Eliminating Bias in the U.S. Consumer Price Index

• Pages: 65-77

Page 65

The possibility that the U.S. consumer price index (CPI) may overstate the rate of inflation for the consumer sector of the U.S. economy has drawn considerable attention from the U.S. Congress, the administration, and users in the academic and business community. The recent report of the "Boskin Commission"(the Advisory Commission to Study the Consumer Price Index, commissioned by the U.S. Senate Finance Committee) has drawn attention to the potential magnitude of this problem for the United States and the fiscal consequences it poses, because the CPI is used to escalate payments in many federal programs and to adjust the marginal tax brackets used for the assessment of income taxes. In an IMF Working Paper entitled Improving the Efficiency of the U.S. CPI, Paul Armknecht discusses the concepts and methods underlying the U.S. CPI and proposes changes to resolve many of the sources of bias that contribute to overstating inflation.

The traditional measures used worldwide to gauge inflation lack some desirable statistical properties and are based on unrealistic economic assumptions, according to Armknecht. The literature on index number theory suggests measures that are superior to the more traditional fixed-quantity weight price index formulas as currently used by the U.S. Bureau of Labor Statistics, producer of the U.S. CPI. Such fixed-weight indexes assume that consumers purchase constant proportions of various goods and services despite the fact that, as relative prices change, consumers substitute less expensive items for more expensive ones. Armknecht presents two improved measures that more closely approximate consumer behavior and suggests how one of these can be implemented without great difficulty using existing Bureau of Labor Statistics procedures.

Many countries use scientific sampling techniques to select the geographic areas and the retail shops in their CPI surveys. The United States, however, is the only country that uses probability proportionate to sales sampling techniques to select the individual varieties of goods and services withinPage 77 each shop. This sampling within shop increases the costs of data collection substantially. Armknecht observes that significant efficiencies can be achieved by sharing statistical information from the detailed records of the U.S. censuses of business enterprises and the census of population and housing. Such data sharing between the Bureau of Labor Statistics and the Bureau of...