responsible for setting its value. Others may kibbutz about the home's worth, but it's only the homeowner alone who sets determines the final price. If the Gov't truely believed homes were an investment commodity like stocks, they would allow us to add them to our 401k plans!
I wasn't attacking the BLS, indeed I was defending Abraham, the BLS head at the time. I was chastizing Academic Economists for not coming to Abraham's defense. Read K. Abraham's response to the Boskin Commission's findings, read what Barry Bosworth of Brooking & Dean Baker of CEPR had to say about the last 1 point reduction in CPI. My point was & is that after Greenspan set the table & Feldstein & others (with a seemingly) polital agenda jumped aboard, those who could have made a difference chose to walk away. I believe that that last 1 point reduction in CPI WAS enormous in influencing present & future fiscal & monetary policy, and especially enormous in influencing peoples' perceptions about how well our country's economy is.
In an '02 paper Mike Bryan went to lengths to explain the difference between cost of living & inflation. How many people in the country can make a distinction? I'd have a better chance if the FED would openly share with me what the "money" is that Mike repeatedly refers to. The Fed's stated goal is to become MORE transparent, yet our most important economic indicators have lost their way & the Fed jumps amongst indicators with the dexterity of a schoolkid playing hopscotch. It is incomprehensible, in 2005 with all the data we have to draw from, that we rely upon hogwash surveys for defining & collecting data we use to set the economic course for our country. Now think about the enormous changes we've adopted in our move to a glogal business world & consider the complexities if we can't even agree upon common definitions of the most basic statistical criteria.
I think it was Dewey who rightly proclaimed some 80 years ago that we're living in a relative world. It seems clear we've moved a long way to an agreement that all things are not economic, they're political, & that the Academic Economics community is just the latest group to lie down & be run over.

Posted by: bailey | July 18, 2005 at 02:07 PM

Hi Bailey,
Yeah, you are correct, the BLS website is a bit hard to follow. I understand it as the following: The BLS asks homeowners what they think their house would rent for when determining the WEIGHT of housing in the CPI, but they go to actual rentals to compute how the rents change over time. I believe Chairman Greenspan has taken issue with the household survey used to compute CPI weights because he thinks people consistently overestimate certain values (like how much their house is worth) and underestimate other things (like how much alcohol they drink.) The weights used to compute the PCE measure come from expenditure data and are thought by some to be more accurate--the PCE weighs owner-occupied housing much less than the CPI (about half as much if I remember correctly.) And I agree with you that the language economists use can be confusing, if not sloppy. Inflation, price increase, cost of living increase, etc. are sometimes used interchangably, when they shouldn't be. The distinction between a cost-of-living increase and inflation is this: cost of living changes are "real", while inflation is purely nominal. Consider this hypothetical example of a doubling of oil prices. Suppose I cut oil production in half and the price of oil doubles. I would call that a cost of living increase, not inflation. It is real, your life is harder, and because the Fed doesn't produce oil, there isn't a thing in the world they can do to make that pain easier. If, however, the Fed were to double the amount of money in circulation, and the price of oil doubles as a result, that is inflation. This, of course, the central bank CAN remedy. I believe that the accurate measurement of monetary induced price increases--inflation--require statistical approaches that are fundamentally different than those used to measure cost-of-living increases--but this is a controversial position on my part and certainly not nececessarily endorsed by anyone other than me. My view on this is easily found in my publications. Thanks for the discussion.

Posted by: Mike Bryan | July 18, 2005 at 04:19 PM

Post a comment

Comments are moderated and will not appear until the moderator has approved them.

If you have a TypeKey or TypePad account, please Sign in

Google Search

Recent Posts



Powered by TypePad