Tuesday, November 30, 2010

CPI - Confusing Price Index

I am fairly busy on the home and work front at the moment so not a lot of time for a new (meaningful) post.  I wanted to post a link to a wonderful John Mauldin essay recently dissecting US inflation data

I am working on a similar piece for the Australian CPI data - which I believe from initial research to even more severely understate price inflation than the US in its imputed/equivalent rent component.  Note that I personally consider any version of CPI to simply be one indicator of price inflation, and never exhaustive.  Moreover, the financial crisis has shown the need to move beyond a total reliance on consumer price inflation in determining monetary policy - in other words to also consider asset price inflation and (more importantly) monetary inflation (i.e. the supply and 'value' of money).
I quote the key passages in Mauldin's essay below, but the overall thrust of his argument is that by ignoring the cost of housing in terms of asset prices, and replacing it with an imputed rent version of the cost of housing, CPI data were depressed through throughout the last decade.  This (likely) contributed to the Fed's policy of keeping the cash rate so low in the early 2000's (and therefore the subsequent house price boom and bust). 

Powerful stuff!!

Mauldin starts by questioning the current suite of inflation data showing moderate positive inflation in the US, and then goes on to say:

...What if the way we measured inflation was flawed in some regard? Let's play a thought game. Back in the early '80s there was some consternation about using the price of houses as a way to calculate inflation. Read this paragraph from the BLS website (emphasis mine):

"Until the early 1980s, the CPI used what is called the asset price method to measure the change in the costs of owner-occupied housing. The asset price method treats the purchase of an asset, such as a house, as it does the purchase of any consumer good. Because the asset price method can lead to inappropriate results for goods that are purchased largely for investment reasons, the CPI implemented the rental equivalence approach to measuring price change for owner-occupied housing. It was implemented for the CPI-U in January 1983."

Homeowner equivalent rent is 25.2% of the input when they calculate the Consumer Price Index (CPI). Thus it makes a big difference how you calculate the price of housing. It is extraordinarily difficult to find historical data on homeowner equivalent rent in the BLS database. You can find "shelter costs," which include energy, insurance, etc. and are 41% of the CPI, but for our purposes today I want to focus more narrowly.

I did find an old release that shows the index value for the year 2000 to be 198.7 (http://www.bls.gov/cpi/cpid00av.pdf). The index value as of this October was 256.8. That means the rise in housing costs over the last decade was about 25% or roughly 2.5% a year, although in the last few years that number has gone deadline flat. And you can see that in the graph of total housing costs below.

CPI 10-Year Chart

But house prices went up by more than double that amount, and about 65% in the seven years from 2000 to 2007 (back-of-the-napkin estimate). That is an asset inflation of about 9-10% a year.

US House Price Index 1975-2010

What if we had been using actual home prices as the measure of inflation? Let's look at the year-over-year change in inflation for the last ten years:


Homeowners equivalent rent is 25% of the index. So take the home price rise, divide it by four, and add it to the inflation index. Inflation in the middle of the decade would have been running 4-5-6% and in 2005 would have been over 7%!

Would rates have been kept low "for an extended period of time" if we had still been using actual home prices? I rather doubt it - alarms bells would have been sounding. The Fed would have been leaning into the rise in "inflation." Thus no housing bubble would have developed. And then no credit crisis.  And the difference all stems from how you measure inflation. These details matter.


In the 1980s the BLS (under Reagan, so not a liberal plot!) decided a home was an investment and not a roof over our head. It also conveniently allowed for lower official inflation, which is what Social Security and other government programs are tied to. But that change had significant unintended consequences.


  1. The Australia measure of CPI is pretty average too, and not transparent at all.

    Here's a particularly useful paper

  2. Thanks Cameron. I appreciate the link. I had a quick look at the links you posted on housing shortage earlier and they are most helpful also. You have done a lot of good work on the topic. I am trying to think of a way of breaking it down to a very simple analysis that shows that (at a minimum) it is not as simple as a story of 'chronic' shortage.

    I think it all comes down to illustrating that there can be excess supply in the market if prices are too high, as is currently the case, and that this is obvious when you consider other assets like sports cars (a lot of demand, but not enough sports cars to meet all of it! Must be a shortage!). Furthermore the methodology of assuming all forecast future demand must be met with equivalent units of supply, regardless of price, would be considered ludicrous in any other asset market.

    Thanks again!