The revised Consumer Price Index: changes in definitions and availability.
Although the CPI is a measure of price change for a market basket of constant quality and quantity, it also needs to retain its relevance to consumers' experience by pricing items currently purchased. New consumer purchasing patterns occur as a result of changes in a number of factors, such as relative prices, income, tastes, demographic characteristics, technological changes, and population shifts. Thus, periodic revisions of the CPI are necessary to incorporate updated versions of the market basket.
This article is one of a series that provides detailed information about the CPI revision. It highlights the changes that will occur in the availability and in the definitions of indexes. Many of the changes derive from shifts that have occurred in the spending patterns of the American public. Nearly five decades of spending patterns as reflected in the CPI expenditure weighting patterns and the corresponding relative importance of major groups are shown in table 1. A later article will discuss the new expenditure weights in detail.
One clear trend in consumer spending has been a reduction in the relative importance of expenditures for food, especially grocery food. Although it is not immediately obvious from the data in table 1, there have been corresponding increases in importance for new products and services such as video recorders and day care. To ensure the most accurate CPI possible, it has been necessary to allocate more pricing and calculation resources to these new and growing expenditure categories, with the correlative result that proportionately fewer resources will be available for items of declining importance. This means new indexes for previously unpriced products will become available. But it also means that there will be some reduction in product detail for expenditure categories with declining relative importance.
The discontinuation of an index does not mean that the item is no longer priced for the CPI. All of the previously priced items will continue to be priced, but with much smaller samples. The relative proportions that these items constitute of the new combined strata (class of similar items) to which they are assigned will also be subject to annual updating through the sample rotation process.
Exhibit 1 summarizes the definitions for new item indexes and explains the definitional changes that some other indexes are undergoing. These definitional changes arise from the need to combine some previously separate items, the addition of some previously unpriced items, and conceptual or coverage changes which enhance the measurement or interpretation of the index.
Some of the items that are being discontinued as separate item strata because of their reduced relative importance have significant applications independent of their use in the CPI. To accommodate users of these indexes, BLS will continue to publish a limited number of them as special sub-strata CPI-U indexes. These sub-strata indexes will be based on extremely small samples and will be less reliable than the pre-1987 numbers. Footnote 3 lists the item strata that are being discontinuted and indicates whether a corresponding sub-strata index will be available.
Beginning with the CPI-U for January 1983, BLS adopted an improved method--called rental equivalance--for estimating homeowner shelter costs. (The change was made in January 1985 for the CPI-W.) The 1987 CPI revision continues the definitional and coverage features associated with that change. In addition, it incorporates two new refinements consistent with that change. First, the new index for materials, supplies, and equipment for home repairs, which combines three more detailed old indexes, will include for pricing only those types of items that would be purchased by tenants and exclude those typically purchased by landlords for major repairs or capital improvements. Second, the rental value of owner-used vacation property is included in lodging while out of town.
The definitional treatment of premium costs for health care insurance will have a change which will affect the structure of the expediture weights for health insurance, but not the methodology for estimating price changes affecting the costs of health insurance. Beginning with January 1987, the CPI will define the cost of health insurance as the portion of premium payments which is retained by the insurer in the form of profits and operating expenses. The portion of the premium which is either paid directly by the insurer to health care providers or as reimbursements to policyholders will no longer be defined as a helath insurance expenditure, but rather as a direct medical care expenditure.
This definitional change will slightly modify the method used for estimating changes in health insurance premiums. An indirect method has been used to estimate the changes in costs of health insurance. The expenditure for health insurance has been defined as total consumer premium payments. The price change for these premiums has been estimated with a combination of the changes in cost for covered medical services and the changes in premium retained by insurers for operating costs and profit.
The revised definition will result in the portion of the health insurance expenditure that is paid as benefits by the insurer being included in the directly priced medical care strata, for example, physician services, hospital room, and eye care. The result of this treatment is that the expenditure weights of these directly priced medical care strata will be increased, and they will also recieve a greater proportion of the price quotations in the CPI. The expenditure weight for health insurance will represent only the portion of the premium retained by the insurers, and changes in its costs will continue to be estimated using an indirect pricing procedure that relies both on CPI changes for covered medical expenses and secondary data on premiums retained by insurers. This definitional change will result in the discontinuation of the index for "Other medical care services."
Local area indexes
As announced in 1984, BLS has allocated the price quotation samples among the 91 pricing areas in a sample design which will produce the most accurate national CPI possible with the funds authorized. The decision to improve the national CPI estimate will reduce the frequency of publication for some areas. Beginning in 1987, semiannual average indexes will replace bimonthly indexes for 12 currently published areas. (See table 2.) These semiannual average indexes, which are the averages of the 6-month periods from January through June and from July through December, will be published with the release of the July CPI in August and the January CPI in February.
The method of calculating the averages for a semiannual average index derives from the one currently used for calculating annual average indexes which BLS publishes at the end of each year. Because montly and bimontly indexes are not published in those areas, the first step will be intermediate monthly and bimonthly calculations for use in the average computation. For those items which are priced monthly, such as food at home, an intermediate montly calculation will be prepared for each of the 6 months. These six calculated numbers will be summed and then divided by six to obtain the semiannual index. A similar but more complex technique is used for items priced bimonthly in each area. An intermediate calculation will be compiled for each of the 3 months that items are actually priced during the 6-month period. The monthly calculation for each of the other 3 months will be interpolated by calculating a geometric mean of the months adjacent to the one being estimated. For example, in an area priced in even-numbered months, a January interpolation would be estimated by taking the geometric mean between the indexes calculations for December and February. Interpolations would be made in a similar manner for March and May. The three intermediate numbers for February, April, and June, calculated with collected prices, would be summed with the three interpolations and divided by six to obtain the semiannual average index for the first 6 months of the calendar year.
The calculation of semiannual indexes for areas in which bimonthly items are priced only in odd-numbered months would use the same methodology except that the data for February, April, and June would be interpolated by using the geometric mean between the calculations for their adjacent months. For example, the June interpolation would be estimated from the calculations made for May and July.
Although BLS will calculate semiannual indexes for these 12 areas from intermediate estimates of monthly data, the samples are much too small to produce a reliable bimonthly or monthly estimate of price change. Estimates based on a small number of observations in a single month would be subject to extremely high volatility resulting primarily from sampling error. The average of 6 months of price data increases the reliability of the estimate. The new semiannual index estimates of price change will be as accurate as the old bimonthly index estimates which they replaced for the 12 local areas identified in table 2.
The BLS has systematically advised users that local area CPI's (including the new semiannual averages) are subject to substantially more statistical error and variablity than the national index. As a result, local indexes should not be used in escalation provisions. Some individuals may have already adopted escaltion clauses using one of the local indexes that will change from bimonthly to semiannual publication. After the data for December 1986, it will not be possible to use individual monthly estimates for these 12 areas, and parties to agreements using these indexes may want to agree on some alternative. BLS does not provide contract interpretation assistance but can provide limited technical assistance for transition, if requested by both parties to an agreement.
The use of the new Consolidated Metropolitan Statistical Area (CMSA) definitions, issued by the Office of Management and Budget, resulted in a number of the published areas becoming larger in terms of their sampled geography. Of the 27 local areas to be published, 5 (Anchorage, Buffalo-Niagara Falls, Honolulu, Milwaukee, and San Diego) did not have changes to the geographic coverage currently being priced by BLS. Only Dallas-Fort Worth became smaller, with Wise County being removed from the official definition.
Several areas have had significant expansions of their sampled geography. For example, the New York area now includes Danbury and other parts of Connecticut; the Philadelphia area has added Wilmington and Trenton; Boston now includes some parts of New Hampshire; the Chicago area has three additional counties including Kenosha, WI; Houston has added Galveston; Los Angeles includes Riverside-San Bernardino; and San Francisco includes San Jose. Table 2 contains a complete list of counties for each local area with a published CPI.
Table 2 also shows the population for both the CPI-U and CPI-W in each of the publication areas as a percentage of their respective total U.S. 1980 urban population. If these weights are compared with the weights shown for 1970, one can ascertain the degree of relative population change in each area since 1970. For example, the weight for the CPI-U population in the Northeast region declined from 26.521 in 1970 to 23.997 in 1980. This decline reflects the faster growth rate of the population of the South and West in recent years, compared with the Northeast. Even though the New York area has expanded since 1970, its relatve population weight has declined.
The population weight for the San Francisco-Oakland-San Jose area has become larger than that for the Detroit-Ann Arbor area. Based on population growth since 1970 and the expansion of its definition, the San Francisco area has superseded the Detroit area as the fifth largest area covered by the CPI indexes. For that reason, the San Francisco area, beginning with the data for January 1987, will be published monthly while the Detroit area will be published bimonthly (even-numbered months). The publication of indexes for Cleveland are changeing from even-numbered months to odd-numbered months, D-size (that is, urban areas with populations under 75,000) strata indexes will not be published in the Northeast and West, and indexes will no longer be published for the Scranton-Northeast Pennsylvania area.
Both the CPI-U and the CPI-W for January 1987 will be linked to the present series of each index as of December 1986 to provide a continuous series. For most indexes, the linking will be accomplished by setting the index levels of the revised CPI with the updated expenditure weights and samples equal to those published for the present series in December 1986. Each index will move upward or downward from the December 1986 level in accordance with subsequent changes in prices. The local area indexes which are calculated and published for the odd-numbered months will be linked to their present series in November 1986 and subsequent changes in prices measured from that point in time. For new items and for those items that have undergone significant definition changes, indexes will be introduced with November or December 1986=100.
As in the past, BLS will publish selected indexes using the old expenditure weights for 6 months after the issuance of the revised CPI. Unlike earlier revisions, these overlap indexes will be calculated from the updated item, outlet, and area samples and will differ from the revised indexes only by their expenditure weights. As a result of a number of enhancements made in the CPI during this and the previous revision, the costly activities of replacing the entire set of item, outlet, and area samples prior to the issuance of the revised CPI have been eliminated. Substantial cost reductions in the revision process were achieved by replacing only those item, outlet, and area samples which were necessary for estimating a CPI based on the 1980 population and the 1982-84 market basket of expenditures. In earlier revisions, the 6-month overlap old series indexes used not only the old expenditure weights but also the old item, outlet, and area samples. The base period for the revised CPI for January 1987 will be 1967=100, the same as for the present index.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||changes in methodology, items included|
|Author:||Marcoot, John L.; Bahr, Richard C.|
|Publication:||Monthly Labor Review|
|Date:||Jul 1, 1986|
|Previous Article:||An analysis of regional employment growth, 1973-85.|
|Next Article:||Reconciling divergent trends in real income.|