Economic Trends
Concurrent Changes In The CPI As A Measure Of Inflation
Research Report
Satyendra Verma, AARP Public Policy Institute
December 2000
The Consumer Price Index (CPI), which is administered and constructed by the Bureau of Labor Statistics (BLS), is the main source of information on consumer prices and inflation in the United States. The CPI influences both public and private sectors' policy decisions as well as the finances of the federal, state, and local governments. The measure of inflation has a multi-dimensional effect on the economy. Both budget outlays and revenues are very sensitive to inflation. In the fiscal year 1998, a half trillion dollars of federal spending was linked to price changes measured by the CPI. Besides federal programs, the CPI indexing also affects private contracts, the allocation of funds, and taxes.
In order to maintain the accuracy of the CPI, two kinds of revisions are continually applied to it. One, the "market basket" of goods and services upon which the expenditure weights are calculated, is updated regularly to adjust for the new items in the consumption pattern. Two, methodological changes are introduced that range from the reselection and reclassification of items and outlets to the formulation of the index numbers. In 1998, comprehensive revisions of the CPI included the use of a geometric mean index instead of a Laspeyres index for 61 percent of the items in the market basket, and a hedonic regression approach to make adjustments for the quality of products. The comprehensive revisions of 1998 were planned to be completed by 2000. 1
The motivation behind the methodological changes was to have a more accurate measure of inflation. Following the recommendations of the Advisory Commission To Study The Consumer Price Index (the Boskin Commission, 1996) appointed by the Senate Finance Committee, and the proposed changes the BLS made on its own, measured inflation, which was thought by many to be overestimated, was to be scaled down by 0.46 to 0.56 percentage points. The new formula alone (geometric mean index), effective from January 1999, was estimated to reduce inflation by 0.20 to 0.26 percentage points per year.
While new improvements have been applied to the CPI from time to time, the CPI historical series was never revised unless data processing errors were discovered. The routine methodological changes were never applied to the historical series of index numbers. In 1999, however, for research purposes, the BLS prepared a research series of the CPI (CPI-RS) from 1978 forward to show how the measure of inflation would have differed if all concurrent changes were applied retroactively to the CPI-U. The research series was intended to serve a dual purpose: One, it would show how inflation was overestimated in the past; and two, it would prove how these improvements, if applied, would reduce measured inflation in the future. The CPI-RS increased 141.2 percent from December 1977 to December 1998 compared with 163.9 percent for the CPI-U. This difference implied an average annual increase of 4.28 percent instead of the 4.73 percent reported by the CPI-U. This was 0.45 percentage points lower — about the same magnitude as the Bureau had estimated in their 1998 CPI revisions.
The research series, CPI-RS, applied three kinds of improvements:
(a) Rental equivalence as a flow-of-service to measure homeowners cost. The CPI-RS also made improvements in the shelter component of "housing" and reported residential and owners' equivalent rent in a separate index, CPI-U-XL. The rent equivalence had an upward bias of 0.1 percent in the measure of inflation.
(b) Formula changes — use of geometric means to correct the consumer substitution bias. The Laspeyres index, unlike the geometric mean index, assumes that consumers use fixed quantities of given products, and do not substitute one product for another. In 1997, the BLS published a new index based on the geometric mean index, CPI-U-XG, for 1991- 98, applied to 61% of CPI-U items.
(c) Quality changes and other changes — use of hedonic regression, a statistical estimation technique, to make quality adjustments in electronics such as camcorders, TVs, personal computers, VCRs, as well as autos (both new and used), housing, and apparel. Also, the replacement of brand name by generic drugs (when patents on brand names expire) is treated as a price decrease, and mandated pollution control measures as a price increase. For pricing hospital services, the new methodology obtains sample prices of specified treatments for particular diseases rather than for number of days in the hospital.
Using formula changes alone reduced the annual rate of inflation by 0.28 percentage points in 1978-82, 0.26 percentage points in 1983-86, and 0.41 percentage points in 1987-97. The quality improvements, on the other hand, increased the price index. The use of rental equivalence reduced inflation only once when applied in 1978; subsequent years had negligible effect (see Fig. 1).
The research series serves as a valuable source to envision the impact of combined methodological changes on measured inflation historically. Only by retracing the impact backwards, can we visualize the impact of these changes in the future calculations. Figure 2 shows the difference between the research series and the actual CPI-U between 1978 and 1998-the difference by which measured inflation was overestimated. For example, inflation was overestimated by 1.8 percentage points in 1980, 0.6 percentage points in 1991, 0.4 percentage points in 1993, and 0.2 percentage points in 1995-98.
All the changes combined caused larger corrections in the beginning of the series than towards the end, because inflation was higher in the early1980s. Secondly, corrections show larger impact when first applied. Once corrections are in place, their impact dampens over time. The CPI revisions of 1998 would continue reducing the measure of inflation by 0.2 percentage points per year in 2000 and beyond. Besides the proposed changes, the BLS has also decided to revise the expenditure weights every two years. The next revision, in 2001-02, will replace the current 1999-2000 weights. The BLS has also decided to publish a "superlative index" in 2002 along with the CPI-U and CPI-W, which would reduce consumer substitution bias even further. The superlative index, according to the Bureau, will be more suitable for cost-of-living index (COLI) calculations--the ratio of minimal cost needed to achieve a given standard of living in two time periods.
Conclusion
The research series CPI-RS does have some limitations. Although a close proxy of all the changes between 1978-1998, it does not take into account those improvements which have negligible effects, such as treatment of seasonal items, discount airfares, hospital services, and sample rotations, etc. In the future, most of the changes in the CPI-U will be driven more by formula changes and less by items and sample rotations.
Top of PageFootnote
1 The Laspeyres index is a weighted arithmetic mean of prices with base year quantities as weights. The geometric index is based on the geometric mean of prices with shares of expenditure as weights. The latter formulation is more suitable when consumers substitute one product for another or introduce new items in their budget.
Written by Satyendra Verma, Ph.D.
December 2000
©2000 AARP
May be copied only for noncommercial purposes and with
attribution; permission required for all other purposes.
Public Policy Institute, Public Affairs, AARP, 601 E Street, NW,
Washington, DC 20049
Pub ID: DD51