Accuracy of the CPI

The CPI-U (Consumer Price Index for Urban Consumers) is used to calculate the inflation adjustments on Series I savings bonds, TIPS, and most inflation-indexed annuities; the very similar CPI-W is used to adjust Social Security payments. In assessing the value of inflation-indexed assets, it is obviously important to know whether the CPI is fairly and accurately calculated. This is a controversial topic.

Some reasons for believing the CPI is understated
These are some frequently-voiced reasons for thinking that the CPI figures are understated:


 * Several changes have been made in the definition of the CPI: the change in measuring homeownership costs from an asset-based approach to owner equivalent rent in 1983; the adoption of the use of the geometric mean for averaging prices within narrow categories in 1999; and the use of hedonic regression models, starting in 1988. Each of these changes had the effect of reducing the CPI at the time it was introduced.


 * A website, http://www.shadowstats.com publishes charts showing their alternate calculation of an "SGS Alternate CPI," which is currently about 6% higher than the official CPI-U. (It is not clear what methodology is used for this calculation).


 * John C. Bogle has said that he thinks the CPI is understated. In his 2010 book Don't Count On It, p. xx, he writes: "Our government also contributes to this distortion of reality. David Einhorn, leader of Greenlight Capital, has observed that 'over the last 35 years, the government has changed the way it calculates inflation several times.... using the pre-1980 method, [inflation] would be over 9%, compared to about 2% in the official statistics.'"


 * The CPI does not take account of food or gasoline pricing. (This is simply incorrect, but it is nevertheless often stated as a reason for not trusting the CPI).


 * It costs the government huge sums of money to pay CPI-linked adjustments in TIPS, I bonds, and Social Security benefits. Since it is also the government that computes the CPI statistics there is a clear conflict of interest, and a clear motivation to understate the CPI.

Some reasons for believing the CPI is reasonably accurate

 * The government isn't a monolith. The CPI is calculated by the Bureau of Labor Statistics. The parts of the government that are affected by CPI calculations would be the Social Security administration and the Treasury. Different bureaucracies within the government tend to be independent and protective of their own turf. The BLS is part of the Department of Labor. It's not obvious why the BLS would want to carry water for those other agencies.


 * The calculation of the CPI is transparent. It doesn't emerge from a secret black box. On the BLS website you can find out more than you'd ever want to know about the composition of the market basket, how owner-equivalent rent is computed, and what past CPI figures would have been if calculated according to current methods. The CPI could be openly fudged--for example, the President could order the BLS to freeze the CPI at some fixed number--but fudging couldn't be concealed.


 * Groups with a direct interest in the CPI, such as the American Association of Retired Persons (AARP), are not complaining about the CPI. A direct email query to AARP elicited a reply that they believe the CPI calculation is accurate; their only quarrel with it is that CPI-W doesn't reflect the expenses of the elderly well. They want Social Security indexed to a market basket that more closely reflects the expenditures of Social Security recipients. The BLS is currently calculating such an index experimentally under the name CPI-E. It's not hugely different from CPI-W.


 * The CPI has been calculated by the Bureau of Labor Statistics since 1919. During that time there have been some periods of high inflation, notably during the Nixon era, yet it is not commonly suggested that the CPI was fudged during those periods. For example, the CPI continued to rise during the Nixon price freeze, showing that the price freeze was ineffective, even though this was presumably embarrassing to the government.


 * Some sources that have complained about the CPI being fudged maintained that the use of owner-equivalent rent was deceptive, and that the CPI should be calculated on the basis of home prices, not owner-equivalent rent. If you accept this, then the CPI is currently significantly overestimating inflation.


 * Treasury securities are a popular investment of people of all shades of the political spectrum who want to minimize risk. As of 2007, for example, two notable conservatives were reported to have large investments in Treasuries. One of them in particular had "$2.5 to 11 million" of his "$21.4 million to $100 million" invested in the Vanguard TIPS fund.


 * The Bureau of Labor Statistics itself calculates a series known as CPI-U-RS, which tracks the CPI from 1978 to the present when computed by current methods. CPI-U-RS is lower than CPI-U, but only slightly. Over the period 1978-1998, the annualized difference was 0.45%. The stated theory behind the changes is, of course, that the older method was overstating inflation and that the changes were a correction. If, however, you believe the older method was the correct one, then the current method is understating inflation--but only by about 0.45%.


 * The MIT "Billion Prices Project" estimates inflation for many countries, completely independently from the Bureau of Labor Statistics, by automatically checking online prices. In the case of the U. S. their estimate is quite comparable to the Consumer Price Index.




 * Although the BLS (Bureau of Labor Statistics) does use various techniques for adjusting components of the CPI, they work in both directions. For example, the BLS has been criticized in a paper, CPI Bias from Supercenters: Does the BLS Know that Wal-Mart Exists?, which says "Our estimates are that the BLS CPI-U food at home inflation is too high by about 0.32 to 0.42 percentage points." (In this case, the claim is that BLS methodology underestimates the effect of prices at places like Wal-Mart by implicitly assuming that the lower prices reflect a different and lower quality of goods.)