This Blog Has Moved!

My blog has moved. Check out my new blog at

Your Ad Here

Tuesday, July 1, 2008

The CPI is Biased

I'm astonished when people say "The CPI is a fair and unbiased measure of inflation". That shows a level of political sophistication equal to the statement "We have always been at war with Eastasia."

I tried searching on the Internet for exactly how the CPI is calculated. I could not find the EXACT FORMULA. The details are carefully hidden, on purpose.

More than 10 years ago, the CPI was calculated using a basket of goods HELD CONSTANT OVER TIME. Now, the CPI is calculated using a complicated re-weighting formula. Some people have recalculated the CPI using the old methodology, and they get inflation figures that are closer to the ones I use (M2 and gold).

There is a political incentive for CPI to understate true inflation. Many payments, such as Social Security, are indexed to the CPI. A Congressman was grilling Ben Bernanke about inflation. He said "I only care about CPI-inflation. I am not discussing other measures of inflation."

Here is an illustration of the "dirty tricks" that are used to bias the CPI downwards. Suppose that one year, the price of beef rises 20% while the price of chicken rises 2%. The CPI calculators will say "People will switch from beef to chicken. Therefore, the 'meat' component of CPI rises 2%." Next year, the price of beef rises 2% while the price of chicken rises 20%. The CPI calculators will say "People will switch from chicken to beef. Therefore, the 'meat' component of CPI rises 2%." In this manner, the CPI understates the true inflation rate.

Money supply inflation does not lead to uniform price inflation. Therefore, the CPI can be manipulated via reweighting tricks.

The complicated reweighting formula causes the CPI to understate inflation. Even worse, the weights are calculated retrospectively. On January 1, the CPI calculators don't say "These will be the weights for this year!" Instead, on December 31, the CPI calculators look at last year's prices and say "These are the weights we chose for last year!" This allows the people who choose the CPI weights to have selection bias.

I don't know if the CPI is reweighted annually or quarterly. My guess is quarterly, because that maximizes the amount of abuse.

If you insist "CPI is a fair measure of inflation", I'll offer you the following trade. I'll pay you CPI+2%, and you pay me the return on a gold investment. Is that a fair trade? If you believe the CPI measures inflation properly, then this should be a great deal!


Anonymous said...

Shadow Stats has a white paper on CPI and other measures. Have you considered test driving a subscription?

Anonymous said...


This subject comes up often as it should (what inflation really is). I tried to explain to my friend that last year when you purchased 100 dollars of groceries (as example) this year same groceries would cost 104. Now I saw a report for 2006 I think where some agency was quoting a government source as inflation was 4.8 or something. Now considering the increase in the costs of gasoline (doubled) in bush days and therefore the snowball of rising food and other prices on retail and industrial items. Where do you think inflation really is. I was thinking 6 to 7 percent. But in the last 3 years I'm not sure that is high enough...

Luke said...

Do you have a reference about the chicken-beef trick? I'm not too familiar with the CPI, but I was under the impression that the CPI can't be cheated this way. Further I was under the impression the CPI reflected actual buying habits.

Everyone knows the CPI is flawed. There are many ways to measure inflation and they are all flawed. Most investors trade oil or gold rather than the CPI, for example. At the least, it misses important items such as commodities. At the extreme, one cannot even define inflation, because there exists no asset of constant value.

However, it does serve a practical purpose of measuring a partial "cost of living". Other factors, such as real estate, energy, and commodities, need to be included for the whole picture. It's not like the CPI is the only number the Fed looks at.

Anonymous said...

Dear FSK,

I believe your chicken-beef example does not happen in reality. True, chain-weighted calculation of CPI/GDP does underestimate inflation compared to base-year calculations. But most economists believe that chain-weighting method is much more accurate than the base-year method. This is true especially in very recent times when consumption of certain goods rose or fell steeply in very short periods of time. (i.e. laptops, cell phones, MP3P, etc.).

As for manipulating numbers for consumer behavior, the calculators depend completely on the past data. After all, CPI is an indicator of the past, not the present. Sure, CPI is flawed. But any rigorous calculation of inflation is flawed.

This Blog Has Moved!

My blog has moved. Check out my new blog at