Skip Navigation

The Economic and Budget Outlook: An Update September 1997

The Bureau of Labor Statistics (BLS), the agency that compiles the consumer price indexes (CPIs), has instituted a number of changes in recent years in the way that it calculates price indexes. Moreover, it has announced additional changes for 1998 and 1999. Although the methods used to construct the CPIs have been modified many times over the long history of the series, the changes in the 1995-1999 period are particularly important. The BLS estimates that, by early 1997, the changes made from 1995 through early 1997 probably reduced the measured increase in inflation by 0.2 to 0.3 percentage points a year compared with what the former methods would have generated. The Congressional Budget Office (CBO) estimates that the changes slated for the next two years will reduce CPI growth by approximately 0.4 percentage points more by the end of 1999.

The changes that have been embodied in the CPI since 1994 modified the measurement of price changes for a wide number of categories. In January 1995, the modifications corrected an upward bias in the owners' equivalent rent measure and a downward bias in the measure of nonowner rent, made a better link between prices of generic drugs and prices of corresponding drugs that have gone off patent, and changed the method of introducing new food items into the sample (in a process called "sample rotation") to eliminate an upward bias. In mid-1996, the method of introducing new food items was extended to the sample rotation of all other nonshelter categories of items in the CPIs. Another change, which removed a small upward bias, was made to the way in which new items are brought into the sample when the existing item can no longer be priced. Then, in January 1997, the measure of hospital prices was changed to try to better reflect the actual services that are provided and the prices of transactions. That change also probably reduced the measured growth of prices.

The changes to be introduced in 1998 and 1999 will further reduce CPI growth.⁽¹⁾ During 1998, the BLS will begin to use new weights for calculating the aggregate price indexes. The current CPIs are calculated using weights based on patterns of consumer expenditures in the 1982-1984 period. Starting in 1998, the weights will reflect patterns of expenditures from 1993 through 1995. That rebenchmarking has been done about every 10 years in recent decades--the last rebenchmarking was done in 1987. Shifting to more current weights tends to reduce the growth of the aggregate price index because current weights put more emphasis on items whose prices have been growing less rapidly in recent years. The CBO economic forecast detailed in Chapter 1 assumes that the 1998 rebenchmarking will initially reduce CPI growth by 0.15 percentage points.

In January 1998, the BLS will also adopt a new procedure for measuring prices for personal computers and peripheral equipment that will reduce the growth of the CPI. The procedure, called a hedonic model, has been used in the producer price index (PPI) since 1991, and it provides a way to estimate the value of changes in quality.

Because of rapid improvements in the quality of computers and peripheral equipment, the PPI for electronic computers has fallen dramatically, particularly last year when it plunged about 30 percent. The CPI for information-processing equipment fell only about 10 percent last year, but the current CPI does not use the hedonic model for computers (although hedonic models are used for some other CPI categories). The BLS has not estimated the effect of using the hedonic model for computer pricing on the growth of the CPI. However, the combined effect of a new, larger weight-ing for computers (because of the rebenchmarking) and the new method of measuring prices is likely to be significant. Whereas the old category for information-processing equipment depressed overall CPI growth last year by about 0.01 percentage point, the new category for personal computers and peripheral equipment--if it had been in place--would have depressed overall CPI growth by about 0.06 percentage points. Of course, if computer prices do not continue to fall as fast as they have in recent years, the effect will be smaller in the future.

In 1999, two changes in CPI methodology will be instituted. The weighting procedure for compiling many of the subaggregates of price change will switch from a simple arithmetic weighting to a geometric weighting. Under certain assumptions about how consumers change their consumption patterns in response to changes in relative prices, the geometric weighting results in a more accurate approximation of a cost-of- living index. The BLS has estimated that the switch will probably reduce CPI growth by up to 0.25 percentage points a year depending on how many subag-gregates the new weighting scheme is used for. CBO's economic forecast assumes the change will reduce CPI growth by 0.14 percentage points a year.

The second change in 1999 will affect the rotation of the sample--in other words, the way in which new stores and items are periodically brought into the sample of items selected for price quotes. Like rebench-marking, rotation updates the sample to reflect current patterns of expenditures. But it affects the outlets and items selected rather than the weighting, and it is an ongoing process that can affect the CPI every year rather than just every 10 years.

Sample rotation now occurs in one-fifth of the geographic areas every year, and a significant lag exists between the time of the survey used to select the outlets and the actual introduction of those outlets in the sample. Therefore, new goods are not fully captured in the CPI nationwide until long after they have been introduced. Starting in 1999, sample rotation will be done according to the category of expenditure rather than geographic area. In addition, the time between the survey and the actual introduction of the outlets will be sharply reduced, perhaps even cut in half. Rotating the sample by category of expenditure will also permit the BLS to focus its rotation on those categories that are undergoing more rapid change, and thereby permit the CPI to reflect consumer spending on future new goods, such as high-definition televisions, in a more timely way.

The Bureau of Labor Statistics has not estimated the effect of that rotation change on the growth of the CPI, but the change will probably reduce CPI growth. The prices of new goods often decline during the first five or 10 years that the goods are on the market as the growing market permits economies of scale and as competitors increase. Therefore, the earlier that the CPI captures new goods, the more likely that CPI growth will be dampened. The effect of that change in methodology on the CPI depends on how much faster new goods are captured, the actual pricing patterns of new goods in the future, and how many important new goods are introduced. The change in methodology could reduce CPI growth by as much as 0.1 percentage point a year, but estimates of the effect are extremely uncertain.

Clearly, the combined effect of the BLS's changes on the CPI over the 1995-2000 period will be substantial. Taken together, the changes may reduce measured inflation on the order of 0.7 percentage points compared with what would have occurred if no changes in methodology had been made. Therefore, CBO's forecast for CPI inflation in 2000, which is 3 percent, cannot be directly compared with CPI inflation measured using pre-1995 procedures. If CBO had forecast the same CPI concept that was in use in 1994, the forecast for CPI inflation in 2000 would have been higher--pre-sumably by about 0.7 percentage points. The changes that the BLS is making improve the accuracy of the CPIs, but they also complicate the assessment of the change in inflation over recent years and over the next few years.

1. The 1998 and 1999 changes are discussed in detail in Department of Labor, Bureau of Labor Statistics, Monthly Labor Review (December 1996).

---