The Boskin Commission Report and Problems with the CPI

In 1997, Michael Boskin chaired a committee appointed by the Senate Finance Committee to examine problems associated with the CPI. The motive was to see whether the CPI accurately reflected the "cost of living", since many transfer payments such as Social Security are indexed to the CPI. To over-compensate for the cost of living is to put strains on a system that will predicably be burdened in any case. Boskin concluded that the CPI overstates the "true" cost of living increases by as much as 2 percent, and although there has been a good deal of debate about the size of the overstatement, there is no doubt that some bias is present. The commission distinguished four sorts of bias:

  1. Substitution bias

    The CPI consumption basket is fixed in the base year, so that it is assumed that consumers go on buying the same basket no matter what the changes in prices are. Actually, consumers protect themselves against some of the effects of inflation by substituting towards relatively less expensive goods.

  2. Formula bias

    The data on the original consumption basket was originally collected by looking at consumer expenditures and dividing by quantities, so that in the formula, prices are a calculated value rather than an observed value. If an item happened to be on sale when the BLS made the observation, the sale price would emerge from the formula, leading to an unusually large price increase when the good returned to its normal price. This may sound trivial, but it is probably as big a source of error as substitution bias.

  3. New outlet bias

    The original data for the CPI consumption basket was collected at a time when discount stores (Wal-Mart, etc) were not as common as they are now. The BLS statisticians keep going back to the original stores to get updated prices, neglecting the new and cheaper outlets. This is now being corrected, but was a source of bias.

  4. New product and quality bias

    Personal computers were a very small part of the consumption basket in 1982 (the base year for the CPI and the year of IBM's first PC). They still get the weight in the consumption basket that they then had. It is also difficult to fully account for changes in quality -- PC's have obviously improved, as have autos and even apartments (the average apartment is a good deal bigger and much more likely to have central air conditioning now than in 1982.) Part of any price increase reflects a payment for improved quality, not the same good.