oversight

Consumer Price Index: Impact of Commodity Analysts' Decisionmaking Needs to Be Assessed

Published by the Government Accountability Office on 1999-06-15.

Below is a raw (and likely hideous) rendition of the original report. (PDF)

                United States General Accounting Office

GAO             Congressional Requesters




June 1999
                CONSUMER PRICE
                INDEX
                Impact of Commodity
                Analysts’
                Decisionmaking Needs
                to Be Assessed




GAO/GGD-99-84
      United States
GAO   General Accounting Office
      Washington, D.C. 20548

      General Government Division

      B-279259

      June 15, 1999

      The Honorable John L. Mica
      Chairman, Subcommittee on Criminal Justice,
        Drug Policy, and Human Resources
      Committee on Government Reform
      House of Representatives

      The Honorable Christopher Shays
      House of Representatives

      One of the most important economic indexes produced by the federal
      government is the Consumer Price Index (CPI). According to the Bureau of
      Labor Statistics (BLS), which publishes the index, the CPI is the principal
      measure of trends in consumer prices and inflation in the United States.
      The CPI is used by the federal government, businesses, and others. In fiscal
      year 1998, $499 billion in federal spending, such as income payments to
      Social Security beneficiaries, was automatically linked to price changes
      measured by the CPI. In addition, because it is used annually to adjust
      various aspects of federal individual income tax for inflation, such as the
      tax brackets and the amounts of personal exemptions, every individual
      income taxpayer is affected by changes in the CPI.

      The CPI tracks the prices of a fixed market basket of goods and services
      that consumers purchase. There are thousands of different products and
      services in the market basket, and BLS attempts to obtain prices on the
      exact same products and services each month. By tracking the exact same
      items each month, BLS seeks to avoid capturing price differences that are
      due to changes in the characteristics of a product or service rather than
      simply changes in price. However, BLS cannot always find the exact same
      item each month; and when this happens, BLS price takers in the field
      “substitute” a new version of the product for the old version. Substitutions
      occur for a variety of reasons, including changes in technology or style as
      well as when an item is out of stock at a store in which prices are being
      collected. In calendar year 1997, according to BLS data, substitutions
      ranged from 1.4 percent in the food and beverages component to
      12.8 percent in the apparel and upkeep component.

      After substitutions are made in the field, BLS’ commodity analysts in
      Washington, D.C., decide if there are significant differences in
      characteristics between the items and their substitutes. When commodity
      analysts determine that the differences are significant, they make what BLS
      refers to as quality adjustments to separate pure price changes from price




      Page 1                                      GAO/GGD-99-84 Consumer Price Index
                   B-279259




                   changes that are due to other factors, such as differences in quality, size,
                   or quantity.1

                   The adjustments made by commodity analysts affect the price changes
                   that go into computing the CPI. Evidence indicates that substitutions (those
                   that are adjusted together with those that are not adjusted) have a
                   significant impact on the CPI. A BLS study estimated that, while less than
                   4 percent of the price quotations used to calculate the CPI in 1995 were
                   substitutions, these substitutions were responsible for about one-half of
                   the price increase in the CPI for the items studied.2 To the extent that more
                   than pure price changes are included in the CPI, the index’s accuracy is
                   affected.

                   As you requested, this report describes (1) how commodity analysts
                   decide whether to make adjustments, (2) the adjustment methods they
                   use, and (3) how supervisors of commodity analysts review the analysts’
                   decisions. For this report, we gathered descriptions from commodity
                   analysts on how they made decisions for specific substitutions that ranged
                   across the major components of the CPI.


                   Commodity analysts use a combination of professional judgment, general
Results in Brief   procedures, specific methods, and limited written guidance in deciding
                   whether and how to make adjustments for substitutions. The relative
                   importance of these four elements varies in the analysts’ decisionmaking,
                   depending upon the specific substitutions. In some cases, judgment is of
                   primary importance. According to BLS, the commodity analysts’
                   supervisors usually review substitution decisions only when they consider
                   the price increases or decreases that result from the analysts’ decisions to
                   be large. Beyond the specific reviews performed by supervisors, BLS does
                   not have a program of assessing the decisionmaking patterns of
                   commodity analysts.

                   In making adjustments, BLS’ objective is to include only pure price change
                   in the calculation of the CPI and, to the extent possible, eliminate price
                   change that is the result of other factors, such as improvements in quality,


                   1
                    In this report, we refer to all adjustments—whether for quality or other reasons—simply as
                   adjustments.
                   2
                    This BLS study included data from about 83 percent of the 1995 item strata. (BLS groups items in the
                   CPI together at broad levels of similarity—food and beverages—and then at sublevels of similarity.
                   Item strata—ground beef, chuck roast—are sublevels.) The study attributed the disproportionate
                   impact of substitutions on the CPI to manufacturers’ tendency to increase prices when new products
                   were introduced, and retailers’ tendency to discount prices when old products were discontinued.



                   Page 2                                                     GAO/GGD-99-84 Consumer Price Index
B-279259




size, or quantity. Commodity analysts receive and review information
about the old and new versions of the commodity from which they make a
series of determinations that revolve around whether the two versions are
similar and, if not, which adjustment method to apply.

Commodity analysts compare the characteristics of the original item with
the replacement and then use their professional judgment to decide if the
two are comparable. For example, a commodity analyst who reviewed a
substitution of an electric blanket with a 5-year warranty for an electric
blanket with a 2-year warranty judged them to be comparable. In calendar
year 1997, analysts concluded that a majority—58 percent—of about
29,000 substitutions were comparable. For other than certain food items,
no current written guidance was available for commodity analysts to
follow in making their decisions. In some instances, according to BLS, the
decisions are straightforward and involve little judgment. In other
instances, the analysts must exercise a significant degree of judgment to
make decisions. For items that are judged to be comparable, no
adjustment is made; and the difference between the prices of the new and
old versions, expressed as a percentage, is used in the calculation of the
CPI for that month.


When a substitution is not comparable with the item it replaces,
commodity analysts either use a direct adjustment method to make an
adjustment themselves or assign one of two indirect methods, in which
case BLS computer programs make the adjustments. In 1997, about
one-third of all nonrent adjustments were made directly; about two-thirds
were made indirectly. Direct adjustments are made when commodity
analysts have data on the ways the old and new versions differ and have
information with which to assess the value of those differences. They are
made when the specific cost of a quality change can be estimated either by
the manufacturer of the items or by using BLS’ statistical models that
incorporate price data. Direct adjustments are also made when an item’s
size or quantity changes. Most direct adjustments in calendar year 1997
were for apparel items and new and used vehicles. Although BLS’ process
for reviewing changes that affect residential rent is somewhat outside the
process that it follows for other CPI items, the majority of adjustments
made to residential rent are direct adjustments.

When an adjustment is judged to be warranted, but a direct adjustment
cannot be made, commodity analysts apply one of two indirect adjustment
methods that impute a rate of price change. Both methods impute the pure
price change by averaging the rates of price changes experienced by the



Page 3                                     GAO/GGD-99-84 Consumer Price Index
B-279259




same type of items in the CPI. The two methods, “class mean” and linking,
differ in terms of the subsets of items included in calculating the rate of
price change. The class-mean method is generally used for products where
new models or product lines are introduced fairly regularly. It is based on
the rate of price changes experienced by other substitutions—of the same
type of product or service—in the particular geographic location. These
other substitutions are those that the commodity analyst had judged for
that month to be comparable or had directly adjusted. Thus, the
class-mean method relies exclusively on an analyst’s judgments for related
substitutions. The linking method is not limited to price changes resulting
from substitutions. This method includes all items of the same type and in
the same location as the item in question, and it is most heavily influenced
by items that had not changed; that is, those that were not substitutions.
The linking method includes a larger array of products and services than
the class-mean method.

According to BLS, there are no guidelines or policies in writing for
supervisors to follow in selecting and reviewing the substitution decisions
of commodity analysts. In practice, according to BLS, there is an unwritten
policy that supervisors are to review substitution decisions when they
consider the price increase or decrease to be large. Few other adjustments
are reviewed. BLS has no policy to randomly or otherwise select and review
substitution decisions.

Beyond the specific reviews performed by supervisors, BLS does not have a
program to assess the decisionmaking patterns of commodity analysts.
However, studies have been conducted from time to time, and three were
conducted in the 1980s and early 1990s. The studies found the
decisionmaking process to be susceptible to producing errors and
inconsistencies and recommended actions intended to promote greater
controls over the decisionmaking process. According to officials we
interviewed, however, BLS now takes the position that such controls are
not required for experienced commodity analysts.

We found no evidence to indicate whether errors or inconsistencies in
commodity analysts’ decisions or lack of comprehensive reviews of those
decisions has had a material effect on the calculation of the CPI. However,
it is sound management practice for BLS to assess, on a periodic basis,
whether errors and inconsistencies in commodity analysts decisions
materially affect the CPI and we make a recommendation to that effect.




Page 4                                      GAO/GGD-99-84 Consumer Price Index
             B-279259




             Every month, usually by the middle of the month, BLS, which is a part of
Background   the U.S. Department of Labor, publishes a new CPI based on data collected
             in the previous month. Two CPIs are published, and the data for each are
             arrayed in various ways, such as by nationwide average for urban areas
             and by selected local areas.3 To produce these indexes, BLS collects and
             processes large amounts of data.

             According to BLS data, an average of about 77,000 price quotations—price
             and characteristics about a product or service—were collected each
             month in calendar year 1997. To obtain these quotations, approximately
             30,000 retail and service establishments and nearly 4,000 landlords and
             tenants were visited or contacted every month. All together, BLS tracked
             the prices of about 94,000 specific items in 1997, although every item was
             not priced every month. The many kinds of products and services under
             which these thousands of items were categorized ranged from white bread
             to funeral expenses (see app. VIII).

             Although some pricing information is gathered by BLS headquarters
             personnel, most price quotations are collected by BLS field representatives
             who are also referred to as price takers. According to BLS, each price taker
             is assigned specific outlets (e.g., supermarkets, department stores, car
             dealers, housing units, and doctors offices) to visit and a list of goods and
             services within those outlets to price.

             If an outlet does not have the exact item, BLS requires the price taker to
             select a substitute item in that outlet. Depending on the item, the price
             taker may visit or contact the outlet more than once to find the missing
             item before making a substitution. In calendar year 1997, acceptable
             substitutions were made for about 3.3 percent of the 872,829 nonrent price
             quotations collected.4 Price takers are to select substitutions that are as
             similar as possible to the items that were not found.

             Substitutions are reviewed by commodity analysts who work in the
             Consumer Prices Branch of the Division of Consumer Prices at BLS’
             headquarters. The branch, which is headed by a branch chief, is divided


             3
               The two CPIs are the CPI for All Urban Consumers (CPI-U) and the CPI for Urban Wage Earners and
             Clerical Workers (CPI-W). According to BLS, the CPI-U represents about 87 percent of the U.S.
             population, and the CPI-W represents about 32 percent of the U.S. population. BLS began publishing the
             CPI-U in 1978. Until then, it published only the CPI-W.

             4
              Upon review, BLS may classify a substitution as an unacceptable replacement if, for example, a
             substitution occurred outside of the time frame BLS has designated for a seasonal item—a spring or
             summer raincoat for a fall or winter coat. Unacceptable substitutions are discarded, and BLS does not
             include them in its statistics on substitutions.



             Page 5                                                     GAO/GGD-99-84 Consumer Price Index
              B-279259




              into five sections, each of which is headed by a section chief. There were
              three supervisors in April 1998 who, in addition to the section chiefs,
              supervised 29 commodity analysts.

              Appendix II provides more information about the general construction of
              the CPI and the collection of prices.


              For our first and second objectives—to describe how commodity analysts
Scope and     decide whether to make adjustments and the adjustment methods they
Methodology   use—we (1) obtained relevant documents and data from BLS; (2) had
              commodity analysts walk us through selected substitutions that they had
              reviewed, asking them to explain what they did and why; and
              (3) discussed the methods used to make adjustments with the supervisors
              of commodity analysts.

              Among the documents we reviewed were BLS manuals and handbooks
              pertaining to the CPI, such as an instruction manual for price takers,5 a
              handbook that included descriptions of the adjustment methods,6 and a
              handbook that described procedures for reviewing housing rental data.7 In
              addition, we reviewed BLS’ descriptions of the adjustment procedures that
              were published either as internal research papers or in professional
              journals. With the exception of certain food items, BLS did not have a
              current set of written procedures that commodity analysts followed when
              reviewing nonrent substitutions.

              At our request, BLS provided us with 1997 summary data on the number of
              price quotations collected, the number of substitutions, and the number of
              times the different methods of adjustment were used. These data are
              reproduced in appendix VIII.

              To obtain an understanding of the commodity analysts’ decision
              processes, we talked with analysts about specific substitutions. We
              interviewed a judgmental selection of 19 out of 28 commodity analysts




              5
                Consumer Price Index: C&S Pricing Data Collection Manual, BLS (Washington, D.C.: U.S. Department
              of Labor, n.d.).
              6
               BLS Handbook of Methods, BLS (Washington, D.C.: U.S. Department of Labor, April 1997).
              7
                Housing Commodity Analyst Handbook, BLS (Washington, D.C.: U.S. Department of Labor,
              January 1996).



              Page 6                                                   GAO/GGD-99-84 Consumer Price Index
B-279259




about their decisions on 120 specific substitutions.8 (See appendix I for the
details of how we selected the analysts and substitutions.) In addition, we
talked to a supervisor about another 16 specific substitutions because the
analyst who had reviewed these substitutions no longer worked at BLS. We
asked the analysts and supervisor questions, such as what characteristics
changed between the substitution and the item it replaced, what led to the
decision that a substitution was comparable or not comparable, and why a
particular adjustment method was applied.

After these interviews, we judgmentally selected 13 substitutions to serve
as illustrative examples for this report. These examples cover
(1) substitutions from the six major components of the CPI, (2) comparable
substitutions, and (3) the major adjustment methods. See appendix I for
the details of how we selected the examples.

For our third objective—to describe how supervisors review commodity
analysts’ decisions—we interviewed supervisors and reviewed BLS studies
that considered issues relating to supervision. We also discussed the
methods used to make adjustments with the supervisors. The supervisors
we interviewed were the chiefs of the five sections into which the
commodity analysts are divided and three supervisors who were not
section chiefs. In addition, we also interviewed the Branch Chief for
Consumer Prices, who is responsible for all five sections. The studies we
reviewed were an assessment of analysts’ decisionmaking over time,9 a
quality assurance report on analysts’ decisions,10 and an evaluation of a
project to develop decisionmaking computer software that would assist
analysts and supervisors.11

Because the procedures that commodity analysts follow in reviewing
substitutions for most CPI items are unwritten, we relied mostly on our
interviews with commodity analysts and their supervisors and managers to
piece together what those procedures were. To the extent possible, we

8
 One of the 29 commodity analysts had not worked for the CPI in 1997 and was excluded from our
survey. Four of the 19 commodity analysts reviewed substitutions connected with residential housing
rent. The others reviewed substitutions for various goods and services, such as over-the-counter drugs.
Analysts who review residential rent follow different procedures than the others. Appendix VII
provides information on the procedures for residential rent.
9
Jack Galvin, “A Control Chart Analysis of Commodity Analyst Review Activity.” Unpublished study,
BLS, October 1985.
10
 Paul A. Armknecht, “Commodity Analyst Updates During 1985 for C&S Survey Quality Assurance
Report.” Unpublished study, BLS, May 1986.
11
 Bob Adkins et al., “The Development and Testing of the CPI Commodities and Services
Comparability Expert System.” Unpublished study, BLS, February 1993.



Page 7                                                      GAO/GGD-99-84 Consumer Price Index
                     B-279259




                     verified what we were told by cross-checking what one person said with
                     what another person said. We did not verify the computerized data that BLS
                     provided to us, such as statistics on substitutions. Nor did we verify the
                     studies or any other materials BLS provided to us.

                     This report describes procedures which, according to BLS, are intended to
                     contribute to the accuracy of the CPI but does not assess the accuracy of
                     the methods BLS uses to make adjustments or estimate the effects of those
                     adjustments on the CPI. Similarly, our work is not intended to evaluate the
                     overall accuracy of the CPI.

                     We did our audit work in Washington, D.C., from November 1997 through
                     January 1999 in accordance with generally accepted government auditing
                     standards. We requested comments on a draft of this report from the
                     Secretary of Labor or her designee. The Commissioner of BLS provided
                     written comments, which are discussed near the end of this letter and
                     reprinted in appendix IX along with our additional comments.


                     BLS has two processes for reviewing changes in products and services that
Description of How   are in the CPI: one for residential rental units and another for all other
Analysts Decide to   items in the CPI. Although BLS does not substitute one residential unit for
Make Adjustments     another, according to BLS officials, changes in a rental unit may cause BLS
                     to adjust the reported rent. For example, if a landlord added a clothes
and the Adjustment   washer and dryer to the unit since BLS last collected data, BLS would adjust
Methods              the current rent for the value of the addition of the washer and dryer so
                     the rental unit would be comparable to what it was earlier. BLS’ process for
                     reviewing changes to rental units is somewhat outside the process that it
                     follows for other CPI items. Appendix VII discusses the procedures and
                     adjustment methods used for residential rent.

                     Using information gathered from our interviews with commodity analysts
                     and their supervisors and confirmed by BLS officials, we sketched out how
                     commodity analysts review nonrent substitutions and the adjustments
                     they make. Figure 1 illustrates that process. It begins when commodity
                     analysts receive commodity review listings, after which they make a series
                     of determinations, such as whether a substitution is acceptable, whether
                     an acceptable substitution will be adjusted, and which adjustment method
                     will be applied.

                     BLShas not documented the process it uses to review and adjust most
                     nonrent substitutions; but BLS officials have agreed that, although there are



                     Page 8                                       GAO/GGD-99-84 Consumer Price Index
B-279259




probably some exceptions, the figure 1 flowchart reasonably illustrates
that process.




Page 9                                     GAO/GGD-99-84 Consumer Price Index
                                         B-279259




Figure 1: BLS’ Process for Making Nonrent Substitutions and Adjustment Decisions




                                                                          Substitution is
                                                                         “killed,” it is not
                                                                         used in the CPI,
                                                                           andCommodity
                                                                                 a price
                                                                              analyst kills the
                                                                             imputation
                                                                                substitution.
                                                                              is made.a



                                                                            No


                                                 Did the
                Commodity analysts            price taker      Yes             Is this
                                                                           an acceptable                Yes
                receive commodity             identify the
                  listing reviews.             item as a                    substitution?
                                             substitution?




                                                                                                          Is the            No
                                          No
                                                                                                       replacement                A
                                                                                                      comparable to
                                                                                                        the original
                                                                                                           item?

                                             Should the       Yes    Commodity analyst
                                               item be                  creates a
                                              classified                                              Yes
                                                                       substitution.
                                                 as a
                                            substitution?

                                                                                                         Does the
                                                                                                    commodity analyst
                                                                                                    have information to     Yes
                                          No                                                                                      B
                                                                                                   make the replacement
                                                                                                    more similar to the
                                                                                                         original?
                                             Processed
                                             with other
                                          items in the CPI.

                                                                                                        No

                                                                                                  Replacement is used
                                                                                                  in the CPI this month
                                                                                                   with the commodity
                                                                                                      analyst directly
                                                                                                  comparing the prices
                                                                                                  of the original and its
                                                                                                       replacement.




                                         Page 10                                                  GAO/GGD-99-84 Consumer Price Index
                                      B-279259




                                                      Is the
                  Does the                          item one
    No       commodity analyst       No          for which BLS         Yes        Is the class-        Yes       Replacement is not used
          have information on the               has designated                    mean method                     in the CPI this month,
A          value of the difference               that the class-                   appropriate                    and a price imputation
            between the original                 mean method                        to apply?                        is made using the
                item and its
               replacement?                      can be used?                                                      class-mean method.a




              Yes                                No                                 No


            Replacement is used
    Yes     in the CPI this month,         Replacement is not
B             with the commodity           used in the CPI this
               analyst making a            month, and a price
               direct adjustment.          imputation is made
                                             using the linking
                                                 method.a




                                      a
                                       Price imputation is a term used by BLS to indicate that the actual price of the substitution is not
                                      used. Instead, an average is calculated from the price changes experienced that month by the
                                      same type of items in the CPI to handle a missing or unusable price quotation.


                                      Source: BLS.




                                      Page 11                                                      GAO/GGD-99-84 Consumer Price Index
                             B-279259




Commodity Review             Commodity review listings (CRL) provide commodity analysts with
Listings Provide Analysts    information for making various judgments about substitutions and are the
With Information for         primary tool used by analysts in the process of reviewing changes in
                             products and services. CRLs are generated each month from the data that
Decisionmaking               have been collected by price takers. CRLs contain data about a product or
                             service, such as its characteristics (referred to as specifications by BLS)
                             and price history. For example, for a cola soft drink, the CRL may list the
                             packaging (e.g., 12 pack), container construction (e.g., metal can), and
                             caffeine content (e.g., caffeine free). CRLs for items that were reported as
                             substitutions by price takers include the specifications and prices for both
                             the substitution and the item it replaced. All specifications are listed
                             according to a hierarchy of importance for comparing the two versions,
                             and specifications that differ between the two versions are automatically
                             noted.

                             According to BLS, two conditions usually cause a CRL to be generated for
                             substitution review: (1) the price taker reported an item as a substitution
                             and (2) BLS computers, which have been programmed to identify changes
                             in specific characteristics that the price taker recorded for the item,
                             identify a change in a key specification. Computer routines, according to
                             BLS, compare the item reported one month with the item that was reported
                             the previous month in which the item was priced. In these
                             computer-identified cases, price takers would not have reported the
                             current item to be a substitution.

                             Upon receiving these CRLs, one of the first judgments that a commodity
                             analyst is to make is (1) whether an item was correctly reported as a
                             substitution or (2) whether the change in specification should cause the
                             item to be treated as a substitution. The analyst may decide that the price
                             taker inappropriately identified an item as a substitution based on the
                             specifications that the price taker provided, in which case, according to
                             BLS, the substitution is deleted and not used in the CPI. If the
                             computer-identified change in specification is significant in the analyst’s
                             judgment, the analyst can classify the current month’s item as a
                             substitution.


Determining Whether a        After the pool of substitutions has been identified, then the commodity
Substitution Is Acceptable   analysts determine if each of them is acceptable for use in the CPI.
                             According to a BLS official, there are several reasons why a substitution
                             may be unacceptable. For example, a substitution is deemed unacceptable
                             when the commodity analyst waits to see if the old version is only



                             Page 12                                     GAO/GGD-99-84 Consumer Price Index
                           B-279259




                           temporarily unavailable at the retail outlet. Unacceptable substitutions are
                           “killed” in BLS’ terminology—deleted and not used in the CPI—and,
                           according to a BLS official, excluded from statistics on substitutions
                           (including those presented in this report). This BLS official told us that
                           1,065 substitutions were killed between October 1997 and September 1998.

                           In calendar year 1997, commodity analysts determined that 28,881 nonrent
                           substitutions were acceptable and eligible for use in the CPI. These 28,881
                           substitutions represented about 3.3 percent of the 872,829 nonrent prices
                           collected by BLS in 1997.


Determining Whether the    For substitutions that are accepted, the next step commodity analysts take
New and Old Versions Are   is to judge whether the price change resulting from the substitution can be
Comparable                 used in the CPI without adjustment or whether an adjustment is necessary
                           to account for differences between the substitution and the item it
                           replaced. BLS officials reported that the analysts make this determination
                           based on the extent of the differences between the old and new versions
                           of the substituted item and the methods and information available to them.
                           In some instances, according to BLS, the decisions are straightforward and
                           involve little judgment. In other instances, a significant degree of judgment
                           is required.

                           When the new version and the old version are judged similar enough to
                           preclude the need for an adjustment, they are said by BLS to be
                           comparable. In 1997, commodity analysts judged about 58 percent of the
                           28,881 nonrent substitutions to be comparable. For example, in one
                           substitution we reviewed, the level of membership in a tennis club
                           changed from “Tennis plus” to “Gold tennis” but the analyst determined
                           that the memberships were essentially the same and comparable. In
                           another substitution, the manufacturer increased the warranty of an
                           electric blanket from 2 years to 5 years, but the analyst concluded that the
                           blankets were comparable.

                           However, if the new version had characteristics that the commodity
                           analyst considered significantly different from the old version, such as
                           changes in materials, features, or size, the analyst would have to consider
                           whether, and how, to make an adjustment. In such cases, the substitutions
                           and the items they replaced are generally judged by BLS to be not
                           comparable. For example, in one substitution we reviewed, the size of a
                           bed dust ruffle changed from twin (old item) to queen (substituted item),
                           causing the commodity analyst to judge them to be not comparable. In



                           Page 13                                     GAO/GGD-99-84 Consumer Price Index
                         B-279259




                         another case, there were differences in ingredients and size between two
                         packages of soup, and the analyst classified them as not comparable.

                         When the two versions are judged to be comparable, the price of the old
                         version is compared with the price of the new version and the result,
                         expressed as a percentage, is used to calculate the CPI for that month. For
                         example, if two coats were comparable and the new version cost $115 and
                         the old version cost $95, the rate of change for the coat would be
                         21 percent. However, when an adjustment is made, the percentage price
                         change that results from that adjustment is used in calculating the CPI. If
                         the two coats were not comparable because the new one had a lining and
                         the old one did not, and the cost associated with the lining was $5, an
                         adjustment of $5 would be made. The rate of change used in the CPI for
                         that month would be an increase of 15 percent instead of 21 percent.

                         Most commodity analysts that we interviewed did not have written criteria
                         to guide them in making their comparability decisions. But a few food
                         commodity analysts showed us criteria that they had developed with their
                         supervisors for specific types of products to help them decide whether
                         substitutions are comparable.

                         Regardless of whether comparability criteria were available, all of the
                         commodity analysts we interviewed indicated that they examine the
                         specifications on the CRL and decide if the differences in characteristics
                         between a substitution and the item it replaced warrant an adjustment. If
                         they decide that the differences are not major, the commodity analysts
                         said they will code the substitution as comparable. Their supervisors
                         stated that, in some instances, the differences did not allow for clear-cut
                         decisions, and that the analysts’ judgment had to be exercised.


Adjustment Methods and   When a substitution is not comparable with the item it replaced,
Determining Which        commodity analysts make a direct adjustment or cause an indirect
Method to Use            adjustment to be made. When direct adjustments are made, BLS has
                         information on the ways the new version changed from the old version
                         and the value of those changes to adjust the price of the substitution
                         directly. BLS uses several different direct adjustment methods. Indirect
                         adjustments are made when there is insufficient information available with
                         which to make a direct adjustment. Unlike the direct adjustment methods,
                         which are based on information that is specific to the item being adjusted,




                         Page 14                                     GAO/GGD-99-84 Consumer Price Index
                                         B-279259




                                         the two indirect adjustment methods that BLS uses, class-mean and linking,
                                         are based on averages of other price changes experienced that month.12

                                         The number of direct and indirect adjustments made to nonrent
                                         substitutions in calendar year 1997 totaled 12,131. This number was about
                                         42 percent of the 28,881 nonrent adjustments. Table 1 shows the number
                                         of direct and indirect adjustments by major CPI component. As the table
                                         indicates, when the number of substitutions is compared to the number of
                                         price quotations, substitutions occur unevenly among the CPI components.


Table 1: CPI Price Quotations, Substitutions, and Methods of Adjustment by Major Components for 1997
                                                                                   Adjusted substitutions by adjustment
                                                                                                 method

                                   Number of           Number of          Substitutions                                Indirect adjustments
                                        price        substitutions         not adjusted                Direct          Class-mean           Linking
Major components                   quotations          (accepted)         (comparable)           adjustments              methoda           method
Food and beverages                    459,635                  6,485                3,640                   107                    31          2,707
        b
Housing                               136,430                  4,047                2,453                   132                1,023               439
Apparel and upkeep                     76,736                  9,797                6,598                 1,223                1,796               180
Transportation                         94,336                  5,660                2,699                 1,837                   823              301
Medical care                           50,237                  1,116                   355                  336                      3             422
Entertainment                          32,985                  1,327                   763                    95                  321              148
Other goods and services               22,440                    422                   215                    40                   52              115
All items                             872,829                 28,881               16,750                 3,770                4,049           4,312
                                         a
                                          This column includes some adjustments made using a method that BLS was phasing out of use.
                                         BLS estimated the number of these adjustments to be at least 77.
                                         b
                                          Housing totals exclude residential rent adjustments. The full version of this table, which is
                                         contained in appendix VIII, contains data on rent.

                                         Source: BLS.



                                         Regardless of whether an adjustment is direct or indirect, the basic intent
                                         is to keep the known differences in characteristics between substitutions
                                         and original items from affecting the CPI’s measurement of price change.
                                         The CPI is designed to include only “pure” price changes, and the
                                         adjustments that BLS makes in connection with substitutions are intended
                                         to separate pure price increases from increases due to other factors, such
                                         as improvements in quality.13

                                         12
                                          BLS refers to these two methods as “imputations.” For ease in reading we use the term “indirect
                                         adjustments” to refer to the class-mean and linking methods.
                                         13
                                             The difference in price that remains after pure price is referred to generically as quality by BLS.



                                         Page 15                                                         GAO/GGD-99-84 Consumer Price Index
                         B-279259




Direct Adjustments       About 31 percent of the 12,131 nonrent adjustments in 1997 were made
                         using direct adjustment methods. On the basis of BLS’ use, these direct
                         adjustments can be classified into a manufacturers’ cost method; a
                         statistical modeling method, which BLS refers to as a hedonic regression
                         method; and a final method that can best be described as a catchall
                         “other,” which includes adjustments for changes in size or quantity or for
                         error.

                     •   For the manufacturers’ cost method, BLS uses cost information from
                         manufacturers to identify individual characteristics or options that have
                         changed and the cost of those changes. In 1997, all of the manufacturers’
                         cost adjustments were in the new and used vehicle item strata. BLS collects
                         information from automakers on the changes they make each model year
                         and the cost of those changes. After screening these changes to make sure
                         they meet BLS’ criteria for quality, BLS then uses the information to make
                         adjustments. BLS collects this change and cost information for a sample of
                         domestically produced models each year. The great majority of the 1,837
                         direct adjustments made in the transportation component of the CPI used
                         the manufacturers’ cost method.

                         In an automobile substitution that we reviewed, the commodity analyst
                         used an automaker’s information to determine that a 1998 model had
                         lower fuel emissions and safer air bags than the 1997 model. Using
                         information from the automaker, the analyst valued these improvements at
                         $135. The analyst then made a direct adjustment, reducing the difference
                         in price between the 1997 model and the 1998 model by $135 to account
                         for the change in quality.14 The remaining difference in price (expressed as
                         a percentage) went into the CPI as the pure price change.

                     •   Under the hedonic regression method, BLS uses statistical models to
                         estimate a value for individual characteristics of a product, such as the
                         value of adding a lining and hood to a coat. According to BLS, in 1997,
                         statistical models were used to make most of the 1,223 direct adjustments
                         to apparel items, which were the only type of item for which BLS had
                         statistical models.15




                         14
                          As of January 1999, BLS no longer treats changes made solely to meet air quality standards as quality
                         improvements for determining the rates of price changes for the CPI.
                         15
                           BLS has developed a statistical model for computer items, which it began using in 1998, and another
                         for televisions, which it began using in January 1999.



                         Page 16                                                    GAO/GGD-99-84 Consumer Price Index
    B-279259




    For one of the apparel substitutions we reviewed, for example, the
    commodity analyst used a statistical model to assign value to certain
    differences—design, fiber content, and cleaning method—between two
    versions of a woman’s coats. Based on information from the model, the
    analyst assigned these differences a combined value of $33.85. This value
    was then, in effect, subtracted from the difference in price between the old
    and new versions.

•   A broad array of direct price adjustments fall within the other direct
    method, including those for a change in size (e.g., 8 ounces instead of 10
    ounces), for a change in the number of units (e.g., 15 tablets instead of 12
    tablets), and to correct for errors made by price takers (e.g., to correct an
    inaccurate recording of the number of ounces in a container). BLS
    considers adjustments for size or unit count to be direct adjustments
    because they adjust for a particular countable characteristic of a good or
    service. Most of the direct adjustments in the major components of the CPI,
    other than apparel and transportation, were these types of direct
    adjustments.

    Commodity analysts we interviewed who reviewed automobile and
    apparel substitutions said that they often made direct adjustments to
    increase the level of similarity between old and new versions. Without
    these adjustments, the substitutions would have been classified as
    comparable, according to the analysts, because the differences were
    minor. For example, the commodity analyst who made the $135 direct
    adjustment in the previously mentioned automobile substitution said that
    the difference in fuel emissions and air bags between the two models
    would have been insufficient to make them not comparable. In addition,
    regardless of type of product or service, analysts who have the information
    available with which to make a direct adjustment may do so without first
    going through the step of contemplating whether the new and old versions
    are comparable.

    When using the direct adjustment methods based on manufacturers’ cost
    information or statistical models, BLS first estimates the value of quality
    changes and then removes that value from the difference in price between
    the substitution and the original item. The pure price change is the
    residual after the adjustment for quality is made. For example, in the
    automobile substitution we discussed previously, the 1998 model cost
    $23,180 while the 1997 model cost $22,104, a difference of $1,076. If no
    adjustment had been made, this difference would have constituted a price
    increase of 4.9 percent. However, an adjustment of $135 was made for
    improvements in fuel emission levels and air bag safety. As a result, the


    Page 17                                      GAO/GGD-99-84 Consumer Price Index
                         B-279259




                         adjusted price difference between the 1998 model and the 1997 model was
                         $941, or 4.3 percent. Therefore, in this case, 0.6 percent (or $135) of the
                         unadjusted price difference of 4.9 percent was attributed to quality
                         change, and the remainder of 4.3 percent was attributed to pure price
                         change.

                         Appendix IV provides more information about the direct adjustment
                         method, including examples that illustrate its use.

Indirect Adjustments     BLS uses either a class-mean method or a linking method to make indirect
                         adjustments. Both of these indirect methods first estimate the pure price
                         change and exclude any residual, which is reverse of what the direct
                         adjustment methods do where price change is the residual. Both methods
                         impute the pure price change by averaging the rates of price changes
                         experienced by the same type of items in the CPI. This averaging is done by
                         computer program. However, the same type of items are defined narrowly
                         for the class-mean method but are defined much more broadly for the
                         linking method.

                         Because the linking method is, in effect, based on the average rates of
                         price change for all of the same type of items in a geographic location, it
                         cannot affect the average rates of price change for that geographic
                         location. In addition, estimates by BLS researchers indicate that, overall,
                         the linking method causes lower price increases than the class-mean
                         method. The researchers estimated that the average monthly price change
                         for adjustments made with the linking method was less than 1 percent in
                         1995 while the average monthly price change for adjustments made with
                         the class-mean method was about 5 percent. Because the methods by
                         definition use different sets of goods and services in the calculations, BLS’
                         estimates of 1 percent and 5 percent reflect differences due to the manner
                         in which the calculations were made as well as differences in the type of
                         products used in the calculations. BLS designed the class-mean method
                         with the intention to produce more accurate rates of price change for
                         items involving new models and products than the linking method because
                         in BLS’ research, new models and products generally incorporated larger
                         price changes.

Class-Mean Adjustments   According to BLS, the class-mean method is the designated method for item
                         strata where new models or product lines are introduced regularly. These
                         strata have included vehicles and apparel items if direct adjustments could
                         not be made; household appliances; and other household goods.




                         Page 18                                      GAO/GGD-99-84 Consumer Price Index
B-279259




The item that was substituted is put aside under the class-mean method in
computing price changes. Instead, to impute a price change, BLS selects a
subset of substitutions from the same item stratum and geographic
location as the original item. For example, if a price taker who tracks
prices in Urbantown substituted one refrigerator for another, BLS would
use this Urbantown stratum (refrigerators and home freezers) to compute
the price change rather than specific price information on the refrigerator
that the price taker reported. BLS has decided that the most appropriate
subset of substitutions for items involving new models or product lines are
(1) those that were not adjusted (i.e., those that were comparable) and
(2) those that were adjusted through a direct adjustment method. All of
the comparable and direct-adjusted substitutions within a stratum are to
be used. According to BLS, this subset of substitutions is the best
approximation of the pure price changes that come about with the
introduction of new models and product lines.

New class-means are computed each month, and the class-mean for a
particular item stratum is assigned, in effect, to all class-mean
substitutions in that stratum and location. For example, if the class-mean
for Urbantown’s refrigerators and home freezers stratum was an increase
of 10 percent and five class-mean substitutions to that stratum were made,
each of those substitutions would reflect a price increase of 10 percent. In
turn, that percentage would be used five times in computing the CPI index
for Urbantown for that particular month.

In 1997, the class-mean method accounted for about 33 percent of the
12,131 adjustments. One of those class-mean adjustments was for another
car substitution that we reviewed. The analyst knew that improvements
had been made to the 1998 model but did not have sufficient information
to make a direct quality adjustment. In this instance, the 1998 model cost
$14,408 while the 1997 model cost $14,010, a price change of 2.8 percent.
However, the class-mean method set aside this actual price increase and
imputed a price increase based on other new automobile models in the
same geographic area. There were 11 comparable and directly adjusted
new models (substitutions) in that geographic area for which BLS had
collected information. As these 11 models had increased in price by an
average of 4.0 percent, the price of the 1998 model was imputed to have
risen by 4.0 percent.

Appendix V provides more information about the class-mean method,
including examples that illustrate its use.




Page 19                                     GAO/GGD-99-84 Consumer Price Index
                             B-279259




Linking Method Adjustments   According to BLS, the linking adjustment method is to be used when
                             neither a direct adjustment nor a class-mean adjustment can be made. BLS
                             has designated the linking method for item strata where new models or
                             product lines are not introduced fairly regularly, such as food items. It is
                             also the designated method for most services, such as medical services,
                             because, in BLS’ opinion, changes in the quality of services are difficult to
                             measure. The linking method was used for about 36 percent of the 12,131
                             adjustments made in 1997. It was used in nearly all of the nonrent item
                             strata in which an adjustment was made and accounted for a majority of
                             the adjustments in the food and beverage component of the CPI and the
                             medical care component.

                             As with the class-mean method, substitutions under the linking method are
                             put aside in calculating price changes for the month. Instead, a rate of
                             price change is imputed from the same item stratum and geographic
                             location as the original item. All items in the item strata with usable prices
                             are to be used to compute a weighted-average price change, which is
                             expressed as a percentage.16 BLS then, in effect, assigns this percentage or
                             rate of change to all linked adjustments in that item stratum and
                             geographic location for the month in which the calculation was made.

                             The linking method was used for a soup substitution that we reviewed. On
                             a per-ounce basis, the new package of soup cost 34.5 percent more than
                             the old package of soup. However, this price increase was set aside
                             because of differences in soup ingredients and package size. Instead, a
                             price increase of 0.6 percent was imputed under the linking method, based
                             on the items with usable prices in that item stratum and geographic region.

                             In addition to being used for linked substitutions, this method is also used
                             when a substitution is unacceptable and killed. According to BLS, when a
                             substitution is killed, an imputed rate of price change is assigned to the old
                             item (the item for which the killed substitution was going to replace), and
                             the same calculation that is used in the linking method is used for this
                             price change.

                             Appendix VI provides more information about the linking method,
                             including examples that illustrate its use.



                             16
                               BLS defines items with usable prices as those for which (1) no substitutions were made,
                             (2) substitutions were made and no adjustments were necessary, and (3) substitutions were made and
                             then adjusted using a direct method. Items that do not have usable prices are those, for example, that
                             were temporarily unavailable for pricing because they were out of season.



                             Page 20                                                     GAO/GGD-99-84 Consumer Price Index
                      B-279259




                      BLS relies on supervisors to review the substitutions decisions made by
Description of How    commodity analysts. According to BLS, there are no guidelines or policies
Supervisors Review    in writing for supervisors to follow in selecting and reviewing the
Commodity Analysts’   decisions made by analysts. In practice, according to the BLS Branch Chief
                      for Consumer Prices, there is an unwritten policy that supervisors each
Decisions             month are to review large price changes going into the CPI. When those
                      changes involve substitutions, the decisions made by commodity analysts
                      are reviewed. Few other adjustments are reviewed. BLS has no policy to
                      randomly or otherwise select and review substitution decisions.

                      According to BLS officials, supervisors and commodity analysts working
                      together set the levels of price change that trigger a review. These levels
                      can vary by item strata and can differ between price increases and price
                      decreases within a stratum. According to BLS officials, large price changes
                      are reviewed because they are the price changes that could have the
                      greatest impact on the indexes. BLS officials said BLS does not keep count
                      of the number of substitutions receiving supervisory review.

                      According to BLS officials, supervisors are to examine the reasons for the
                      analysts’ decisions for price quotations that produce large price changes.
                      The supervisors reported that they frequently did this by examining the
                      brief explanations that are printed on the price quotations lists that they
                      review. Commodity analysts write these explanations when they review
                      substitutions. If the supervisors are not satisfied with the explanations in
                      these messages, they are to ask the analysts to explain their decisions in
                      person. All the supervisors said that they usually accept the analysts’
                      explanations, either through the written explanations or in person.

                      Beyond the specific reviews performed by supervisors, BLS does not have a
                      program of assessing the decisionmaking patterns of commodity analysts.
                      However, BLS has studied the process by which commodity analysts make
                      decisions at least three times in the 1980s and early 1990s, and the studies
                      found that the decisionmaking process is susceptible to producing errors
                      and inconsistencies.

                      As explained by one study, substitution review is potentially prone to
                      (1) visual search errors, as commodity analysts locate information in
                      product descriptions; (2) comprehension errors, as they extract the
                      content of these descriptions; and (3) consistency errors, if they irregularly
                      apply a particular rule. In addition, according to the study, inconsistencies
                      may occur when one commodity analyst does the work of another
                      commodity analyst or when different commodity analysts use different



                      Page 21                                      GAO/GGD-99-84 Consumer Price Index
              B-279259




              approaches to substitution of their product groups, resulting in different
              products being treated in different ways. Each of the three studies
              recommended actions intended to promote greater control over the
              decisionmaking process to reduce the potential for errors and
              inconsistencies. We found that BLS’ implementation of these
              recommendations has varied, with action in most cases diminishing after
              initial steps were taken. According to officials we interviewed, BLS now
              takes the position that these controls are not required for experienced
              commodity analysts.

              Appendix III provides further information about how commodity analysts
              decide whether to make adjustments and how supervisors review those
              decisions.


              BLS faces a difficult task each month of collecting tens of thousands of
Conclusions   prices, reviewing those prices, computing the CPI, and ensuring its
              accuracy in a timely manner. In this process, BLS commodity analysts
              review the substitutions that price takers make each month for items they
              cannot find. Substitutions are not inconsequential because BLS has
              determined that they can have a significant impact on the CPI.

              To account for substitutions, BLS has developed a set of procedures and
              methods to determine whether a substitution is comparable to the item it
              replaced and, if not, what adjustment to make for its inclusion into the CPI.
              BLS depends on commodity analysts to make the decisions on whether
              substitutions are comparable and, to a lesser extent, which adjustment
              method to apply. By the very nature of the differences that can exist
              between original and substituted items, commodity analysts must exercise
              a degree of professional judgment in making decisions, more with some
              substitutions and less with others. They make these judgments with little
              or no written criteria to follow. In terms of review or quality controls,
              supervisors review large price changes and rely on commodity analysts for
              explanations for their decisions. Beyond these reviews, BLS has no
              program to review commodity analysts’ decisions either preissuance or
              postissuance of the CPI.

              In the past, BLS has studied the process by which the commodity analysts
              make decisions and found that the decisionmaking process was
              susceptible to producing errors and inconsistencies. But BLS did not fully
              act upon the recommendations that came from those studies; its actions
              on most recommendations diminished after it took initial steps.



              Page 22                                      GAO/GGD-99-84 Consumer Price Index
                     B-279259




                     We found no evidence to indicate whether errors or inconsistencies in
                     commodity analysts’ decisions or lack of comprehensive review of those
                     decisions has had a material effect on the calculation of the CPI. We are not
                     suggesting that the CPI is inaccurate. Indeed, any errors or inconsistencies
                     could be random in nature and in effect cancel each other out without
                     material effect on the CPI. However, prior BLS studies have noted that the
                     decisionmaking process was susceptible to producing errors and
                     inconsistencies—a situation that has not been remedied by sustained
                     corrective action. Given this susceptibility, it is a matter of sound
                     management practice to (1) periodically evaluate the degree to which
                     commodity analysts are actually making errors and inconsistent decisions
                     and (2) evaluate the material effects, if any, of errors or inconsistencies
                     upon the CPI. Moreover, the need for such management practices is
                     underscored by the significant uses made of the CPI in the public and
                     private sectors and the effect those uses can have on individuals and
                     businesses nationwide.


                     To help ensure that the CPI is protected from potential effects of errors
Recommendation       and/or inconsistencies resulting from commodity analysts’ substitution
                     decisionmaking, we recommend that the Commissioner of BLS evaluate, on
                     a periodic basis, the degree of consistency and accuracy in analysts’
                     substitution determinations and the resulting effects on the CPI.


                     In a letter dated April 7, 1999, the Commissioner of BLS provided comments
Agency Comments      on a draft of this report for the Department of Labor. The Commissioner
and Our Evaluation   commented that our descriptions of the procedures will be useful to BLS
                     and to many CPI users. In discussing our recommendation, she said that
                     periodically evaluating the degree of consistency and accuracy in analysts’
                     substitution decisions and the resulting effects on the CPI was certainly a
                     desirable thing to do and that BLS would explore ways it could enhance its
                     existing review processes in that area.

                     The Commissioner mentioned several ways in which BLS could implement
                     the recommendation. The first two ways were to (1) evaluate
                     enhancements to the data used to monitor commodity analysts’ handling
                     of substitutions and (2) review the documention of commodity analyst
                     procedures to see if it could be made more complete. We agree that these
                     actions would represent good first steps but note that they would not
                     necessarily implement all the parts of our recommendation. For example,
                     more complete documentation of commodity analyst procedures could



                     Page 23                                      GAO/GGD-99-84 Consumer Price Index
B-279259




improve the analysts’ consistency and accuracy but would not measure the
degree of consistency and accuracy.

A third way mentioned by the Commissioner was for BLS to consider the
possible use of “expert systems” software (i.e., computer software that
helps individuals make consistent and accurate decisions on complex
issues) to assist commodity analysts and enhance the consistency of their
decisions. BLS developed an expert software system several years ago,
which BLS did not subsequently implement for reasons the Commissioner
explains in her comments. BLS demonstrated this software to us. While we
have not evaluated such software, the demonstration to us showed that
such a system could help commodity analysts enhance the consistency of
their decisions. In addition, such a system might also help BLS to assess the
degree of accuracy in its decisions. However, BLS would have to develop
specific methods for using the data from an expert system to meet the
second part of our recommendation, which was to evaluate the effect of
the analysts’ decisions on the CPI.

The Commissioner’s letter is reprinted in appendix IX. She made
additional comments in her letter, which we addressed as appropriate in
appendix IX.


We are sending copies of this report to Alexis Herman, Secretary of Labor;
Katharine Abraham, the Commissioner of BLS; Patsy Mink, Ranking
Minority Member, Subcommittee on Criminal Justice, Drug Policy, and
Human Resources; and other interested parties. We will also make this
report available to others on request.




Page 24                                      GAO/GGD-99-84 Consumer Price Index
B-279259




Major contributors to this report are listed in appendix X. If you have any
questions about this report, please call me on (202) 512-8676.




Laurie E. Ekstrand
Associate Director, Federal Management
  and Workforce Issues




Page 25                                     GAO/GGD-99-84 Consumer Price Index
Contents



Letter                                                                                          1


Appendix I                                                                                     30
                       Methodology Used to Select Examples, Substitution Decisions,            30
Additional               and Commodity Analysts
Information on Our     Survey Methodology                                                      32
Methodology
Appendix II                                                                                    34
                       Construction of the CPI                                                 35
Background
Information on the
Consumer Price Index
Appendix III                                                                                   41
                       Role of Commodity Analysts                                              41
How Commodity          Use of Commodity Review Listings                                        42
Analysts Decide to     Process of Making Decisions About Substitutions                         43
                       BLS Processes and Procedures for Reviewing Substitution                 53
Make Adjustments         Decisions

Appendix IV                                                                                    57
                       Manufacturers’ Costs                                                    57
Direct Adjustments     Example 1 - New Car (Under the New Cars Item Stratum)                   62
                       Hedonic Regression Estimates                                            63
                       Example 2 - Woman’s Coat (Under the Women’s Coats and                   68
                         Jackets Item Stratum)
                       Other Direct Adjustments                                                70
                       Example 3 - Season Ticket (Under the Admission to Sporting              72
                         Events Item Stratum)
                       Example 4 - Canned Soup (Under the Canned and Packaged Soup             73
                         Item Stratum)

Appendix V                                                                                     76
                       Background                                                              76
Class-Mean Method of   How Price Changes Are Calculated With the Class-Mean Method             78
Adjustment             How Quality Improvements Are Accounted for in the Class-Mean            80
                         Method
                       Examples of BLS’ Use of the Class-Mean Method                           81




                       Page 26                                  GAO/GGD-99-84 Consumer Price Index
                        Contents




                        Example 1 - Bedroom Linens (Under the Linens, Curtains,                  81
                          Drapes, and Sewing Materials Item Stratum)
                        Example 2 - Woman’s Parka (Under the Women’s Coats and                   83
                          Jackets Item Stratum)
                        Example 3 - New Car (Under the New Cars Item Stratum)                    85

Appendix VI                                                                                      88
                        Background                                                               88
Linking Method of       How Price Changes Are Calculated With the Linking Method                 89
Adjustment              How BLS Accounts for Quality Change in Using the Linking                 90
                          Method
                        Examples of BLS’ Use of the Linking Method                               91
                        Example 1 - Packaged Soup (Under Canned and Packaged Soup                92
                          Item Stratum)
                        Example 2 - Multivitamin Tablets (Under Internal, Respiratory,           94
                          and Over-the-Counter Drugs Item Stratum)

Appendix VII                                                                                     96
                        Background                                                               96
Measurement of          Adjustments for Changes in Units                                         98
Residential and         Adjustments Are Made by Computer and by Analysts                        100
                        Addition and Deletion of Housing Units                                  104
Homeowners’             Accounting for Quality Change                                           105
Equivalent Rents        Examples of BLS’ Adjustments for Residential Rent and REQ               106
                        Example 1 - Automatic Adjustment                                        106
                        Example 2 - Pricing-Links-Cancel Adjustment                             108
                        Example 3 - Links-Pause Adjustment                                      109

Appendix VIII                                                                                   112

CPI Price Quotations,
Substitutions, and
Methods of
Adjustment for 1997




                        Page 27                                   GAO/GGD-99-84 Consumer Price Index
                        Contents




Appendix IX                                                                                    153

Comments From the
Bureau of Labor
Statistics
Appendix X                                                                                     158

Major Contributors to
This Report
Related GAO Products                                                                           160




                        Abbreviations

                        BLS        Bureau of Labor Statistics
                        CAL        Commodity analyst listing
                        CEX        Consumer Expenditure Survey
                        CRL        Commodity review listing
                        CPI        Consumer Price Index
                        CPI-U      Consumer Price Index representing all urban consumers
                        CPI-W      Consumer Price Index representing all urban wage earners
                                        and clerical workers
                        ELI        Entry level item
                        POPS       Point-of-Purchase Survey
                        PPI        Producer Price Index
                        PSU        Primary sampling unit
                        REQ        Homeowners’ equivalent rent


                        Page 28                                  GAO/GGD-99-84 Consumer Price Index
Page 29   GAO/GGD-99-84 Consumer Price Index
Appendix I

Additional Information on Our Methodology


                      In our scope and methodology section, we explained how we obtained the
                      information necessary to describe the Bureau of Labor Statistics’ (BLS)
                      methods and procedures for reviewing substitutions. This work included
                      judgmentally selecting 13 examples of substitution decisions to illustrate
                      the methods. To select these examples, we discussed 136 substitution
                      decisions with 19 commodity analysts1 in BLS’ Consumer Prices Branch.
                      This appendix explains how we selected these examples, substitution
                      decisions, and commodity analysts. We also explain how we administered
                      a questionnaire to BLS’ commodity analysts.


                      We judgmentally selected a small number of examples to illustrate the use
Methodology Used to   of BLS’ quality adjustment methods across the major components of the
Select Examples,      Consumer Price Index (CPI). We discussed these examples with the
Substitution          commodity analysts who had reviewed them and requested that the
                      analysts consult their supervisors about our selection. The analysts and
Decisions, and        their supervisors agreed that the examples we had selected were broadly
Commodity Analysts    representative of common situations that analysts encounter when
                      reviewing substitutions. Because substitutions for the nonrent items of the
                      CPI are very different from substitutions for rent items, we used different
                      procedures to select examples for each group.

                      For the nonrent items, we first obtained a selection of the commodity
                      review listings (CRL) that commodity analysts use when reviewing
                      substitutions and deciding whether an adjustment should be made. BLS
                      automatically generates a CRL each time a substitution is made, and these
                      lists are the main source of information about substitutions and the items
                      they replace. Because BLS made 28,881 substitutions for nonrent items in
                      1997, we limited our request to 2 months of CRLs. We requested CRLs for
                      substitutions in October and November 1997 because we wanted the most
                      recent months relative to when we began our work, so that the analysts
                      would be most likely to recall the reasons for which they had made their
                      decisions. According to BLS, 6,257 substitutions were made for nonrent
                      items in October and November 1997. BLS provided us with the CRLs for
                      substitutions that had been made in the priced, nonrent item strata of the
                      CPI during those 2 months.2


                      To arrive at a small number of illustrative examples from the 6,257 nonrent
                      substitutions, we made a series of selections to reduce the number of

                      1
                       In one instance, the analyst responsible for the item strata was no longer employed by BLS. However,
                      the analyst’s supervisor discussed the substitutions with us.
                      2
                       According to BLS data, there were 204 nonrent item strata, of which 181 were priced.



                      Page 30                                                    GAO/GGD-99-84 Consumer Price Index
Appendix I
Additional Information on Our Methodology




substitutions we were considering. First, we selected 18 of the 181 nonrent
priced item strata for an in-depth review. We made our initial selection by
item strata because the CRLs were organized by item strata and because BLS
makes its decisions on which adjustment methods to apply by item strata.
We judgmentally selected these item strata to ensure that we covered
(1) all of BLS’ direct and indirect adjustment methods and (2) all the major
components of the CPI with the exception of the other goods and services
component.3 The major components of the CPI that we covered were food
and beverages, apparel and upkeep, housing, transportation, medical care,
and entertainment. In a few instances, our selection of the item strata was
influenced by the opinions of experts we had interviewed. For example,
we included poultry items because an official at the Department of
Agriculture had suggested that poultry would be interesting to study
because of recent developments in the ways it was cut and packaged. Our
selection of 18 item strata contained 1,212 substitutions.

Having made this selection of 18 item strata, we examined the CRLs for
these strata and selected a number of them for discussion with the
commodity analysts. We selected CRLs that appeared to illustrate the
adjustment methods used in each item stratum. We based our selection on
(1) the number of substitutions and types of adjustments that were made
in each item stratum and (2) preliminary discussions with commodity
analysts.4

We met with the 9 commodity analysts and 1 supervisor who were
responsible for the 18 item strata and had detailed discussions with them
on 106 CRLs. BLS assigns commodity analysts to particular item strata and,
therefore, one analyst reviews all the substitutions in an item strata. We
discussed the selected CRLs with the analysts and also asked them to
explain the procedures they normally follow when reviewing substitutions.

After these interviews, we judgmentally selected 10 nonrent substitutions
to serve as illustrative examples for this report. These examples cover the
six major components of the CPI. They also cover comparability decisions

3
 The “other” component was a miscellaneous collection of other items that did not fit into the other
major categories of the CPI. It consisted of items such as personal-care appliances and services, school
books, and day-care services. The “other” component had fewer than 100 of the 6,257 substitutions
that occurred in October and November 1997.
4
 We conducted some preliminary discussions to obtain an idea of what analysts do, so that we would
be better able to later hold discussions with the analysts responsible for the 18 item strata. For these
preliminary discussions, we simply picked out a few CRLs—25 in 6 additional item strata—that looked
as if they would be useful to talk to analysts about (e.g., involved the various adjustment
methods)—and then discussed them with the responsible analysts. These 25 substitutions are included
in the 136 we report that we conducted. We used information on 2 of these 25 substitutions in the
letter portion of this report.



Page 31                                                     GAO/GGD-99-84 Consumer Price Index
                     Appendix I
                     Additional Information on Our Methodology




                     and BLS’ direct and indirect adjustment methods. We did not include an
                     example of the only other method that was used, the overlap method,
                     because it was used for less than 100 of the 28,881 substitutions and is
                     being phased out, according to BLS officials. We selected our examples to
                     meet the following criteria: (1) it reflected a common pattern of
                     decisionmaking in that item stratum or item strata, (2) it lent itself to a
                     straightforward description, (3) it was not atypical or unusual in any way,
                     and (4) it did not contain errors made by the analysts. To ensure that our
                     selections did indeed meet these criteria, we showed our examples and
                     criteria to the analysts and their supervisors; and we asked them if our
                     examples were good illustrations of their decisionmaking in those strata,
                     and if the examples met the criteria. The analysts and their supervisors
                     agreed this was the case.

                     In addition to the 18 nonrent item strata, we also covered 2 rent item
                     strata: residential rent and homeowners’ equivalent rent. The procedures
                     used to make adjustments in these item strata differ from those used
                     elsewhere in the CPI. See appendix VII for information about these
                     adjustments.

                     The procedures we used to select CRLs as illustrative examples for the rent
                     item strata differed from those we used in the nonrent item strata. We
                     randomly selected one example for each of the three main methods of
                     adjustment used in the rent strata. We discussed these examples with the
                     commodity analysts and their supervisor for the rent strata; and, when the
                     analysts and/or their supervisor noted that the initial selections were
                     atypical or problematic, we randomly selected alternative examples. In all,
                     we discussed five rent item CRLs with four commodity analysts and their
                     supervisors.

                     To select the examples, we discussed 120 specific substitutions (115
                     nonrent plus 5 rent substitutions) with 19 commodity analysts (15 nonrent
                     plus 4 rent commodity analysts). In addition, we discussed 16 nonrent
                     substitutions with one nonrent supervisor. Following these discussions,
                     we selected 13 substitutions (10 nonrent plus 3 rent substitutions) to serve
                     as illustrative examples in this report.


                     To gain a better understanding of the processes the commodity analysts
Survey Methodology   follow when making substitution decisions and their educational and
                     professional backgrounds, we asked them to complete a written
                     questionnaire. We divided the commodity analysts into two groups: those



                     Page 32                                     GAO/GGD-99-84 Consumer Price Index
Appendix I
Additional Information on Our Methodology




who review residential rent and those who review all nonrent items in the
CPI. The analysts who review residential rent follow different procedures
than the others; therefore, they could not be asked all the same questions.
All four residential rent analysts, as of September 1998, responded to a
limited version of this survey. All 24 individuals who were nonrent
analysts in October and November 1997 and were still analysts in
April 1998 responded to our full survey. Three nonrent supervisors, who
also had some responsibilities for deciding whether to make substitution
decisions and adjustments, also responded.




Page 33                                     GAO/GGD-99-84 Consumer Price Index
Appendix II

Background Information on the Consumer
Price Index

              BLS produces the CPI by measuring the average change over time in the
              prices paid by urban consumers for a fixed market basket of consumer
              goods and services. The selection of items for the market basket is
              determined from detailed records of purchases made by thousands of
              individuals and families, as reported on periodic surveys. The items
              selected for the market basket, such as potatoes, are to be priced each
              month at specific retail outlets, such as grocery stores and supermarkets,
              in urban areas throughout the country. According to BLS, in 1997, price
              takers collected the prices of about 94,000 items (goods and services) in 85
              urban areas of the country. These prices were collected from about 30,000
              retail and service establishments and from about 46,000 landlords and
              tenants, who provided data on housing units.

              The CPI is used as a measure of price changes to make economic decisions
              in the private and public sectors. According to BLS, the CPI has three major
              uses indicated as follows:

              Economic indicator of inflation. The administration, Congress, and the
              Federal Reserve use trends in the CPI as an aid to formulating fiscal and
              monetary policies. Business and labor leaders as well as private citizens
              use the CPI as a guide to making economic decisions.

              Escalator for wages, benefit payments, and tax brackets. The CPI is used by
              collective bargaining units to adjust the wages of workers. Also, it is the
              basis for automatic changes in some federal benefit payments. For
              example, in December 1997, as a result of changes in the CPI, 44 million
              Social Security beneficiaries and 6.5 million Supplemental Security Income
              recipients had their benefits adjusted for inflation. More than 21 million
              food-stamp recipients in 1997 were affected by changes in the CPI.1 Also,
              millions of railroad, military, and federal civilian retirees and survivors are
              affected by changes in the CPI. The CPI is also used to adjust key elements
              of the individual income tax to limit the extent to which individuals must
              pay higher taxes solely because of inflation. For example, the amount
              allowed for personal exemption, the amount of the standard deduction,
              and tax brackets are adjusted annually according to changes in the CPI.

              Deflator of selected economic statistical data series. The CPI is used to
              adjust selected economic statistical series for price changes and to
              translate these series into inflation-free dollars. Examples of data series


              1
               The federal expenditure estimates given on page 1 of this report do not include the food stamps
              program or the school-lunch program because these programs are affected by the “food away from
              home” CPI subindex, whereas the other programs use the overall CPI that includes all items.



              Page 34                                                   GAO/GGD-99-84 Consumer Price Index
                      Appendix II
                      Background Information on the Consumer
                      Price Index




                      that are adjusted by the CPI include retail sales, hourly and weekly
                      earnings, and components of the National Income and Product Accounts.

                      The CPI was initiated during World War I, when rapid increases in the
                      prices of goods and services, particularly in shipbuilding centers where
                      workers were demanding wage adjustments, made such an index essential
                      for calculating cost-of-living adjustments. In 1921, BLS began regular
                      publication of an index representing the expenditures of urban wage and
                      clerical workers, which was then called the Cost-of-Living Index. The
                      name of the index was changed to the CPI following controversy during
                      World War II over the index’s validity as a measure of the cost of living.
                      According to BLS, the CPI has always been a measure of the changes in
                      prices for goods and services purchased for family living.

                      Major revisions were made to the CPI about once each decade to update
                      the fixed market basket, with the most recent revision occurring in
                      January 1998. Because consumers’ buying habits change, new studies were
                      made of what goods and services consumers were purchasing, and major
                      revisions to the CPI were made in 1940, 1953, 1964, 1978, and 1987 as well
                      as 1998.2 In the 1978 major revision, BLS began publication of a new index
                      for all urban consumers—the CPI-U. According to BLS, the CPI-U, which
                      represents the expenditures of about 87 percent of the population, takes
                      into account the buying patterns of professional employees, part-time
                      workers, the self-employed, the unemployed, and retired people as well as
                      those previously covered in the CPI. BLS has continued publication of the
                      older index, the CPI-W, which represents the expenditures of urban wage
                      and clerical workers or about 32 percent of the population.


                      Construction of the CPI begins by selecting a collection of goods and
Construction of the   services that is usually bought by the reference population in the index.
CPI                   The collection of goods and services, called items, is known as the market
                      basket. The CPI market basket is developed from detailed expenditure
                      information that is provided by thousands of families and individuals who
                      participate in the Consumer Expenditure Survey (CEX), which is conducted
                      for BLS by the Bureau of the Census over several years. For example, the
                      1987 CPI revision was based on CEX data collected from 1982 through 1984,



                      2
                       For this report we reviewed item substitutions that BLS made during October and November of 1997.
                      At that time, the last major revision to the CPI was in 1987. Therefore, the information contained in
                      this report is based on the CPI structure for the 1987 revision rather than the 1998 revision. Although
                      there were changes between the 1987 and 1998 revisions, the general steps BLS follows to construct
                      the CPI described in this appendix did not change between the two CPI revisions.



                      Page 35                                                     GAO/GGD-99-84 Consumer Price Index
                           Appendix II
                           Background Information on the Consumer
                           Price Index




                           from about 29,000 individuals and families.3 Expenditure data from the CEX
                           are used to select the categories of items from which specific, unique
                           commodity and service items are selected to be priced for the CPI.

                           BLS measures price changes each month by checking the prices of the
                           items in the market basket and then comparing the aggregate costs of the
                           market basket with those for the previous month. BLS price takers obtain
                           prices for most of the items by visiting or contacting thousands of
                           retailers, service providers, and landlords and tenants each month.


Classification of Market   BLS classified all CEX expenditure items for the 1987 CPI revision into 206
Basket Items               item strata, which are arranged into 7 major components:4 (1) food and
                           beverages; (2) housing; (3) apparel and upkeep; (4) transportation;
                           (5) medical care; (6) entertainment; and (7) other goods and services, such
                           as haircuts, college tuition, and bank fees. Taxes that are directly
                           associated with the prices of specific goods and services, such as sales and
                           excise taxes, are also included.5

                           The 206 item strata are divided into specific subcategories, which are
                           called entry level items (ELI). For example, item stratum 0101 flour and
                           prepared flour mixes has two ELIs: flour (01011) and prepared flour mixes
                           (01012). All item strata have at least one ELI, and some strata have more
                           than one. Appendix VIII lists item strata and related ELIs.


Expenditure Weights of     Expenditure weights are used to give proportionate emphasis for price
Market Basket Items        changes of one item in relation to other items in the CPI. Expenditure
                           weights allow the CPI to distinguish between items that have a major
                           impact on consumers and to provide appropriate emphases to price
                           changes associated with these items. For example, if ground beef were
                           assigned a weight representing about one-third of 1 percent of the
                           expenditures of the typical urban consumer and if sirloin steak were
                           assigned a smaller weight representing less than one-tenth of 1 percent,

                           3
                            The 1998 CPI revision was based on CEX data collected from 1993 through 1995, from about 36,000
                           individuals and families.
                           4
                            In the January 1998 revision, the major categories changed from seven to eight and include food and
                           beverages, housing, apparel, transportation, medical care, recreation, education and communication,
                           and other goods and services. The number of item strata also changed from 206 to 211.
                           5
                            The CPI includes various governmental-charged user fees, such as water and sewerage charges, auto
                           registration fees, and vehicle tolls. Taxes not directly associated with the purchase of consumer goods
                           and services, such as income and Social Security taxes, are excluded. In addition, the CPI does not
                           include investment items, such as stocks, bonds, real estate, and life insurance because they relate to
                           savings, not daily living expenses.



                           Page 36                                                     GAO/GGD-99-84 Consumer Price Index
                         Appendix II
                         Background Information on the Consumer
                         Price Index




                         then the price changes of ground beef would have about 3 times as much
                         impact on the overall CPI as similar price changes for sirloin steak.

                         Weights derived from consumers’ expenditures, as reported in the CEX, are
                         assigned to the 206 item strata. To compute the weights, BLS first totals the
                         amount spent on an item stratum, such as white bread, during the base
                         weighting period by CEX respondents, who BLS refers to as consumer units.6
                         BLS then divides that total by the number of consumer units, which results
                         in an average expenditure per unit. Next, the average expenditures per
                         unit are weighted with data from the decennial census to represent the
                         U.S. urban population. To do so, the average expenditures per unit are
                         multiplied by certain factors to represent the geographic dispersion of the
                         urban population. Finally, these nationwide urban expenditures on the
                         market basket items are totaled into an aggregate amount. The 206
                         expenditure weights are the percentages of this aggregate amount that are
                         spent on each of the 206 item strata (e.g., white bread).

                         Expenditure weights remain fixed until the next major revision of the CPI
                         and serve as a benchmark from which price comparisons are calculated.
                         The weights of the components for the 1987 major revision are those that
                         have derived from the 1982 through 1984 CEX.


Relative Importance of   Relative importance is related to, but not the same as, expenditure
Market Basket Items      weights. Relative importance can be used to show the direct effect an item
                         has on the overall CPI price change because it shows the share of total
                         expenditure that would occur if consumed quantities of the items remain
                         constant. Although the expenditure weights remain fixed until a major
                         revision, which had occurred about every 10 years, the relative importance
                         changes over time, reflecting the effect of price changes.

                         Expenditure weights equal the relative importance percentages at the time
                         of a major revision. But since BLS maintains the quantities of the items as
                         the same amounts that were consumed in the base period, the relative
                         importance percentages change as a result of changing prices. Items
                         registering a greater-than-average price increase become relatively more
                         important. Conversely, items registering a smaller-than-average price
                         increase become relatively less important. Therefore, as the time between
                         major revisions increases, items with higher-than-average rates of inflation
                         have increasing rates of influence upon the CPI. As shown in figure II.1, the

                         6
                          The CEX collects data from “consumer units,” which are defined by BLS as either financially
                         independent, unrelated individuals or groups of individuals who pool their resources to make joint
                         consumption decisions.



                         Page 37                                                    GAO/GGD-99-84 Consumer Price Index
                                      Appendix II
                                      Background Information on the Consumer
                                      Price Index




                                      relative importance of medical care in the index for all urban consumers,
                                      which was 5.7 in December 1986, increased to 7.4 in December 1997
                                      because medical prices increased at a greater rate than the rate for the all
                                      items CPI—the overall CPI. During the same period, the relative importance
                                      of apparel and upkeep fell from 6.3 percent to 5.3 percent because apparel
                                      and upkeep prices increased at a lower rate than the all items CPI.


Figure II.1: Relative Importance of
Components in the CPI-U, 1986 and     Percent
1997
                                      40




                                      30




                                      20




                                      10




                                       0
                                            ing



                                                        ion




                                                                              es




                                                                                              p



                                                                                                           rvi d
                                                                                                                 s


                                                                                                                            re



                                                                                                                                           nt
                                                                                                              ce
                                                                                             ee



                                                                                                        se s an



                                                                                                                          ca




                                                                                                                                         e
                                                                         rag
                                            us



                                                       tat




                                                                                                                                      nm
                                                                                             pk




                                                                                                                          al
                                                                                                          d
                                           Ho




                                                                         ve
                                                       or




                                                                                        du




                                                                                                                                     tai
                                                                                                                      dic
                                                                                                       oo
                                                   sp



                                                                     be




                                                                                                                                  ter
                                                                                        an



                                                                                                   rg



                                                                                                                     Me
                                                  an



                                                                    nd




                                                                                                                                 En
                                                                                                  he
                                                                                   rel
                                                  Tr



                                                               da




                                                                                                  Ot
                                                                                   pa
                                                                o



                                                                               Ap
                                                             Fo




                                                        1986

                                                        1997



                                      Source: BLS.




                                      Page 38                                                                                    GAO/GGD-99-84 Consumer Price Index
                              Appendix II
                              Background Information on the Consumer
                              Price Index




Collecting Prices of Market   Each month, BLS price takers visit or call thousands of retail stores, service
Basket Items                  establishments, rental units, and doctors’ offices all over the United States.
                              Each month, they record the prices of about 80,000 items.7

                              To determine which retail outlets its price takers should visit to obtain
                              monthly price quotations for nonrent items, BLS sponsors the
                              Point-of-Purchase Survey (POPS), which is conducted by the Bureau of the
                              Census. The survey respondents are asked, by item categories such as
                              doctors, whether they made specific purchases and, if so, the names and
                              locations of all places of purchases and the expenditure amounts. BLS uses
                              the results from the survey to select outlets for pricing.

                              BLS price takers visit each selected retail outlet to initially select items that
                              will be priced either monthly or bimonthly. For each outlet, categories of
                              items are selected for pricing. Using probability selection methods that are
                              based on revenues and volume information that is provided by the retail
                              outlet, BLS price takers use a table of random numbers to select for pricing
                              a unique item within the specified categories.

                              BLS collects rent prices for rental units in a different manner than that used
                              to identify and price other items in the market basket. BLS uses monthly
                              price changes of rental units in the CPI housing survey for the residential
                              rent and homeowners’ equivalent rent items in the CPI housing component.8
                              Residential rent and homeowners’ equivalent rent are estimated from
                              approximately 36,000 rented units and 26,000 owned units in the survey.
                              Each month, BLS price takers obtain information from renter units on the
                              rent for the current month, the previous month, and the services that the
                              landlord provides. These data are used to measure changes in rent prices
                              for residential rent as well as homeowners’ equivalent rent. (For detailed
                              information about this process see app. VII).


Replacement of Market         Because the CPI uses a fixed market basket of goods and services, BLS price
Basket Items No Longer        takers are instructed to collect price information for the same item (e.g.,
Available for Pricing         one dozen pink carnations, with greenery, wrapped in paper, and not
                              delivered) each time they visit the retail outlet or rental unit. However, in
                              many instances, the same identical item is not available for purchase in
                              each subsequent visit. In these situations, price takers are to follow certain

                              7
                               Prices are not collected monthly on all items in the CPI. Some are collected bimonthly and rent is
                              collected every 6 months for housing units.
                              8
                               BLS determines the value of owner-occupied housing by using a rental equivalent method, which
                              estimates the amount of rent that would be paid if it were rented.



                              Page 39                                                     GAO/GGD-99-84 Consumer Price Index
Appendix II
Background Information on the Consumer
Price Index




procedures to make a substitution—selection of a new version
(replacement) that is similar to the old version of the item that is no longer
available.9 BLS has different procedures for the price takers to follow to
bring into the CPI new products or services from the POPS that are not
substitutions for items that are in the fixed market basket.

In selecting a substitution the price takers are to follow specific guidance
for choosing the new version. In general, the price taker is to select the
item with specifications most consistent with the old version. After the
price taker selects a new version and records the information about the
item, the information is sent to BLS headquarters in Washington, D.C.,
where it is coded, entered into computer systems, verified, and examined
by commodity analysts for inconsistencies. Appendix III describes further
the procedures commodity analysts are to follow in examining
substitutions.




9
 As described in appendix VII, price takers in the CPI housing survey return to the same address in
each collection period and record information about the residential unit at that address. Substitutions
do not take place between residential units as they do elsewhere in the CPI. However, as described in
appendix VII, adjustments are made to make the current unit similar to what it was at the prior price
collection.



Page 40                                                     GAO/GGD-99-84 Consumer Price Index
Appendix III

How Commodity Analysts Decide to Make
Adjustments

                    Commodity analysts review information collected by the price takers and
                    judge whether the substitution (new version) is an acceptable replacement
                    for the item that disappeared (old version) and, when acceptable, whether
                    it is comparable—similar—to the old version. Where versions are not
                    comparable, the analysts then select a method of adjustment. When the
                    substitutions have large price increases or decreases, supervisors review
                    the outcomes of the commodity analysts’ decisions.

                    According to BLS, the analyst’s knowledge and judgment about the
                    comparability of the two versions is very important because the analyst’s
                    decision determines the rate of price change that will be used in
                    computing the CPI. BLS further states that bias can enter the index if
                    substitutions are not carefully reviewed by commodity analysts.

                    This appendix provides information about (1) BLS’ procedures and
                    practices that are used by the commodity analysts to make their decisions
                    for substitutions of nonrent commodities and services and (2) the review
                    of those decisions by supervisors. (See app. VII for information about the
                    measurement of residential rent and homeowners equivalent rent.)


                    According to BLS, when receiving the price information that was collected
Role of Commodity   by the price takers and entered into BLS’ computer system, the commodity
Analysts            analysts—who are to have detailed knowledge about the particular goods
                    or services—check the data for completeness, accuracy, and consistency.
                    For example, if there are accuracy questions, analysts are to obtain a copy
                    of the form used by the responsible price taker to record price data to
                    verify that the actual collected values were captured in the system. The
                    analysts then make any necessary corrections. Analysts make adjustments
                    for changes in quality between new and old versions of a product or
                    service. They also make adjustments for differences not necessarily
                    related to quality. For example, they make adjustments to correct for
                    errors, to account for differences in size or quantity between new and old
                    versions, and to account for substitutions that, although acceptable, are
                    dissimilar to the items replaced (e.g., a pizza pan for a pie pan).

                    BLS describes commodity analysts as economists responsible for validating
                    and analyzing price data and for explaining short-run and long-run price
                    trends. Each analyst is responsible for specific item strata (e.g., the
                    commodity analyst for the pork item stratum would be responsible for
                    bacon, pork chops, ham, and other pork items, including sausage). The
                    analysts are expected to learn as much as they can about the items and the



                    Page 41                                     GAO/GGD-99-84 Consumer Price Index
                   Appendix III
                   How Commodity Analysts Decide to Make
                   Adjustments




                   consumer markets assigned to them. Drawing on that knowledge and
                   considering item-specific information supplied by price takers, commodity
                   analysts decide whether and how certain price information will be used in
                   computing each month’s CPI.

                   As part of our study of the procedures BLS follows to handle substitutions
                   and make adjustments, we collected information from 24 commodity
                   analysts about their educational and professional backgrounds and their
                   job responsibilities. The 24 analysts completed a questionnaire we gave
                   them in April 1998. The survey included all 24 individuals who had been
                   commodity analysts in October and November 1997 (the months in which
                   the substitutions we reviewed were made) and who were still commodity
                   analysts in April 1998.1

                   Nearly all of the commodity analysts (22 of 24) reported that they had
                   college degrees, with most of them reporting a bachelor’s degree and a
                   major in economics. They varied in terms of job experience from 1 to 27
                   years as a commodity analyst, 11 years was the median. As a commodity
                   analyst, they also varied from 1 year to 27 years working in the same
                   principal item strata, 9 years was the median.

                   Commodity analysts have a number of responsibilities, we were told in the
                   survey. The analysts reported that they spent the most time reviewing
                   price quotations, which include substitutions. The median amount of work
                   time they reported spending on these reviews was about 40 percent. The
                   commodity analysts generally said they research the industries in the item
                   strata for which they had responsibility and developed and modified the
                   forms that price takers use to collect price quotes. Some analysts reported
                   doing statistical analyses to develop regression models. In 1997, the
                   commodity analysts also had to prepare for the January 1998 revision of
                   the CPI.


                   The primary tool used by commodity analysts in the process for reviewing
Use of Commodity   changes in products and services is the commodity review listing (CRL).
Review Listings    After the data have been collected by the price takers and entered into BLS’
                   computer system, CRLs are generated for review by commodity analysts.
                   CRLs are computer printouts of data, such as the characteristics and price
                   history, on items meeting certain conditions. CRLs are generated when

                   1
                    These 24 commodity analysts reviewed nonrent substitutions, which are the subject of this appendix.
                   In addition to these 24 analysts, we later surveyed the 4 commodity analysts who were reviewing rent
                   data as of September 1998. These four analysts all had college degrees and had worked 2 to 14 years as
                   commodity analysts.



                   Page 42                                                    GAO/GGD-99-84 Consumer Price Index
                    Appendix III
                    How Commodity Analysts Decide to Make
                    Adjustments




                    conditions indicate a substitution (e.g., item was reported as a substitution
                    by the price taker) and for conditions having nothing to do with
                    substitutions.

                    Among the data CRLs list for an item are the item’s specifications, which
                    show various characteristics of the item, such as the packaging (e.g., 12
                    pack), container construction (e.g., metal can), and caffeine content of
                    cola drinks. If the CRL is for an item that the price taker identified as a
                    substitution, it includes the specifications for both the substitution and the
                    item it replaced. Specifications that changed from the prior price
                    collection are specially noted for the commodity analyst, and all
                    specifications are listed according to a hierarchy of importance for judging
                    the comparability of the two versions. For example, in the “canned fish or
                    seafood” entry level item (ELI), the order of importance is type (e.g.,
                    salmon), followed by variety (e.g., pink), and then form (e.g., solid),
                    thereby signaling that type and variety are more important than form of
                    the canned seafood in making comparability decisions. In some apparel
                    ELIs, the specifications are grouped by order of importance into three tiers
                    to guide the price takers in making substitutions and commodity analysts
                    in making their decisions about the substitutions.


                    According to BLS officials, the process through which commodity analysts
Process of Making   make their decisions about substitutions begins when the commodity
Decisions About     analyst receives CRLs. CRLs are generated for substitution review usually
Substitutions       for one of two conditions: (1) the item was reported as a substitution by
                    the price taker or (2) a change occurred to a key specification, although
                    the item itself was not reported by the price taker to be a substitution.

                    According to BLS, there may be a change or modification to an item being
                    priced, such as a change in color, that does not warrant substitution. BLS
                    refers to such changes as specification corrections. However, BLS
                    computers are programmed to identify changes in specific characteristics;
                    and, when any of the key specifications change, a CRL will be printed for
                    substitution review by the commodity analyst.2

                    The general process for making substitution and adjustment decisions is
                    illustrated in figure 1 in this report, and includes a series of questions that



                    2
                     Some changes, however, will not generate a CRL for substitution review. According to BLS,
                    commodity analysts have identified the specifications that when changed are least likely to cause the
                    item to be classified as a substitution.



                    Page 43                                                     GAO/GGD-99-84 Consumer Price Index
                             Appendix III
                             How Commodity Analysts Decide to Make
                             Adjustments




                             analysts answer in determining how best to deal with substitutions.3 We
                             note, however, that the process is not followed in all cases because many
                             analysts have some leeway in the process as a result of special
                             circumstances related to their respective item strata. The process
                             description that follows is based on our interviews of commodity analysts,
                             their supervisors, and BLS managers. This process has not been thoroughly
                             documented by BLS.


Determining Whether          According to BLS officials, the commodity analysts review CRLs. For the
Substitutions Were           price quotations that the price takers identified as substitutions, the
Identified                   analysts determine whether the price takers appropriately identified them
                             as substitutions. For price quotations that the price taker did not treat as
                             substitutions, the analysts determine whether these should be classified as
                             substitutions.

                             According to the BLS officials, the majority of CRLs reviewed for
                             substitutions are those identified as such by the price takers. Based on the
                             specifications for an item that the price taker reported, the commodity
                             analyst may decide that the price taker inappropriately identified an item
                             as a substitution; in that case, the CPI will treat it as if there were not a new
                             version. The price-taker-identified substitutions that the commodity
                             analysts deem appropriately identified go to the next stage of substitution
                             processing.

                             For items that have changed in characteristics but were not identified as
                             substitutions by price takers, the commodity analyst can reclassify them
                             as substitutions by “upping the version.” The commodity analyst uses the
                             changes in specifications and, to a limited extent, the price to judge
                             whether the modification in the item was a significant change. The
                             commodity analyst reclassifies an item as a substitution by making the
                             item priced in the current month a new version and making the item
                             priced in the previous collection period the old version. By doing so, the
                             commodity analyst converts the item into an acceptable substitution.


Determining Whether a        According to BLS officials, when the commodity analysts review the CRLs,
Substitution Is Acceptable   they determine if the price quotation that the price taker identified as a


                             3
                              The decisions leading to the application of the overlap method are not included in the figure or in this
                             appendix because, according to a BLS official, this method was used for a small number of
                             substitutions in 1997. BLS estimated that at least 77 substitutions were adjusted with this method in
                             1997. In addition, BLS was phasing out the use of this method of adjustment for substitutions at the
                             time of our study.



                             Page 44                                                      GAO/GGD-99-84 Consumer Price Index
    Appendix III
    How Commodity Analysts Decide to Make
    Adjustments




    substitution is eligible for use in the CPI. In 1997, commodity analysts
    determined that 28,881 price quotations that they reviewed were
    acceptable substitutions for use in the CPI. This number excludes price
    quotations for rent.

    If in the analyst’s opinion it is not an acceptable substitution, the new
    version is “killed” or deleted by the analyst. In these instances, the price
    taker in the next collection period must select another replacement for the
    item that disappeared. A BLS official said that the reasons for killing a
    substitution included when

•   the new version is out of scope (outside the definition of possible
    substitutions in the item’s ELI);
•   the substitution occurred outside of the time frame for making a
    substitution for an item that has been designated by BLS as a “seasonal”
    item, such as substituting a spring or summer raincoat for a fall or winter
    coat;
•   the commodity analyst waits to see if the old version is only temporarily
    unavailable in the outlet. The commodity analyst bases this decision on
    knowledge of the item and its price history;
•   the commodity analyst, using industry and item knowledge, believes that
    the price taker did not follow selection criteria to find the closest
    substitution to the item that disappeared; and
•   the specifications recorded by the price taker for the new version are
    unclear or incomplete.

    When a substitution is killed, it is deleted and is not used in the CPI and an
    imputed price is assigned to the item that disappeared.4 The imputation for
    a killed substitution is the same as that used in the linking method (see
    app. VI). That is, a rate of price change is calculated based on other similar
    items that were priced in the killed substitution’s item stratum and
    geographic location. The calculated rate of price change is applied to the
    previous price of the item that disappeared, and the imputed price for that
    item will be used to calculate the CPI in the next collection period.

    BLS was unable to provide us with the number of substitutions that were
    killed in 1997. However, BLS reported that over 12 months from October
    1997 through September 1998, 1,065 substitutions were killed. Since killed
    substitutions are not regarded as substitutions by BLS, they are not
    reflected in any of the tables presented in this report.

    4
     Imputed price is a term used by BLS to indicate that the actual price of the substitution is not used.
    Instead, an average is calculated from the price changes experienced that month by similar items in
    the CPI to handle a missing or unusable price quotation.



    Page 45                                                       GAO/GGD-99-84 Consumer Price Index
                             Appendix III
                             How Commodity Analysts Decide to Make
                             Adjustments




Determining the Need to      According to BLS officials, after substitutions are accepted, they go through
Make an Adjustment           additional steps to determine if adjustments are necessary to account for
                             differences between them and the items they replaced. The first
                             determinations after acceptance, we found, depend on a substitution’s
                             item stratum and the nature of any difference between the two versions. If
                             the substitution is in an automobile or apparel stratum where information
                             about the value of the differences is available to make a direct adjustment,
                             the commodity analysts are likely to consider whether to make a direct
                             adjustment. This is also the case, regardless of stratum, if the analyst is
                             correcting a past recording error or making adjustments to account for
                             changes in the size of a product or in the number of items.

                             If a direct adjustment cannot be made, then another set of processes
                             comes into play regardless of item stratum. In this case, the first question
                             usually asked is whether the two versions of the item are
                             comparable—similar.

Determining Whether a        When the analysts determine that the new version is comparable to the old
Substitution Is Comparable   version, no adjustment is made; and the ratio of the new version’s price in
                             the current period to the old version’s price in the previous period is used
                             in the calculation for the CPI for that month. This process is similar to the
                             way the CPI calculation uses the ratio of current-period price to the
                             previous-period price of items in the CPI that were not substitutions. In
                             1997, of the 28,881 nonrent substitutions reviewed by commodity analysts,
                             16,750, or 58 percent, were determined to be comparable without
                             adjustments. The rate of comparability differed among the CPI’s seven
                             components, with the highest rate in the apparel and upkeep component
                             (67 percent) and the lowest rate in the medical care component
                             (32 percent).

                             BLS uses computer codes—called comparison codes—to control the way a
                             CPI price change calculation handles a price quotation. BLS’ computers are
                             programmed to generate comparison codes to provide the commodity
                             analysts an initial basis for making their comparability decisions. Based on
                             the degree of difference between the “effective” prices of the old and the
                             new versions, one of three comparison codes appears on the CRLs. An
                             effective price is the price taker’s reported price that is adjusted, if
                             appropriate, for the size of the item, such as price per ounce of a food
                             item. If the price change, which is expressed as a percentage, is within a
                             specified interval,5 then BLS’ computers are programmed to generate a

                             5
                              The upper and lower levels of these tolerance intervals are determined by the respective commodity
                             analysts and their supervisors.



                             Page 46                                                   GAO/GGD-99-84 Consumer Price Index
                          Appendix III
                          How Commodity Analysts Decide to Make
                          Adjustments




                          code indicating that the two versions are comparable. However, if the
                          difference in effective prices is outside the specified interval, then the
                          computer generates either one of two adjustment codes (class mean or
                          linking) signaling that the versions may not be comparable. In our survey
                          of commodity analysts, how often the analysts agreed with the
                          computer-generated codes varied widely. The range of agreement was
                          from 25 percent to 100 percent of the CRLs reviewed for substitutions that
                          were made in calendar year 1997, 70 percent agreement with the codes
                          was the median.

                          In the limited number of ELIs that we studied, most commodity analysts
                          did not have written comparability criteria to guide them in making their
                          decisions. But a few food commodity analysts showed us criteria that they
                          had developed with their supervisors for specific ELIs to help them decide
                          whether substitutions are comparable. The analysts used these criteria as
                          guides to identify the characteristics (shown as specifications on the CRL)
                          that, if different between the old and new versions, indicated that the new
                          version was not comparable to the old version. For example, if a can of
                          sardines replaced a can of salmon, the analyst would determine that the
                          two versions were not comparable.

                          These comparability criteria also identified the specifications that were
                          less important in deciding comparability. That is, the old and new versions
                          might differ with regard to these characteristics, but that condition would
                          not usually warrant a decision of not comparable. For example, if the
                          change between two cans of salmon was only the origin (e.g., from
                          imported to domestic), the analyst most likely would determine that the
                          two versions were comparable.

                          Regardless of whether comparability criteria were available, all of the
                          commodity analysts we interviewed indicated that they examine the
                          specifications on the CRLs for substitutions and decide if the differences in
                          the characteristics indicate a difference in quality between the old and
                          new versions. If the commodity analysts determine that the differences do
                          not indicate a major change in quality, they said they would leave the
                          comparison code for the substitution as comparable or code the
                          substitution as comparable if a different code was assigned by the
                          computer.

Comparable Substitution   The following example of a comparable substitution decision comes from
Decision                  our interview with a commodity analyst’s supervisor. According to the
                          supervisor, this example is illustrative of an analyst’s comparability



                          Page 47                                      GAO/GGD-99-84 Consumer Price Index
                              Appendix III
                              How Commodity Analysts Decide to Make
                              Adjustments




                              decisions for this ELI and shows an analyst’s decision when it is fairly easy
                              to make.

                              The example includes a price calculation, showing the percentage change
                              in price that went into the CPI for that month as a result of no adjustment
                              being made. This percentage change in price is applicable only to the
                              particular case illustrated. It is not intended to be representative of the
                              percentage change in price that occurred for similar kinds of comparable
                              substitutions. Nor is the percentage of change intended to be
                              representative of the impact on the CPI of not making an adjustment.

                              The example includes (1) a table showing the specifications of the old
                              version and the new version that replaced it (differences in the
                              specifications between the old and new versions are highlighted by
                              shading), (2) the analyst’s reasons for judging the two versions to be
                              comparable, and (3) the calculation of price change that was made for use
                              in the CPI calculation.

Example - Club Membership     A new version of club membership was substituted for the old version and,
(Under the Club Membership    as shown in table III.1, the “level of membership” differed. The analyst
Dues and Fees Item Stratum)   concluded that the two versions were comparable even with the difference
                              in type of membership.




                              Page 48                                      GAO/GGD-99-84 Consumer Price Index
                                      Appendix III
                                      How Commodity Analysts Decide to Make
                                      Adjustments




Table III.1: Characteristics of Two
Tennis Club Memberships               Description                           Old version                   New version
                                      Specifications
                                       Type of enrollment                   Family                        Family
                                       Type of member                       Existing                      Existing
                                       Level of membership                  Tennis plus                   Gold tennis
                                       Length of period covered             Per month                     Per month
                                       Type of organization                 Health club                   Health club
                                       Bathing facilities                   Yes                           Yes
                                       Towels                               Yes                           Yes
                                       Pool                                 Yes                           Yes
                                       Exercise facilities                  Yes                           Yes
                                       Lessons                              Yes                           Yes
                                       Other clarifying data                Husband, wife, and 2          Husband, wife, and 2
                                                                            children                      children
                                      Price information
                                       Amount                               $148.00                       $150.00
                                       As of                                September                     October
                                       Collected/imputed                    Collected                     Collected

                                               Specifications that differ between old and new versions



                                      Source: BLS.


Analyst’s Reasons for                 A field message from the price taker informed the commodity analyst that
Comparable Decision                   the level of membership had not changed. To confirm the field message,
                                      the analyst reviewed the other specifications and found no changes except
                                      for a $2 increase in price. Accordingly, the analyst determined that this
                                      was a comparable substitution.

Calculation of Price Change           The calculation of price change for comparable substitutions is done
                                      entirely by computer routine without direct involvement by commodity
                                      analysts. For this item the computer calculated a 1.4 percent rate of
                                      change for October 1997, and the CPI in October reflected the same
                                      percentage increase for this tennis club membership, as it would have
                                      done if the name of the membership level had not changed.


Directly Adjusting for                According to BLS officials, the commodity analyst, after deciding that the
Differences                           substitution is acceptable and in conjunction with the comparability
                                      determination, will determine if information is available to do a direct



                                      Page 49                                                     GAO/GGD-99-84 Consumer Price Index
                               Appendix III
                               How Commodity Analysts Decide to Make
                               Adjustments




                               adjustment. If the information is available, the commodity analyst is to
                               make a direct adjustment.

                               Because the information is often available, direct adjustments are
                               commonly made for automobiles and most apparel items. BLS works with
                               automakers each year to obtain information on improvements made to
                               new vehicles and the costs of those improvements. Also, BLS has built
                               statistical models to help it value the various features of most apparel
                               items.

                               Commodity analysts also make direct adjustments to correct errors and to
                               account for changes in an item’s size or quantity. For example, a direct
                               adjustment may be made for a change in the number of events covered by
                               a season ticket. These kinds of adjustments are not limited to any one
                               stratum or group of strata, and the information to make them is usually
                               available from the CRL. BLS does not consider adjustments made only to
                               correct an error or account for a change in size or quantity to be true
                               adjustments for quality.

                               In 1997, of the 28,881 substitutions, 3,770, or 13 percent, were directly
                               adjusted by commodity analysts. As appendix VIII shows, the majority of
                               these direct adjustments were for automobile and various apparel items.
                               The adjustment amount is applied to the price of the old version. Then the
                               adjusted price of the old version is compared with the price of the new
                               version and the change, if any, expressed as a percentage, goes into
                               calculating the CPI. Appendix IV further describes direct adjustments.

Making Direct Adjustments to   Previously, we said commodity analysts compared old and new versions of
Increase Comparability         items to determine whether they were comparable. When analysts working
                               with automobile and apparel substitutions make these comparisons, they
                               do so with a different twist from those working with other substitutions.
                               The analysts we interviewed who reviewed automobile and apparel
                               substitutions said that they may consider the new and old versions of an
                               item to be essentially comparable but will still make a direct adjustment.6
                               In these cases, the analysts said, the adjustments are relatively minor and
                               are made to increase the already high level of similarity between the old
                               and new versions. For example, an apparel commodity analyst said that an
                               adjustment was made for a minor difference in the fiber content of a coat,
                               even though the difference would have been insufficient to make the old

                               6
                                BLS research on direct adjustments that were made in selected apparel item strata for 6 months in
                               1991 indicated that more than two-thirds of the substitutions would have been deemed comparable if
                               they had not been directly adjusted. See Paul R. Liegey, Jr., “Apparel Price Indexes: Effects of Hedonic
                               Adjustment,” Monthly Labor Review, Vol. 117 (May 1994), p. 40.



                               Page 50                                                      GAO/GGD-99-84 Consumer Price Index
                            Appendix III
                            How Commodity Analysts Decide to Make
                            Adjustments




                            coat and new coat not comparable. The new coat had 15 percent less wool
                            than the old coat. The analyst used a statistical model to estimate the value
                            associated with less wool and then reduced the price of the old coat by
                            that amount, making the two coats more alike.

                            According to the analysts, when they make a direct adjustment, the
                            substitution is counted with other not-comparable items. However, BLS
                            could not identify the number of times this happened because it does not
                            track whether quality-adjusted price comparisons would have otherwise
                            been comparable. Since information that is used to make adjustments to
                            increase comparability is restricted to automobile and apparel ELIs, for
                            which the commodity analysts have information to make direct
                            adjustments, only a limited number of comparable substitutions would be
                            directly adjusted for quality changes.


Using Class-Mean and        In 1997, 29 percent of the nonrent substitutions were not comparable and
Linking Methods to Adjust   could not be directly adjusted for differences between the old and new
for Differences             versions. Two indirect methods—linking and class mean—were
                            commonly used to adjust these 8,361 substitutions. In both methods the
                            price of the new version is set aside for the current month, and an
                            adjustment determined from price movements of the same type of items is
                            applied to the previous price of the old version. The class-mean method is
                            to be used in item strata where new models or product lines are
                            introduced regularly, whereas the linking method is to be used where new
                            models or product lines are not regularly introduced. In 1997, BLS had
                            designated the class-mean method to be applicable to 53 of the 206 item
                            strata. Of the substitutions that were not directly adjusted in 1997, the
                            substitutions were almost equally divided between the linking and
                            class-mean methods.7

                            As previously described in this appendix, when there is a substitution, BLS’
                            computers will indicate one of three codes on the CRL. One code indicates
                            that the computer will apply the class-mean method of adjustment if the
                            substitution is in an item stratum for which BLS has designated the
                            class-mean method. Another code indicates that the linking method will be
                            applied if the substitution is not in a class-mean designated item stratum.
                            The third code indicates that the substitution and the item it replaced are
                            comparable and, therefore, no adjustments are needed. However, the
                            commodity analysts can overrule these computer-generated codes. When

                            7
                             BLS officials also noted that substitutions are not equally distributed across all ELIs. As shown in
                            appendix VIII, some ELIs experience more substitutions than others.



                            Page 51                                                       GAO/GGD-99-84 Consumer Price Index
                              Appendix III
                              How Commodity Analysts Decide to Make
                              Adjustments




                              commodity analysts overrule those substitutions coded as comparable,
                              they are to determine if the substitution is in an item strata for which the
                              class-mean method has been designated in deciding the adjustment
                              method to apply.

                              Although the new version is set aside when the linking or class-mean
                              method is applied, it is used beginning in the following collection period.
                              The substituted product or service will become the item that the price
                              taker is to price for the CPI in future price collection periods.

Applying the Class-Mean       If the commodity analyst decides that the class mean is the appropriate
Method                        adjustment method for the substitution, then the price of the new version
                              is set aside for the current month, and a procedure is followed by BLS
                              computers to impute a rate of change. This imputation is based on a
                              particular subset of items in the CPI that BLS considers to be most
                              appropriate for the calculation. The class mean imputation is described in
                              detail in appendix V. In 1997, of the 28,881 substitutions, 4,049, or
                              14 percent, were adjusted with the class-mean method.

Applying the Linking Method   According to BLS, if the substitution that the commodity analyst deemed as
                              not comparable resides in an item stratum that is not a class-mean
                              designated item stratum and was not directly adjusted for quality change,
                              then the linking method will be applied. In the linking method, the price of
                              the new version is set aside, and a procedure is followed by BLS computers
                              to impute a rate of change that is based on all the other items with usable
                              prices in the geographic area and item stratum. This procedure is further
                              described in appendix IV.

                              Some of the commodity analysts who work on class-mean designated item
                              strata told us that under certain circumstances, they use the linking
                              method for substitutions that are in class-mean designated item strata. For
                              example, the analyst for new automobiles reported that he used the
                              linking method when the price taker appeared to have collected incorrect
                              information.

                              Of the 28,881 substitutions in 1997 that commodity analysts reviewed,
                              4,312 (about 15 percent) were adjusted with the linking method. Of these
                              linked-adjusted substitutions, 371 were in class-mean item strata.




                              Page 52                                      GAO/GGD-99-84 Consumer Price Index
                            Appendix III
                            How Commodity Analysts Decide to Make
                            Adjustments




                            BLS relies on supervisors to review the substitution decisions made by
BLS Processes and           commodity analysts. Although there are no written guidelines for
Procedures for              supervisors to follow in selecting and reviewing the decisions made by
Reviewing                   analysts, BLS officials stated that the supervisors follow unwritten policies.
                            There is no unique set of procedures for selecting and reviewing these
Substitution Decisions      decisions. The reason for most supervisory reviews, according to BLS, is a
                            large price change, with the definition of large varying among and within
                            item strata. Supervisory reviews are scheduled three times during the
                            month’s pricing periods and after all of the pricing information has been
                            collected. For those substitutions selected for review, supervisors
                            normally rely heavily on commodity analysts in carrying out the scope of
                            their reviews.


Supervisors Review          According to BLS, supervisors are to review large price changes going into
Substitutions That Result   the CPI each month, with the definition of large varying among and within
in Large Price Changes      item strata. Large price changes are reviewed because they are the price
                            changes that could have the greatest impact on the indexes. As part of this
                            review, substitution decisions made by commodity analysts are reviewed
                            when those decisions produce large price changes. Substitution decisions
                            that produce price changes that are too small to meet the selection criteria
                            are usually not subject to review. The branch chief for consumer prices
                            reported that there is no random review of analysts’ decisions. According
                            to BLS, it does not track the percentage of substitution decisions that are
                            reviewed because of large price changes.

                            A BLS official reported that supervisors and analysts working together set
                            the levels of price change that trigger a review. Levels are set for each
                            individual item stratum. To establish these levels, the analyst and
                            supervisor responsible for an item stratum consider that stratum’s current
                            and historical price data but also rely on their own judgment and industry
                            knowledge. As a result, the levels of price change that require a
                            supervisory review can vary by item strata and can differ between price
                            increases and price decreases within the same stratum. For example, the
                            supervisor for the household-goods item stratum reviews price increases
                            or decreases of 15 percent or more. However, in the women’s coats and
                            jackets item stratum, price increases of 25 percent or more are to be




                            Page 53                                      GAO/GGD-99-84 Consumer Price Index
                           Appendix III
                           How Commodity Analysts Decide to Make
                           Adjustments




                           reviewed, while price decreases require a change of 20 percent or more.
                           Price changes that do not meet or exceed these levels are not reviewed.8


Supervisory Reviews        For the purpose of collecting CPI prices, BLS divides each month into three
Occur After Data           collection periods, which it calls pricing periods. At the end of each
Collection Periods End     pricing period, BLS officials stated, all the price quotations in which a price
                           change meets or exceeds the trigger levels for supervisory review are
                           identified by computer routines, including some that resulted from the
                           substitution decisions of commodity analysts. At the end of each pricing
                           period, computer routines produce lists of these quotations, called
                           supervisory query review lists, that are organized by each commodity
                           analyst, item stratum, and geographic region. These lists, which are
                           produced at the end of each pricing period, provide basic information for
                           each price quotation that was flagged for review. This information
                           includes the old and new prices of the items, the percentage of price
                           change, and any short messages the analysts provided to explain the
                           change.9 The lists can be lengthy; one chief of a section said that the three
                           supervisor query lists have generated a total of more than 1,000 price
                           quotations in some months.

                           The Branch Chief for Consumer Prices reported that BLS does not have any
                           written policies or guidance for supervisors to follow when reviewing
                           analysts’ decisions. However, the branch chief stated that the unwritten
                           policy requires supervisors to review and approve every price change that
                           appears on the supervisory query lists. One section chief said that, in
                           general, each of the three reviews was completed in about a day.


Review Also Occurs After   After the supervisory review of price changes is complete, BLS computer
Preliminary Indexes Are    routines produce preliminary basic indexes for each item stratum and
Produced                   geographic area. The supervisors are required to review these preliminary
                           indexes. According to BLS, review of these preliminary basic indexes is an
                           additional quality control that has been built into the CPI to identify and
                           verify large price changes. After any changes that result from the review of
                           the preliminary basic indexes are entered into the computer, the final

                           8
                            These levels were for “regular” prices of women’s coats and jackets (i.e., the prices of items that were
                           not on sale). BLS normally set different levels of price change for regular and sale prices. For example,
                           substitutions for women’s coats and jackets at regular prices were reviewed if the new version had
                           decreased by 20 percent. However, substitutions for women’s coats and jackets at sale prices were
                           reviewed only if the new version decreased by 75 percent or more.
                           9
                            The lists do not routinely include information about whether the price quotation was a substitution
                           and, if so, what decision the analyst made. However, this information might be included in the
                           analysts’ explanatory messages.



                           Page 54                                                      GAO/GGD-99-84 Consumer Price Index
Appendix III
How Commodity Analysts Decide to Make
Adjustments




basic indexes are calculated and combined to form the CPI indexes that are
published.

More than 8,000 preliminary basic indexes were produced each month in
1997, according to BLS officials.10 The production schedule was supposed
to allow supervisors about 2 days each month to review all preliminary
basic indexes.11 However, because of production delays, supervisors often
had much less time to review the indexes, BLS officials said.

The Branch Chief of Consumer Prices stated that the supervisors are to
identify large price changes in the preliminary indexes, and investigate the
price quotations responsible for those changes. According to BLS officials,
there are no set levels of price changes that require investigations. Instead
the supervisors are to decide what levels require investigation. By way of
illustration, one section chief stated that the rates of price change that
would prompt him to investigate an index varied by item stratum, and
could change over time, but generally ranged between 12 to 20 percent at
the time of our interview.12

During an investigation, the supervisors typically are to identify the
quotations that are causing increases from a list containing all the price
quotations for the month.13 This list, which is called the Commodity
Analyst Listing (CAL), contains basic information for each quotation, and
includes the old and new prices, the percentage price change, the analyst’s
decision, and any short explanatory messages provided by the analyst.

According to the Branch Chief for Consumer Prices, when the supervisors
identify quotations that cause large changes in a preliminary basic index,
they can ask the commodity analyst responsible to explain what had
occurred. However, it is possible that, because of the three supervisory
query reviews that occurred earlier in the month, quotations with large
price changes may have already been examined.

10
 The preliminary basic indexes were produced for each of BLS’ 183 priced item strata in each of its 44
geographic regions. Following the 1998 revision of the CPI, the number of preliminary basic indexes
diminished because the number of geographic regions decreased to 38.
11
  The chiefs of the Consumer Price Branch’s four Commodities and Services sections were responsible
for the supervisory review. However, at the time of our field work, two of the chiefs also had a
supervisor in their sections who assisted them in the review.
12
  The levels of price change that prompt supervisors to review an index can be different from the
levels of price change that require supervisors to examine a price quotation during the earlier
supervisory query review.
13
 About 80,000 price quotations were collected each month in 1997. The section chiefs of the
Consumer Price Branch received information for all the price quotations in their sections.



Page 55                                                     GAO/GGD-99-84 Consumer Price Index
                         Appendix III
                         How Commodity Analysts Decide to Make
                         Adjustments




Normal Scope of          The Chief of the Consumer Prices Branch stated that the supervisors are
Supervisory Review       required to examine the reasons for the analysts’ decisions for price
                         quotations that produce large price changes. The supervisors reported that
                         they frequently did this by examining the short messages that commodity
                         analysts attached to each price quote and that are printed on the
                         supervisory query and CALs. These messages were short, in large part,
                         because the computer system limits messages to 128 characters. We found
                         the following examples of short messages for price quotations that were
                         substitutions for apparel items in October and November 1997:

                     •   “Approximately same specs;”
                     •   “Price change O.K.;”
                     •   “Same regular price;” and
                     •   “Change in fiber.”

                         If the supervisors are not satisfied with explanations in the short
                         messages, they are to ask the analysts to explain their decisions in person.
                         All of the supervisors said that they usually accept the analysts’
                         explanations, either through the short messages or in person. Three of the
                         four supervisors stated that it was very rare for them to formally review an
                         analyst’s decision by examining the characteristics and price histories of
                         the quotations themselves. One supervisor stated that he did not review
                         the characteristics and price histories because of the confidence he had in
                         his commodity analysts.




                         Page 56                                     GAO/GGD-99-84 Consumer Price Index
Appendix IV

Direct Adjustments


                           A direct adjustment is a price adjustment for specific differences between
                           the characteristic of a good or service that the CPI priced in one collection
                           period and the characteristics of a substitute good or service that the CPI
                           priced in the next collection period. Direct adjustments can be classified
                           on the basis of BLS’ use into the following:

                       •   a manufacturers’ cost method, which uses cost information from
                           manufacturers to identify the cost of individual characteristics or options
                           that have changed;
                       •   a hedonic regression method, which uses statistical models to estimate a
                           value for (or implicit “price” of) the individual characteristics of a good,
                           such as the cost of adding a lining and hood to a coat; and
                       •   an “other” method, which covers a broad array of direct price adjustments,
                           including those made for a change in the number of units (e.g., 15 tablets
                           instead of 12 tablets) or the size of an item (e.g., 8 ounces instead of 10
                           ounces), and to correct data-entry errors made by BLS.

                           Similar to the indirect adjustment methods of linking and class mean that
                           BLS uses, the direct adjustment methods divide the total price change into
                           a pure price change component and a quality change component. The CPI
                           is to incorporate only the pure price component.

                           A fundamental difference between the indirect and the direct adjustment
                           methods is that the indirect methods first estimate the pure price change
                           and then assign any remaining or residual portion of the price difference
                           to quality. The direct methods do the opposite. They first estimate the
                           value of quality changes and, in essence, remove that value from the price
                           difference. The pure price change is the residual after the adjustment is
                           made. Hence, unlike the appendixes on the class mean and linking
                           methods, this appendix first describes how quality changes are estimated
                           before discussing how price changes are calculated as a residual.


                           BLS uses manufacturers’ cost information to make direct adjustments for
Manufacturers’ Costs       changes in quality in the new and used vehicles item strata. Each year, BLS
                           collects information from the manufacturers of automobiles and trucks on
                           the costs of the new features in each model and uses this information to
                           make adjustments. For used vehicles, BLS uses data from prior years’ new
                           vehicles.

                           All of BLS’ manufacturers’ cost adjustments were in the new and used
                           vehicle item strata. In 1997, BLS made 1,828 direct adjustments in these



                           Page 57                                      GAO/GGD-99-84 Consumer Price Index
                       Appendix IV
                       Direct Adjustments




                       items. According to BLS officials, about 90 percent of these adjustments
                       were made using manufacturers’ cost information, and the remainder were
                       made using the other method.


Background on          Since 1967, BLS has asked the major domestic automobile and truck
Manufacturers’ Costs   manufacturers for information about the cost of quality improvements in
                       the new models that are used for the Producer Price Index (PPI).1 BLS
                       requests all manufacturers of domestically produced vehicles to
                       participate in this process. Domestically produced vehicles include those
                       produced in the United States by both U.S. and foreign automobile
                       manufacturers. The manufacturers are very cooperative, according to BLS.
                       Manufacturers that produce their models overseas and import them into
                       the United States are not included in this annual information gathering
                       process.

                       For the 1998 model year (introduced in the fall of 1997), BLS requested
                       information on the 20 car models it had selected for pricing in the PPI.
                       Manufacturers provided information on only 15 models in time for the BLS
                       press release on 1998 models published in October 1997. According to BLS,
                       the five models lacking information either had so many changes that the
                       manufacturer could not accurately break out the cost associated with each
                       change, or the information, while usable for direct adjustments when
                       reviewing substitutions, was provided too late to be incorporated in the
                       press release.

                       The manufacturers provide BLS with lists of improvements they have made
                       to the new models in the PPI sample and the cost of producing those
                       improvements. According to BLS, the manufacturers often claim that
                       improvements in style, comfort, and convenience are quality
                       improvements. However, BLS does not consider such improvements to be
                       quality improvements unless they are presented with evidence of
                       functional improvements. BLS has developed criteria for quality
                       improvements that include improvements in safety, durability, and
                       performance. BLS analysts use these criteria to review the manufacturers’


                       1
                        BLS bases the PPI on producers’ output. According to BLS the PPI sample of new model vehicles was
                       representative in the sense that it included the platforms that accounted for the automakers’ largest
                       outputs. The PPI sample selection process focuses on vehicle platforms, officials stated, because a
                       single platform usually includes a number of nameplates, and each nameplate may have numerous
                       models. For example, the H platform for General Motors has three nameplates: the Buick LeSabre,
                       Oldsmobile Eighty-Eight, and Pontiac Bonneville, each of which has its own models. All of the models
                       are built on the same platform and have basically the same power trains. The differences between the
                       models include styling and equipment changes. According to BLS, it generally uses one model to
                       represent an entire platform because of the similarity between the models.



                       Page 58                                                    GAO/GGD-99-84 Consumer Price Index
                          Appendix IV
                          Direct Adjustments




                          lists of improvements and decide whether they are, in fact, quality
                          improvements. Through their review, BLS analysts developed lists detailing
                          every quality improvement and the cost of every quality improvement for
                          all models in the model-year sample.2

                          For the 15 new domestic passenger cars in the 1998 PPI sample, according
                          to BLS, the average increase in the manufacturers’ suggested retail price
                          over the 1997 versions of these models was $363.27. Of this increase, BLS
                          attributed $230.81 (or 63.5 percent) to quality changes. BLS estimated that
                          $52.14 of the $230.81 was for changes in accordance with the 1990 Clean
                          Air Act Amendments and $178.67 was for other quality changes, such as
                          powertrain improvements, corrosion protection upgrades, and changes in
                          the levels of standard or optional equipment.3

                          The CPI program applies this information to models that were in the PPI
                          sample and to similar models that were outside the PPI sample. CPI
                          commodity analysts use information from the manufacturers and publicly
                          available publications, such as Automotive News, to help determine
                          whether models outside the PPI sample shared the same platform as
                          models inside the PPI sample.

                          In some cases, BLS analysts apply manufacturers’ information from the
                          sample to models outside the sample that do not share the same platform
                          with a model in the PPI sample. For example, if a rear defogger was added
                          to a new model outside the sample, the analyst might estimate the average
                          price of defoggers added to models inside the sample and use the estimate
                          to make an adjustment to the model outside of the sample, even though it
                          did not have the same platform. Sample information might also be used to
                          make adjustments to foreign manufacturers’ models imported into the
                          United States, in instances where the domestic manufacturer’s information
                          appeared applicable. In cases such as the ones just described, BLS analysts
                          rely on judgment rather than a set of uniform rules and procedures.


How Quality Changes Are   The CPI’s commodity analyst for new cars uses the lists of quality changes
Calculated Using          when reviewing substitutions. If the substitution (a 1998 model) is in the
                          PPI sample, the analyst looks for improvements that have been made in the
Manufacturers’ Costs
                          model under review. The analyst may find, for example, that the engine of


                          2
                           According to BLS, officials from the CPI and the International Price Program also participate in the
                          review process that determines whether improvements are, in fact, quality improvements.
                          3
                           As of January 1999, BLS no longer treats changes made solely to meet air quality standards as quality
                          improvements in the CPI.



                          Page 59                                                     GAO/GGD-99-84 Consumer Price Index
Appendix IV
Direct Adjustments




a particular model has been upgraded. The manufacturer’s information in
the sample also lists the costs of producing the improvement. For
example, the engine upgrade might cost $100. The analyst will then
multiply this cost by a factor to adjust for the difference between producer
and retail prices. For example, if the analyst used 20 percent as a factor,
the result would be $120 ($100 x 1.20 = $120).

Using this methodology, an adjustment of $120 would be made to ensure
that the new (1998) and old year’s (1997) models were comparable in
terms of quality. If the price of the new model was $11,750 and the price of
the old model was $11,500, BLS would add $120 to the price of the old
model so that the comparison would be between $11,750 and $11,620
instead of between $11,750 and $11,500.

According to BLS, if a model priced in the CPI sample is not in the PPI
sample, the analyst is to use the manufacturers’ and publicly available
information to determine whether the price is similar to one of the models
in the sample. For example, the available information may show that a
Ford Escort is built on the same platform as a Mercury Tracer. If the Ford
Escort was in the sample, but the Mercury Tracer was not, and the analyst
concludes that the two models are sufficiently similar, the manufacturer’s
information on the Ford Escort could be applied to the Mercury Tracer.

If a model is not in the PPI sample and is not built on the same platform as
a model in the PPI sample, the analyst may still be able to use some
manufacturer’s information obtained in the sample. However, in the
absence of manufacturer’s information, it is much more likely, according
to BLS, that the analyst will use the class-mean method of adjustment.4 In
general, prices of new foreign models manufactured abroad and imported
to the United States are more likely than domestically produced items to
be adjusted using the class-mean method.

Direct adjustments for used vehicles rely on manufacturers’ information
that was obtained in previous years. The CPI used-car index is based on the
prices of vehicles that are between 2 and 6 years old. For example, the
used-car index in 1998 might include the substitution of a 1995 for a 1994
Chevrolet Cavalier. In that instance, the BLS analyst would make
adjustments if the manufacturers’ information for the 1995 and 1994


4
 BLS officials report that, if a new model has been significantly redesigned, they may consider it too
different to be compared to the old model. Under those circumstances, BLS will not use the
manufacturers’ information—if it is available—to make direct adjustments, but will use the class-mean
method of adjustment.



Page 60                                                    GAO/GGD-99-84 Consumer Price Index
                             Appendix IV
                             Direct Adjustments




                             Chevrolet Cavaliers indicated that quality changes had occurred between
                             the model years.

                             A direct adjustment for used cars relies on the information collected in
                             earlier years for new cars, with the dollar amount of the adjustment
                             translated into percentage terms. In the above example, the estimated $120
                             quality improvement equals 1.0 percent of the $11,500 price for the model
                             of prior year (1997). If BLS uses the two models in the price index 3 years in
                             the future (2001), a 1.0 percent quality adjustment would be added to the
                             used-car price collected for the 1997 model before it is compared with the
                             used-car price collected for the 1998 version.


How Price Changes Are        In the example above, an adjustment of $120 would be added to the price
Calculated After Making an   of the previous year’s model to make the new and old year’s models
Adjustment Using             comparable with the terms of quality. Then the adjusted price of the old
                             model ($11,620) would be compared to the price of the new model
Manufacturers’ Costs         ($11,750). The CPI would incorporate an increase of 1.1 percent to reflect
                             the $130 difference between the two prices. Of the $250 total price
                             difference between the two models in the example, BLS procedures
                             allocate $120 to quality and $130 to price.


BLS’ Use of Manufacturers’   The following example results from our interviews with a BLS commodity
Costs                        analyst. According to the commodity analyst and supervisor, it is
                             representative of the adjustments that employ manufacturers’ costs to
                             adjust for quality improvements in vehicles.

                             The example includes a price calculation that shows the percentage
                             change in price that went into the CPI for that month, as a result of the
                             direct adjustment. This percentage change in price is applicable only to
                             the particular case illustrated. It is not intended to be representative of the
                             percentage changes in price that occurred for similar kinds of
                             substitutions in which direct adjustments are made. Nor is the percentage
                             change in price intended to be representative of the impact that direct
                             adjustments have on the CPI.

                             The example includes (1) a table showing the characteristics of the old
                             version and the new version that replaced it (differences between the old
                             and new versions are highlighted by shading), (2) the analyst’s reasons for
                             judging the two versions to be comparable, (3) the rationale for selecting
                             the method of adjustment, and (4) the price calculation that was made.



                             Page 61                                       GAO/GGD-99-84 Consumer Price Index
                                     Appendix IV
                                     Direct Adjustments




                                     Although the specifications for the 1998 and 1997 models were alike, the
Example 1 - New Car                  commodity analyst made an adjustment because the analyst had been
(Under the New Cars                  provided with information by the automaker that showed improvements
Item Stratum)                        had been made to the 1998 model.


Table IV.1: Characteristics of Two
Versions of a New Car                Description                        Old version                     New version
                                     Specifications
                                      Model year                        1997                            1998
                                      Model                             4-door sedan                    4-door sedan
                                      Engine size                       3.8 liter engine                3.8 liter engine
                                      Number of cylinders               Six cylinders                   Six cylinders
                                      Transmission                      4-speed automatic               4-speed automatic
                                      AM/FM radio and cassette          Yes                             Yes
                                      Rear window defroster             Yes                             Yes
                                      Power-assisted driver seat        Yes                             Yes
                                     Price information
                                      Amount                            $22,104                         $23,180
                                      As of                             September                       October
                                      Collected/imputed                 Collected                       Collected

                                              Specifications that differ between old and new versions



                                     Source: BLS.




Analyst’s Reasons for                Even though the specifications of the old model and the new model were
Deciding the Items Were              unchanged, the analyst knew that changes had been made to the new
Comparable                           model because of information provided by the automaker. This
                                     information showed that the engine had been improved to reduce fuel
                                     emissions and that the air bags had been depowered to improve passenger
                                     safety. In the analyst’s opinion, despite the improvements, the new and old
                                     models were broadly comparable. Nevertheless, the analyst concluded
                                     that an adjustment for the engine improvements and depowered airbags
                                     would make them even more comparable.


Rationale for Selecting the          To make the adjustment, the analyst had to override another adjustment
Method of Adjustment                 code that was computer generated. A computer routine automatically
                                     calculates any change in prices between the old and new versions of a



                                     Page 62                                                     GAO/GGD-99-84 Consumer Price Index
                       Appendix IV
                       Direct Adjustments




                       substitution. If the price change is below a predetermined level, the
                       computer is programmed to default to a comparable substitution code. If
                       the price change exceeds the predetermined level, the computer defaults
                       to a not-comparable code. In the new cars item stratum, the
                       not-comparable code would indicate that the class-mean method should
                       be used. The computer routine cannot default to a direct adjustment code.
                       In this case, the computer generated the code for a comparable
                       substitution.

                       If the analyst had not made the adjustment the models would have been
                       considered comparable, and the 4.9 percent difference between the
                       collected price of the new model and that of the old model would have
                       been incorporated into the CPI. However, the analyst had the necessary
                       data from the sample of manufacturers to make an adjustment and
                       believed that an adjustment would make the new and the old models more
                       comparable in terms of quality.


Calculation of Price   The automaker had reported that the wholesale cost of the engine
Change                 improvements was $105, and the cost of depowering the airbags was $10.
                       The analyst added these two costs together ($105 + $10 = $115), and then
                       multiplied the total by 118 percent to account for the difference between
                       the wholesale costs and their retail equivalent. The analyst used a markup
                       factor of 118 percent because, according to BLS, the available data
                       indicated this was the most appropriate markup for this type of vehicle in
                       1998. The cost total of $115 multiplied by 118 percent was $135.70, which
                       the analyst rounded to $135.

                       To make the adjustment, the analyst added the $135 to the price of the
                       1997 model ($22,104 + $135), creating an adjusted price of $22,239. As a
                       result of this action, BLS’ computer routines compared the adjusted price
                       of $22,239 for the old model with the actual price of the new model,
                       $23,180, and recorded a price increase of 4.3 percent for the new model.
                       This price increase was included in the CPI calculations for October 1997.


                       BLS first introduced the use of adjustments based on hedonic regression
Hedonic Regression     estimates into the CPI in 1988 to adjust for the effect of aging on housing
Estimates              units. In 1991, hedonic regression estimates were applied to adjust apparel
                       items. According to BLS, the extensive and frequent seasonal and
                       style-driven changes in characteristics of apparel items pose numerous
                       problems for the maintenance of a constant quality market basket of



                       Page 63                                     GAO/GGD-99-84 Consumer Price Index
                           Appendix IV
                           Direct Adjustments




                           apparel items in the CPI. The use of hedonic regression methods in apparel,
                           BLS reported, has helped address these problems.


                           In 1997, BLS used hedonic regression estimates exclusively for apparel
                           items. According to BLS officials, at least 95 percent of the 1,223 direct
                           adjustments that were made for apparel items in 1997 used hedonic
                           regression estimates; the remainder used the other method.

                           BLS is undertaking a major research program with the goal of applying
                           adjustments based on hedonic regression estimates to other consumer
                           items where satisfactory models can be developed. The effort will place
                           particular emphasis on consumer durable goods, such as electronic
                           products, likely to experience significant quality change over time. In
                           January 1998, hedonic regression models that were initially developed for
                           the PPI began to be used for computer items in the CPI. Effective with the
                           release of the CPI for January 1999, BLS introduced hedonic adjustments in
                           the television stratum of the CPI.


Background on Hedonic      Hedonic regression estimates are used to make direct adjustments to the
Regression Estimates for   price of an item when it is replaced by a substitute item so that BLS can
Adjustments                compare its price with the price of its replacement. The values, or the
                           implicit prices, of the various quality characteristics that affect the total
                           price of a particular good are estimated using regression techniques.

                           A fundamental assumption underlying the hedonic price framework,
                           according to economists, is that an individual good can be viewed as a
                           combination of a number of observable characteristics.5 The hedonic
                           modeling technique—based on the prices paid by consumers for the goods
                           for which the CPI collects prices—provides an estimate of the value or
                           implicit price of each characteristic of the good that is modeled in the
                           hedonic regression equation.

                           The implicit prices for individual product characteristics that are
                           estimated using hedonic regressions can be used to adjust the prices of
                           two items with different characteristics. After these adjustments, the
                           remaining difference between the prices of the two items can be
                           considered as an estimate of the pure price difference between them. This
                           pure price change can then be incorporated in a constant-quality price
                           index such as the CPI.

                           5
                           See, for example, Triplett, Jack E., “The Economic Interpretation of Hedonic Methods,” Survey of
                           Current Business, January 1986, p. 40, and Berndt, Ernst R., The Practice of Econometrics: Classic and
                           Contemporary, Addison-Wesley, New York, 1991, p. 117.



                           Page 64                                                    GAO/GGD-99-84 Consumer Price Index
                                        Appendix IV
                                        Direct Adjustments




How Quality Changes Are                 For the period reviewed in our report (October and November 1997), the
Calculated Using Hedonic                application of hedonic techniques among nonrent items was limited to
Regression Models                       apparel items. BLS has found that measuring price change for goods such
                                        as apparel items presents considerable problems because these items
                                        undergo numerous changes in characteristics as fashions evolve.6 Since
                                        January 1991, BLS has been using hedonic regression models to make
                                        direct adjustments for quality differences in apparel items. According to
                                        BLS, the hedonic regression models develop estimates for the values of
                                        product features. These estimates are then used to make adjustments to
                                        the prices of apparel items for the changes in quality. Previously, BLS relied
                                        heavily on the linking method to adjust apparel items.

                                        Hedonic regression modeling in apparel utilizes the general framework in
                                        which an item can be viewed as a collection of characteristics, which,
                                        taken together, provide satisfaction or value to the consumer. A woman’s
                                        jacket, for example, can be considered an aggregation of its features, such
                                        as its fiber content (e.g., percentage wool) and type of closure (button or
                                        zipper), each of which contributes to the value of the jacket in the eyes of
                                        the consumer.

                                        Figure IV.1 describes the hedonic regression equation that BLS uses for
                                        apparel items.

Figure IV.1 Description of a Hedonic
Regression Equation for Apparel Items
                                        The standard hedonic regression equation for apparel commodities uses the natural
                                        logarithm of the item’s price as the dependent variable and several independent
                                        variables capturing different characteristics of the apparel item. The values of the
                                        independent variables are measured linearly (i.e., measured in levels, rather than
                                        logarithms). The coefficients in such a semilog specification provide an estimate of the
                                        proportional change in price that results from a one-unit change in a quality
                                        characteristic. In most cases, an independent variable representing a characteristic is
                                        dichotomous, in the sense that it indicates whether or not the item possesses the
                                        characteristic. The value of the variable is 1 for an item with the characteristic and 0 for
                                        an item without it. In some cases, an independent variable measures a continuous
                                        quantity, such as the content of a particular fiber. An item’s percentage of wool, for
                                        example, would range from 0 to 100. Additional variables called control variables usually
                                        also are included to capture the effect of price variations by city size, region, and type of
                                        business.
                                        Source: BLS.



                                        Determining the best set of characteristics to explain prices in apparel
                                        poses a challenge because fashion influences price. BLS analysts found,

                                        6
                                         BLS Handbook of Methods, U.S. Department of Labor, BLS, (April 1997) p. 184.



                                        Page 65                                                   GAO/GGD-99-84 Consumer Price Index
                       Appendix IV
                       Direct Adjustments




                       however, that fashion can be approximated to some extent through such
                       characteristics as whether the item is a store or national brand and the
                       type of closure (e.g., zipper or buttons). According to BLS, other
                       characteristics such as lining and fiber are always included in the apparel
                       regression models because they are fundamental to the price of an apparel
                       item. According to BLS, all these characteristics can be observed and tested
                       to see the degree of influence, if any, they exert on price.

                       For example, a 1997 version of the hedonic regression model for women’s
                       coats provides estimates of the effect on price of almost 30 different
                       characteristic and control variables. These estimates included
                       dichotomous variables indicating whether or not the item was one of
                       several types of coats, such as a windbreaker, a parka, a trench coat,
                       among others. Continuous variables included the percentage content of
                       cashmere, camel hair, wool, cotton, and other fibers. The equation
                       included variables to control for other factors, such as whether the price
                       was collected at a discount department store, in a large city, and from the
                       Northeast.

                       According to BLS guidelines, the value to consumers of particular product
                       attributes change over time. Hence, the hedonic regression estimates need
                       to be updated periodically. For example, BLS’ detailed written guidelines
                       for the apparel modeling process include a requirement that the apparel
                       regression models be updated at least every 12 to 15 months. Should a
                       sharp market change occur, however, BLS’ guidelines call for earlier
                       updating.


How BLS Accounts for   BLS’ exclusion of quality differences using the hedonic regression estimates
Quality Change Using   can be demonstrated with a hypothetical case based on a woman’s trench
Hedonic Regression     coat. For example, assume that the new version of the trench coat had a
                       lining, but the old version did not, and the two versions had no other
Estimates              differences. In this example, assume that the BLS commodity analyst
                       determined that the two versions were not comparable but was able to use
                       the estimated hedonic regression equation for women’s coats to adjust for
                       the effect of the difference.

                       For this illustration, assume that the price of the old trench coat was $100
                       and that the price of the new version was $125. The estimated hedonic
                       regression coefficient for the lining characteristic was .13470725, which
                       represents the lining’s logarithmic effect on the price of the coat. To
                       determine the quality adjusted price of the old version, the logarithmic



                       Page 66                                      GAO/GGD-99-84 Consumer Price Index
                             Appendix IV
                             Direct Adjustments




                             price effect is exponentiated and multiplied by the old price: $100 x
                             exp(.13470725) = $114.42.

                             The hedonic quality adjustment has the effect of adding $14.42 ($114.42 -
                             $100.00) to the price of the old item to make it comparable to the new
                             item. In effect, the hedonic regression adjustment implies that if the old
                             version had a lining, the lining would have added $14.42 to its price.


How BLS Calculates a         Following the hedonic regression adjustment described above, the old
Price Change After Making    item’s adjusted price of $114.42 is compared with the new item’s price of
a Hedonic                    $125.00. The 9.2 percent difference between the two prices is incorporated
                             in the calculation of the CPI as a pure price increase. Thus, of the $25
Regression-Based             actual price difference between the two versions of the coat, $14.42 was
Adjustment                   allocated to a quality difference; and the remaining $10.58 was determined
                             to represent a pure price increase.


Example of BLS’ Use of the   The following example results from our interviews with a BLS commodity
Hedonic Method               analyst. According to the commodity analyst and supervisor, it is
                             representative of the methods in which they use hedonic regression
                             estimates to adjust for quality change in apparel items.

                             The example includes a price calculation that shows the percentage
                             change in price that went into the CPI for that month as a result of the
                             direct adjustment. This percentage change in price is applicable only to
                             the particular case illustrated. It is not intended to be representative of the
                             percentage changes in price that occurred for similar kinds of
                             substitutions in which direct adjustments are made. Nor is the percentage
                             change in price intended to be representative of the impact that direct
                             adjustments have on the CPI.

                             The example includes (1) a table showing the characteristics of the old
                             item and the new item that replaced it (differences between the old and
                             new versions are highlighted by shading), (2) the analyst’s reasons for
                             judging the two versions to be not comparable, (3) the rationale for
                             selecting the method of adjustment, and (4) the price calculation that was
                             made.




                             Page 67                                       GAO/GGD-99-84 Consumer Price Index
                                     Appendix IV
                                     Direct Adjustments




                                     A new version of heavyweight woman’s coat was substituted for the old
Example 2 - Woman’s                  version and, as shown in table IV.2, several important characteristics of
Coat (Under the                      the two versions differed. The analyst decided that the two versions were
Women’s Coats and                    not comparable, and made a direct adjustment by using values from a
                                     hedonic regression.
Jackets Item Stratum)
Table IV.2: Characteristics of Two
Versions of a Woman’s Coat           Description                           Old version                     New version
                                     Specifications
                                         Type                              Heavyweight                     Heavyweight
                                         Design                            Shirt jacket/Battle jacket      Parka
                                         Body fiber                        100 percent cotton              60 percent cotton and 40
                                                                                                           percent polyester
                                         Brand/label name                  Local brand name                Different local brand name
                                         Size range                        Junior/misses                   Junior/misses
                                         Length                            Waist                           Waist
                                         Cleaning method                   Machine washable                Dry clean only
                                                                                                           a
                                         Type of pockets                   2 slot
                                         Lining                            Yes                             Yes
                                         Method of closure                 Button closure                  Zipper and button closure
                                         Color                             Biscuit (Tan)                   Ivory/Hunter green
                                      Other information                    Corduroy                        Drawstring waist
                                     Price information
                                         Amount                            $32.028                         $125.00
                                         Type of price                     Sale                            Regular
                                         As of                             August                          October
                                         Collected/imputed                 Imputed                         Collected

                                                 Specifications that differ between old and new versions


                                     a
                                     There was no information about the type of pockets in the new version.


                                     Source: BLS.




Analyst’s Reasons for                The commodity analyst noted that there were several major differences
Deciding the Items Were              between the old and new versions in the design, body fiber, and method of
Not Comparable                       cleaning. The analyst said that BLS research had shown that these
                                     differences had a significant effect on the price of women’s coats and




                                     Page 68                                                        GAO/GGD-99-84 Consumer Price Index
                              Appendix IV
                              Direct Adjustments




                              jackets. The differences were sufficient for the analyst to determine that
                              the items were not comparable. The analyst used hedonic regression
                              estimates to adjust for the quality differences and make the items
                              comparable. Without the adjustments the class-mean method would have
                              been applied to the item by default.


Rationale for Selecting the   To make the adjustment, the analyst had to override another adjustment
Method of Adjustment          code that was computer generated. A computer routine automatically
                              calculates any change in prices between the old and new versions of a
                              substitution. If the price change percentage is within a predetermined
                              interval, the computer is programmed to default to a comparable
                              substitution code. If the price change is outside the interval, the computer
                              defaults to a not-comparable code. In the women’s coats and jackets item
                              stratum, the not-comparable code would indicate that the class mean
                              should be used. Because the computer routine is based on price change, it
                              cannot default to a direct adjustment code. In this case, the
                              computer-generated the code for a not-comparable substitution, which
                              means that the class-mean method of adjustment would have been used.

                              However, the hedonic regression model for women’s coats and jackets
                              provides estimates of the price effects of differences in design, body fiber,
                              and method of cleaning. Therefore, the analyst was able to apply the
                              hedonic regression estimates to adjust for quality differences and make
                              the substitution comparable.


Calculation of Price          The analyst used the results of the hedonic regression to calculate the
Change                        adjustment attributable to the quality change in design, body fiber, and
                              method of cleaning. The analyst calculated this to be $33.850 and added
                              this value to the price of the old version to derive an adjusted price of
                              $65.88 ($32.028 + $33.850).7 The difference between the adjusted
                              price of the old version ($65.88) and the price of the new version
                              ($125.00) represented a pure price increase of 89.7 percent, which was
                              incorporated into the CPI.




                              7
                               In this instance, the price of the old version was an imputed price. Imputed price is a term used by
                              BLS to indicate that the old version did not have a usable price. Instead, an average was calculated
                              from the price change experienced in the previous collection period by the same type of items in the
                              CPI to handle a missing or unusable price quotation.



                              Page 69                                                     GAO/GGD-99-84 Consumer Price Index
                            Appendix IV
                            Direct Adjustments




                            A third method of direct price adjustments includes adjustments for an
Other Direct                item’s size or the number of units it includes. BLS considers adjustments
Adjustments                 for unit count or size to be direct adjustments because they adjust for a
                            particular characteristic of a good or service. BLS also includes data-entry
                            corrections for individual items in this direct adjustment method.

                            In 1997, BLS made direct adjustments to at least 719 items to account for
                            unit-count changes, size differences, or data errors, which represented
                            about 6 percent of the 12,131 adjusted substitutions in 1997 that involved
                            nonrent items. Of these 719 direct adjustments, 336 involved medical-care
                            items, 132 were for housing-related items, 107 were in food and beverage
                            items, 95 involved entertainment items, and the remaining 49 were in
                            transportation and the “other” goods and services ELIs.

                            According to BLS, a large number of adjustments for unit count or size
                            differences are made automatically in the CPI. For example, most food and
                            beverage prices are automatically converted to an effective price on a
                            per-ounce basis. These conversions are not counted as substitutions and
                            are not reflected in the count of direct adjustments.


How BLS Accounts for        For purposes of measuring the price change of an item of constant quality,
Quality Changes             the item should remain the same in unit count or size between collection
Associated With             periods. Before calculating the price change between two versions of an
                            item that have experienced a size or unit count change, BLS adjusts the
Differences in Unit Count   price of the old version to the level it would have been if the old version
or Size, and Corrections    had the same unit count or size as the new version.


How BLS Calculates Price    To illustrate a price calculation with an “other” direct adjustment method,
Changes After Direct        suppose for example, that a package of gum has five sticks in one period
Adjustments for             and sells for 50 cents, and at the next price collection period the package
                            has six sticks of gum and sells for 60 cents. According to BLS, the
Differences in Unit Count   difference in unit count would make this substitution not comparable.
or Size, and Corrections    However, the BLS commodity analyst can make a direct quality adjustment
                            to the price of the old version to account for the difference in unit count.
                            The analyst would calculate the price per stick of gum in the previous
                            collection period as 10 cents. The 10 cents is added as a direct adjustment
                            to the 50 cents price of the old version to price it as if it had six sticks.
                            After the adjustment, the old price for the pack of gum is 60 cents. The 10
                            cents is what BLS would count as quality change in the context in which BLS




                            Page 70                                      GAO/GGD-99-84 Consumer Price Index
                          Appendix IV
                          Direct Adjustments




                          generally uses the term quality—all changes that are not pure price
                          changes.

                          If no adjustment for unit count had been made, the price difference
                          between the two packages of gum would have been reflected as a
                          20-percent increase (the increase from 50 cents to 60 cents per pack).
                          However, after the direct adjustment to the price of the old version, based
                          on the price per stick of gum in the package, the price of the old version
                          was 60 cents. Because the price of the new version was also 60 cents, the
                          CPI would reflect no price change (0 percent) for the item.



Examples of BLS’ Use of   The following two examples of BLS’ use of direct adjustments for unit
Direct Adjustments for    count or size, or corrections come from our interviews with BLS
Size or Unit Count, and   commodity analysts. According to the commodity analysts and
                          supervisors, these examples are illustrative of the substitutions in which
Corrections               they use such direct adjustments in their respective CPI components. These
                          examples show analysts’ decisions when the difference in unit count or
                          size between versions are large enough to make the items not comparable,
                          even though, according to BLS, it also makes such adjustments when the
                          differences are not as large. In addition, in these examples the analyst has
                          sufficient information to make a direct adjustment to the price of the old
                          version, and compare that adjusted price with the price of the new
                          version.

                          The examples include a price calculation that shows the percentage
                          change in price that went into the CPI for that month as a result of the
                          direct adjustments. These percentage changes in price are applicable only
                          to the particular cases illustrated. They are not intended to be
                          representative of the percentage changes in price that occurred for similar
                          kinds of substitutions in which direct adjustments are made. Nor are the
                          percentage changes in price intended to be representative of the impact
                          that direct adjustments have on the CPI.

                          Both examples include (1) a table showing the specifications of the old
                          version and the new version that replaced it, (2) the analyst’s reasons for
                          judging the two versions to be comparable or not comparable, (3) the
                          rationale for selecting the method of adjustment, and (4) the price
                          calculation that was made. Within each table, differences in specifications
                          between the old and new versions are highlighted by shading.




                          Page 71                                     GAO/GGD-99-84 Consumer Price Index
                                      Appendix IV
                                      Direct Adjustments




                                      The new version of the season ticket included 44 hockey games, while the
Example 3 - Season                    old version included 43 games. The analyst decided to make a direct
Ticket (Under the                     adjustment to account for the difference in the number of games.
Admission to Sporting
Events Item Stratum)

Table IV.3: Characteristics of Two
Versions of a Season Ticket Package   Description                      Old version                       New version
                                      Specifications
                                       Type of event                   Hockey                            Hockey
                                       Level of competition            Professional                      Professional
                                       Admission type                  Season ticket for 43 events       Season ticket for 44 events
                                       Seating location                Middle price                      Middle price
                                       Type of purchase                In person                         In person
                                      Price information
                                       Amount                          $1,617.432                        $1,677.00
                                       As of                           September                         October
                                       Collected/imputed               Imputed                           Collected

                                               Specifications that differ between old and new versions

                                      Source: BLS.




Analyst’s Reasons for                 The commodity analyst decided that, because the number of tickets in the
Deciding the Items Were               season ticket package increased from 43 to 44 hockey games, a direct
Not Comparable                        quality adjustment could be made. According to BLS, it is not clear whether
                                      the analyst would have decided this was a comparable or a
                                      not-comparable substitution if a direct adjustment had not been made.


Rationale for Selecting the           To make a direct adjustment in this case, the analyst overrode the default
Method of Adjustment                  comparison code that the computer had generated. For items in the
                                      entertainment component of the CPI, a computer routine automatically sets
                                      the comparison codes so that the prices of the old and new versions will
                                      be compared unless the analyst sets them to another value. This computer
                                      routine also calculates the percentage change in the versions’ prices for
                                      the convenience of the analyst. When there is a change that meets or
                                      exceeds a predetermined amount, the computer is programmed to
                                      calculate a price increase (or decrease) and a commodity listing review is




                                      Page 72                                                     GAO/GGD-99-84 Consumer Price Index
                       Appendix IV
                       Direct Adjustments




                       forwarded to the commodity analyst. In this case, the computer defaulted
                       to a comparable adjustment code signifying that the old and new versions
                       were alike, and the computer generated a price increase of 3.7 percent.
                       However, the analyst overrode this adjustment code to make the direct
                       adjustment. The analyst calculated the quality adjustment factor by
                       calculating the incremental difference in the increase in the number of
                       hockey tickets and then entered this factor (1/43 = 0.233) as part of the
                       adjustment procedure.


Calculation of Price   The price calculation for the direct quality adjustment method is done by
Change                 computer using the adjustment factor. According to BLS, the price of the
                       old version was increased by 2.33 percent to have it represent what a
                       44-event season ticket would have cost. Using this increased cost of the
                       old version, the computer compared it with the cost of the new version of
                       the season ticket.8 In doing so, according to BLS, the computer calculated a
                       1.3 percent increase for this season ticket, and this percentage increase
                       was incorporated into the CPI for October 1997.


                       The analyst concluded that two soups were not comparable and made a
Example 4 - Canned     direct adjustment to correct for the size of the container.
Soup (Under the
Canned and Packaged
Soup Item Stratum)




                       8
                        In this instance, the price of the old version was an imputed price. BLS officials said that, when prices
                       are unavailable (e.g., because the item is out of season), they impute the prices of those items, using
                       the same imputation that is used for the linking method. BLS imputes prices using averages calculated
                       from the price changes experienced in the same month by the same type of items in the CPI to handle
                       missing or unusable price quotations.



                       Page 73                                                       GAO/GGD-99-84 Consumer Price Index
                                     Appendix IV
                                     Direct Adjustments




Table IV.4: Characteristics of Two
Versions of a Canned Soup            Description                        Old version                     New version
                                     Specifications
                                      Where purchased (in               Shelf                           Shelf
                                      store)
                                      Type                              Condensed                       Condensed
                                      Brand classification              Nationally marketed brand       Nationally marketed brand
                                      Flavor                            Chicken                         Chicken
                                      Other major ingredients           Noodles                         Noodles
                                      Packaging                         Canned                          Canned
                                      Weight                            10.375 oz.                      10.75 oz.
                                     Price information
                                      Amount                            $1.324                          $0.95
                                      As of                             October                         November
                                      Collected/imputed                 Imputed                         Collected

                                              Specifications that differ between old and new versions



                                     Source: BLS.




Analyst’s Reasons for                The commodity analyst decided that the two versions were the same soup,
Deciding the Items Were              but a message from the price taker indicated that the size of the container
Not Comparable                       had been corrected. The size of the soup was previously reported
                                     incorrectly as 10.375 ounces and was now being reported correctly as
                                     10.75 ounces. To make a correction for this weight error, the commodity
                                     analyst made the two soups not comparable.


Rationale for Selecting the          To make a direct adjustment in this case, the analyst had to override
Method of Adjustment                 another adjustment code that was computer generated. In this case, the
                                     computer defaulted to a comparable adjustment code signifying that the
                                     old and new versions were alike, and the computer generated a price
                                     decrease of 30.75 percent. However, because the analyst wanted to correct
                                     the weight error, the analyst overrode this adjustment code. The analyst
                                     calculated the quality adjustment factor by calculating the percentage that
                                     represents the proportion of the price of the old version that was in error
                                     to make it the same as the larger-size new version and then entered this
                                     factor (–0.0597) as part of the adjustment procedure.




                                     Page 74                                                     GAO/GGD-99-84 Consumer Price Index
                       Appendix IV
                       Direct Adjustments




                       For not-comparable substitutions in this item stratum, BLS designated the
                       linking method as the standard method of adjustment when other methods
                       are not usable. However, in this instance, the analyst had enough
                       information to make a direct adjustment to equate the two sizes of soup.
                       The analyst said that if the linking method had been selected, the soup
                       would have been excluded from the calculation for the CPI that month.
                       According to the analyst, a quality adjustment would allow a continuation
                       of the same item in the CPI and allow the analyst to correct the weight
                       error.


Calculation of Price   The price calculation for the direct quality adjustment is done by computer
Change                 using the adjustment factor. According to BLS, the effective price per ounce
                       of the old version was decreased by 0.0597 percent to have it priced at the
                       correct weight. Using this decreased cost of the old version of the soup,
                       the computer compared the corrected unit price with the unit price of the
                       new version. In doing so, according to BLS, the computer calculated a
                       26.4 percent decrease for this canned chicken soup, which was
                       incorporated into the CPI for November 1997.




                       Page 75                                     GAO/GGD-99-84 Consumer Price Index
Appendix V

Class-Mean Method of Adjustment


              According to BLS, the class mean is an indirect method of adjustment that
              is used for not-comparable substitutions in item strata where price
              changes are closely associated with annual or periodic introductions of
              new models or product lines. For example, BLS uses the class-mean
              method for the new cars item strata because researchers noted that
              automobile manufacturers often increase the prices of their automobiles
              when they introduce the new year’s models. Although BLS officials said
              they would prefer to make a direct adjustment under these circumstances,
              often they do not have the information they need to do so. (See app. IV for
              a discussion about direct adjustments.) When a direct adjustment cannot
              be made for a new model, the class-mean method is used because,
              according to BLS, it more accurately reflects the manufacturers’ price
              increase than the linking method, the other available method of
              imputation.1

              BLS reported that the class-mean method is similar to the linking method in
              several ways (see app. VI). Both methods use imputations to estimate
              rates of price change when an item in the CPI is replaced by a
              not-comparable substitution and a direct adjustment cannot be made. (BLS
              also refers to these rates of price change as price relatives.) Both methods
              depend upon two fundamental assumptions: (1) the price change
              applicable to the not-comparable substitution cannot be directly
              calculated and (2) the best available estimate of this price change is the
              rate of price change that occurs for the same type of items in the same
              geographic area.

              The class-mean differs from the linking method in that it is based on the
              price changes of a much more specific group of price quotations. Whereas
              the linking method is based on all price quotations in the same item
              stratum in the same geographic area, the class-mean method is based on a
              specific subset that includes only quotations for substitutions, which are
              judged comparable and/or to which direct adjustments are made in the
              same item stratum in the same geographic area.2


              BLS officials said they developed the class-mean method in response to a
Background    problem they perceived in the linking method in the late 1980s. Until the
              class-mean method was developed, the linking method was the only

              1
               Imputation is a term used by BLS to indicate that the actual price of the substitution is not used.
              Instead, an average is calculated from the price change experienced that month by the same types of
              items in the CPI to handle a missing or unusable price quotation.
              2
               BLS refers to an item stratum in a geographic area as an item stratum-index area.



              Page 76                                                     GAO/GGD-99-84 Consumer Price Index
Appendix V
Class-Mean Method of Adjustment




method of imputation available to BLS for most substitutions that were
judged to be not comparable.3 In BLS’ opinion, the linking method tended to
understate the degree of price increase for items for which new models or
product lines were frequently introduced.

According to BLS, its original research indicated that manufacturers usually
raised the price of an item when they introduced a new model or product
line. As a result, BLS officials stated, new models and products usually had
higher average price increases than unchanged models and products.
However, because the linking method was heavily influenced by
unchanged models and products, it tended to understate the level of price
increase when it was applied to new models and products.4

In response to this problem, BLS developed a method for calculating a price
relative that was based on comparable and directly adjusted substitutions
because the prices of these substitutions, BLS officials said, were more
likely to have changed due to the introduction of new models or products.
BLS officials relied on both logic and research to justify their use of this
subset of substitutions for the class mean.

BLS made several assumptions in arriving at this subset of substitutions for
the class mean. It assumed that this subset of substitutions contained a
large percentage of new models or products lines. BLS further assumed that
the not-comparable substitutions in these item strata also contained a
large percentage of new models or product lines but that the unchanged
items did not.5 Under these circumstances, BLS officials stated, it is more
appropriate to impute the prices of not-comparable substitutions from a
subset of comparable and directly adjusted substitutions than from all
price quotations, most of which consist of items that were not
substitutions and whose characteristics were unchanged.


3
 The only other method of imputation available at the time, the overlap method, could only be applied
to substitutions for items that had been on sale. The purpose of the overlap method, BLS officials
stated, was to prevent biases from entering the CPI as a result of substitutions for items that had been
on sale. BLS officials said that, as a result of the introduction of other methods for preventing biases
from sale price items entering the CPI, the overlap method had almost entirely been phased out by
December of 1997.
4
 According to BLS officials, the class mean was developed in the late 1980s to address the higher than
average price increases that usually accompanied the introduction of new models or product lines.
However, in recent years, BLS has reported that some manufacturers have decreased the prices of new
models or product lines. If this decrease was lower than the average decrease that would have
occurred if the linking method was used, the price changes of the adjusted items would be overstated
if the linking method was used.
5
 By unchanged items, we mean items that were not substitutions and whose characteristics had not
changed.



Page 77                                                      GAO/GGD-99-84 Consumer Price Index
                      Appendix V
                      Class-Mean Method of Adjustment




                      To check these assumptions, BLS conducted research, in the early 1990s,
                      the average price changes for (1) comparable and directly adjusted
                      substitutions and (2) unchanged items. BLS looked at each item stratum
                      individually, and compared average price changes. BLS reported that it
                      found the rates of price change for comparable and directly adjusted
                      substitutions higher than for unchanged items in many item strata where
                      their industry knowledge indicated that new models and products were
                      regularly introduced. The reason for this, BLS officials decided, was that
                      the new models and products in these item strata generally had higher
                      rates of price increase than the unchanged items.

                      In the early 1990s, BLS conducted a broad review of the CPI item strata.
                      Following this review, it decided that the class-mean method would be
                      used in item strata where (1) new models or products were known to be
                      introduced on a regular basis and (2) the average price increase for the
                      comparable and directly adjusted items was significantly higher than for
                      unchanged items. Under these circumstances, BLS officials stated, the class
                      mean was a better method of imputation than the linking method.

                      BLS commodity analysts used the class-mean method in 33 percent of the
                      12,131 not-comparable substitutions in 1997. In December 1997, the class
                      mean was designated for 53 of the 183 priced item strata.6 These included
                      most apparel item strata, some transportation item strata, such as new
                      trucks and cars, and many household-goods item strata. The
                      household-goods item strata included furniture, kitchen appliances,
                      electrical goods, utensils, linens, and cleaning agents and tools. In
                      addition, the class mean was designated for a variety of other items, such
                      as printed items, boats, bicycles, outdoor equipment, photographic goods,
                      pet food, and auto repair services.


                      In the class-mean method, the replacement version is put aside in
How Price Changes     calculating the price change. Instead, a computerized procedure is
Are Calculated With   followed to impute a rate of change that is based on items similar to the
the Class-Mean        old version. A first step in this procedure is for the computer to identify
                      the substitution’s item stratum and geographic area. For example, if a
Method                price taker in Urbantown made a not-comparable replacement for a new
                      car in November 1997, the class-mean method would use the new cars
                      item stratum for Urbantown, November 1997.



                      6
                       After the 1998 revision of the CPI, the class mean was designated for 51 of the 186 priced item strata.



                      Page 78                                                      GAO/GGD-99-84 Consumer Price Index
Appendix V
Class-Mean Method of Adjustment




The next step in the computerized procedure is to identify all the
comparable and direct adjusted substitutions and calculate a price change
rate for the applicable item stratum. As a result, the class-mean
method—unlike the linking method—excludes all items that are not
substitutions from its calculations.7 Because the great majority of the
items are not substitutions each month, this means that the class mean
often is based on a fairly small number of items.8 In the three illustrations
that follow this section, the class means were based on as few as 1
substitution and as many as 11 substitutions. If there are no comparable or
directly adjusted substitutions in the item stratum for the geographic area
in question, BLS said the computer routines would search other item strata
or geographic areas that had been defined as similar until an item stratum
with at least one comparable and/or directly adjusted substitution is
found. For example, if there were no comparable or directly adjusted
substitutions in women’s coats and jackets in Urbantown in
November 1997, the computer would search for comparable or directly
adjusted substitutions in women’s separates and sportswear stratum in
Urbantown in November 1997.9

When comparable and/or directly adjusted items are identified, either in
the original item stratum in the original geographic area or in similar strata
or areas, a weighted-average price change is determined. BLS then assigns
this weighted average to all class-mean replacement items, in that item
stratum, in that geographic area, in that month. For example, assume an
item stratum in a geographic area contained eight items with equal
weights. Of those eight items, one is the item under review, which is a
not-comparable substitution. The other seven are one comparable
substitution whose price has increased by 5 percent, one directly adjusted
substitution whose price has increased by 1 percent, and five items that
are not substitutions. Only the comparable and the directly adjusted

7
 Because the class-mean method only includes comparable and/or directly adjusted substitutions, it
also excludes all linked substitutions from its calculations.
8
 Less than 4 percent of the 872,829 monthly or bimonthly price quotations in the CPI were
substitutions in 1997. However, the rate of substitutions varied by major components, ranging from
12.8 percent for apparel items to 1.4 percent for food items.
9
 According to BLS officials, judgment was used to establish fairly elaborate search routines for the
class mean. Officials said they ordered the item strata in terms of their similarity with the item stratum
in the geographic area under consideration. For example, in the example given here, if there were no
comparable or directly adjusted substitutions in the women’s separates and sportswear item stratum,
the routine would perform a search on the women’s dresses item stratum. If there were no comparable
or directly adjusted substitutions in that item stratum, the computer would perform a search on the
women’s underwear, nightwear, and accessories item stratum. In other item strata, such as new cars,
the computer routine would search other geographic areas for comparable or directly adjusted
substitutions. For example, if there were no comparable or directly adjusted substitutions for new cars
in the New York City suburbs, the routine would search New York City.



Page 79                                                      GAO/GGD-99-84 Consumer Price Index
                       Appendix V
                       Class-Mean Method of Adjustment




                       substitutions would be used to calculate the class mean. The percentage
                       changes would be added together for a total of 6 percent, and divided by
                       the number of substitutions used. The result would be an average increase
                       of 3 percent, which would be the price relative for all class-mean
                       substitutions in that item stratum in that geographic region for that month.

                       Of course, the computations that BLS makes to obtain a weighted-average
                       price change are more complex than those in the illustration. For example,
                       BLS assigns a weight to each item in a stratum to reflect its relative
                       importance to consumers. (The weights are based on consumer spending
                       patterns with the larger weights assigned to items on which consumers
                       spend the most.) BLS uses these weights in calculating the overall net price
                       change (price relative) for the class-mean method. Because the price
                       change of each item is multiplied by the weight of the item, price changes
                       in items with large weights are likely to have a greater impact on the
                       overall net price change than price changes in items with smaller weights.

                       Price relative calculations for the class-mean method are made by
                       computer routine after all of the month’s prices have been collected and
                       the replacement items have been reviewed by the commodity analysts.
                       Each month, a single price relative is computed for each item stratum in
                       each of the CPI geographic areas that BLS has designated for the class mean
                       and that has at least one not-comparable substitution.


                       Quality improvements are accounted for in the class-mean method exactly
How Quality            as they are in the linking method (see app. VI). Conceptually, under the
Improvements Are       class-mean method, BLS divides any difference in price between the old
Accounted for in the   item and the replacement item into two parts—pure price and quality. BLS
                       makes the implicit assumption that the pure price change is an amount
Class-Mean Method      that can be estimated by the rate of price change that occurred for a
                       subset of the same types of items in the same geographic area. In other
                       words, BLS assumes that the pure price change is the price relative
                       calculated through the class-mean method. BLS assumes that any remaining
                       difference in price—the residual—reflects differences in quality between
                       the old item and its replacement. This residual is excluded from the price
                       relative calculation because the CPI is designed to reflect only pure price
                       change.




                       Page 80                                     GAO/GGD-99-84 Consumer Price Index
                       Appendix V
                       Class-Mean Method of Adjustment




                       The following three examples of BLS’ use of the class-mean method come
Examples of BLS’ Use   from our interviews with three BLS commodity analysts. See appendix I for
of the Class-Mean      a detailed discussion of how these examples were selected. According to
Method                 the commodity analysts and their supervisors, these examples illustrate
                       how the class-mean method works. These examples show analysts’
                       decisions when the replacement is different from the old version and is
                       viewed as a dissimilar item.

                       Each example includes a price calculation, showing the percentage change
                       in price that went into the CPI for that month as a result of the class-mean
                       adjustment. Each percentage change in price is applicable only to the
                       particular case illustrated. The examples are not intended to be
                       representative of the percentage changes that occurred for similar kinds of
                       substitutions in which class-mean adjustments are made. Nor are the
                       percentage changes intended to be representative of the impact that
                       class-mean adjustments have on the CPI.

                       Each example includes (1) a table showing the specifications of the old
                       version and the new version that replaced it, (2) the analyst’s reasons for
                       judging the two versions to be not comparable, (3) the rationale for
                       selecting the method of adjustment, and (4) the calculation of price
                       change that was made. Within each table, differences in specifications
                       between the old and new versions are highlighted by shading.


                       The analyst decided that the new version of a bed dust ruffle was not
Example 1 - Bedroom    comparable with the old version. The substitution was adjusted using the
Linens (Under the      class-mean method, which was the designated method of adjustment for
Linens, Curtains,      this item stratum.

Drapes, and Sewing
Materials Item
Stratum)




                       Page 81                                      GAO/GGD-99-84 Consumer Price Index
                                    Appendix V
                                    Class-Mean Method of Adjustment




Table V.1: Characteristics of Two
Versions of a Dust Ruffle           Description                            Old version                      New version
                                    Specifications
                                         Quality                           First                            First
                                         Item priced                       Dust ruffle                      Dust ruffle
                                         Style                             Ruffles                          Ruffles
                                         Size                              Standard, twin, or single        Queen
                                         Fabric                            Woven                            Woven
                                         Fiber                             90 percent cotton                90 percent cotton
                                         Backing                           No backing                       No backing
                                         Brand/label name                  Different from the new           Different from the old version
                                                                           version
                                    Price information
                                         Amount                            $39.12                           $199.99
                                         Type of price                     Sale                             Regular
                                         As of                             August                           October
                                         Collected/imputed                 Imputed                          Collected

                                                 Specifications that differ between old and new versions



                                    Source: BLS.




Analyst’s Reasons for               The analyst said that the new version of the ruffle was different in size
Deciding the Items Were             from the old version because it was a queen-size dust ruffle, as compared
Not Comparable                      to the old version which was a single-size dust ruffle.10 Therefore, the
                                    versions were not comparable, and an adjustment had to be made.


Rationale for Selecting the         For not-comparable substitutions in this item stratum, BLS designated the
Method of Adjustment                class-mean method as the standard method of adjustment for substitutions
                                    that were not comparable. In this instance, the class-mean adjustment
                                    code was generated by computer routines because the price difference
                                    between the two versions exceeded a predetermined level. As the analyst
                                    decided that this substitution was not comparable, the class-mean
                                    adjustment code was not changed, and the class-mean method of
                                    adjustment was used.



                                    10
                                      BLS price takers can report that the bedroom linens fall into one of five sizes: (1) standard, twin, or
                                    single, (2) full, (3) queen, (4) king, or (5) other. Throughout this discussion, we refer to the standard,
                                    twin, or single-size dust ruffle as a single-size dust ruffle.



                                    Page 82                                                         GAO/GGD-99-84 Consumer Price Index
                        Appendix V
                        Class-Mean Method of Adjustment




Calculation of Price    The price calculation for the class-mean method is done entirely by
Change                  computer routines without the direct involvement of the analyst. The
                        method calculates a rate of price change based on the price movement of
                        comparable and/or directly adjusted prices in the same item stratum in the
                        same geographic area. In this instance, according to BLS, there was only
                        one comparable substitution in the same item stratum in the same
                        geographic area, and that served as the class mean. The comparable
                        substitution was for a quilt and comforter whose price had changed from
                        $179.99 to $179.00. Therefore a –0.5 percent rate of change was calculated
                        for this item stratum in this geographic area for use with class-mean
                        adjustments for October 1997, and the price of the old version of the dust
                        ruffle was adjusted by –0.5 percent, from $39.12 to $38.91.11 This means
                        that the CPI in October 1997 reflected the same percentage change for this
                        dust ruffle.


                        The analyst decided that the new version of a woman’s parka was not
Example 2 - Woman’s     comparable with the old version. The substitution was adjusted using the
Parka (Under the        class-mean method, which is the designated method of adjustment for this
Women’s Coats and       item stratum when direct adjustments cannot be made.

Jackets Item Stratum)




                        11
                         The price for the old version was an imputed price. BLS officials said that, when prices are
                        unavailable (e.g., because the item is out of season) they impute the prices of those items using the
                        same imputation method that is used for the linking method.



                        Page 83                                                      GAO/GGD-99-84 Consumer Price Index
                                    Appendix V
                                    Class-Mean Method of Adjustment




Table V.2: Characteristics of Two
Versions of a Woman’s Parka         Description                        Old version                     New version
                                    Specifications
                                     Type                              Heavyweight                     Heavyweight
                                     Design                            Parka                           Parka
                                     Body fiber                        Leather                         Leather
                                     Brand/label name                  Local brand name                Different local brand name
                                     Size range                        Junior/misses                   Junior/misses
                                     Length                            Fingertip length                Fingertip length
                                     Details/ features                 Hood                            No hood
                                     Other information                 Fur-like lining                 Fur-like lining
                                     Other information                 Gold                            Brown
                                    Price information
                                     Amount                            $153.45                         $79.99
                                     Type of price                     Regular                         Regular
                                     As of                             August                          October
                                     Collected/imputed                 Imputed                         Collected

                                             Specifications that differ between old and new versions



                                    Source: BLS.




Analyst’s Reasons for               The analyst noted that there was a big difference in the prices of the old
Deciding the Items Were             and the new versions ($153.45 compared with $79.99). The analyst also
Not Comparable                      said that the research BLS had conducted to develop hedonic models for
                                    direct adjustments had found that hoods could influence the price of a
                                    coat. Furthermore, he was not familiar with the brand names of the old
                                    and new versions and did not know how they might differ in terms of
                                    quality. According to BLS, there are thousands of brands of women’s coats,
                                    and it is not possible for commodity analysts to be familiar with all of
                                    them. Even though there is a hedonic regression model for the women’s
                                    coats and jackets item stratum, the analyst could not make a direct
                                    adjustment because the model did not contain any cost factors for
                                    differences in brand names. Taking all of this into consideration, the
                                    analyst decided this substitution needed to be adjusted by an indirect
                                    method.


Rationale for Selecting the         For not-comparable substitutions in this item stratum, BLS has designated
Method of Adjustment                the class-mean method as the standard method of adjustment if a direct



                                    Page 84                                                     GAO/GGD-99-84 Consumer Price Index
                       Appendix V
                       Class-Mean Method of Adjustment




                       adjustment could not be made. In this instance, the computer routine had
                       generated a class-mean adjustment code because the price difference
                       between the two versions exceeded a predetermined level. As the analyst
                       decided that the substitution was not comparable, and a direct adjustment
                       could not be made, the computer-generated code was not changed, and
                       the substitution was adjusted using the class-mean method.


Calculation of Price   The price calculation for the class-mean method is done entirely by
Change                 computer routines without the direct involvement of the commodity
                       analyst. The method calculates a rate of price change based on the
                       observations for comparable and/or directly adjusted substitutions in the
                       same item stratum in the same geographic area. In this case, a BLS
                       computer routine found three comparable substitutions in the original
                       item stratum and geographic area. These were substitutions for (1) a
                       heavyweight leather jacket, for which the price had risen by 1.9 percent,
                       (2) a heavyweight polyester jacket, for which the price had fallen by
                       29.8 percent, and (3) a heavyweight leather anorak, for which the price
                       had risen by 69.2 percent. There were no directly adjusted substitutions in
                       this item stratum and geographic area.

                       The class mean, which is a weighted average of these three items, was
                       calculated to be 6.1 percent for October 1997. Therefore, the price of the
                       old version was adjusted by 6.1 percent, from $153.45 to $162.81. This
                       means that the CPI in October 1997 reflected the same percentage change
                       for this version of a woman’s jacket.


                       The analyst decided that the new version of a new car was not comparable
Example 3 - New Car    with the old version. The substitution was adjusted using the class-mean
(Under the New Cars    method, which was the designated method for not-comparable
Item Stratum)          substitutions in this item stratum when direct adjustments could not be
                       made.




                       Page 85                                     GAO/GGD-99-84 Consumer Price Index
                                    Appendix V
                                    Class-Mean Method of Adjustment




Table V.3: Characteristics of Two
Versions of a New Car               Description                        Old version                       New version
                                    Specifications
                                     Model year                        1997                              1998
                                     Number of doors                   4-doors                           4-doors
                                     Body type                         Sedan                             Sedan
                                     Number of cylinders               4                                 4
                                     Engine displacement or            1.6 liters                        1.6 liters
                                     size
                                     Transmission                      3-speed standarda                 1722a
                                     Options package
                                       Rear defogger                   Yes                               Yes
                                       Power steering                  Yes                               Yes
                                       AM/FM cassette                  No                                Yes
                                       Tilt wheel                      No                                Yes
                                       Power windows, locks,           No                                Yes
                                       and mirror
                                     Air conditioning                  Yes                               Yes
                                    Price information
                                     Amount                            $14,010                           $14,408
                                     As of                             August                            October
                                     Collected/imputed                 Collected                         Collected

                                             Specifications that differ between old and new versions


                                    a
                                     Although this information was reported on the CRL, the analyst for new cars stated that, in fact,
                                    both the old and new models of this car had 4-speed automatic transmissions.


                                    Source: BLS.




Analyst’s Reasons for               A 1998 model of a car had replaced the 1997 model. According to the
Deciding the Items Were             commodity analyst, the characteristics of the new version, and publicly
Not Comparable                      available industry information, showed that the 1998 model contained
                                    significant quality changes over the 1997 model. However, the analyst did
                                    not have the information that would have allowed a direct adjustment to
                                    be made because the automaker had not provided sufficient information
                                    to BLS, and the information provided by other automakers could not be
                                    applied to this model.




                                    Page 86                                                     GAO/GGD-99-84 Consumer Price Index
                              Appendix V
                              Class-Mean Method of Adjustment




Rationale for Selecting the   In this case, the computer-generated default comparison code was
Method of Adjustment          comparable based on the level of the price change. If the analyst had
                              judged the substitution to be comparable, the price change would have
                              been 2.8 percent. However, the analyst stated that publicly available
                              industry information indicated that there were significant differences in
                              the features of the old and new versions. In the analyst’s opinion, quality
                              changes had been made, but the automaker had not provided the
                              information necessary to make a direct adjustment; and the information
                              provided by other automakers was not applicable to this model. Therefore,
                              the analyst inserted a code indicating that the class-mean method should
                              be used. In the new cars item strata, the analyst stated, the class mean is
                              always used when the manufacturer makes significant quality changes to a
                              new year’s model but does not provide the information to allow BLS to
                              make direct adjustments.


Calculation of Price          The price calculation for the class-mean method is done entirely by
Change                        computer routines without direct involvement by the commodity analyst.
                              The method calculates a rate of price change based on other changes in
                              the same item stratum in the same geographic area. In this instance, a BLS
                              computer routine searched for all the comparable and directly adjusted
                              substitutions in the item stratum of the substituted item in its geographic
                              area. Of the 11 substitutions that were found, 5 were comparable
                              substitutions and 6 were directly adjusted substitutions. The 11
                              substitutions, which had price changes ranging from a 0.3 percent
                              decrease to a 10.5 increase, were all used to calculate a weighted-average
                              price increase of 4.0 percent. Therefore, the price of the 1997 model was
                              imputed to have risen by 4.0 percent, from $14,010 to $14,570. This means
                              that the CPI in October 1997 incorporated the same percentage change for
                              this car.




                              Page 87                                     GAO/GGD-99-84 Consumer Price Index
Appendix VI

Linking Method of Adjustment


              BLS’ linking method is an indirect adjustment method. It uses imputations1
              to produce rates of price change when an item in the CPI is replaced with a
              substitution that is not comparable and no other adjustment method can
              be used. (BLS also refers to these rates of price change as price relatives.)
              With the linking method, BLS makes two fundamental assumptions: (1) the
              price change applicable to the not-comparable substitution cannot be
              directly calculated and (2) the best available estimate of this price change
              is the rate of price change that occurs for the same type of items in the
              same geographic area. If prices for these items went up or down 5 percent,
              then the price of the item in question is imputed to go up or down
              5 percent as well.


              According to BLS, the linking method is one of several adjustments that can
Background    be used when a BLS price taker submits a replacement—a new
              version—that a commodity analyst judges to be very different from the old
              version (the versions are not comparable). The difference between the
              new and old versions is so large, in the analyst’s view, that a determination
              cannot be made as to whether it is due solely to pure price change. In
              addition to such a difference, two other conditions must also exist for a
              commodity analyst to designate the linking method: (1) data are not
              available to use a direct adjustment method and (2) BLS has not designated
              use of the class-mean method for the item’s-in-question item strata. In
              other words, analysts are to use linking when they cannot use any other
              adjustment method. (The direct adjustment and the class-mean methods
              are discussed in apps. IV and V, respectively.)

              When faced with not-comparable substitutions in 1997, BLS commodity
              analysts used the linking method to determine price changes for
              36 percent of the 12,131 adjusted substitutions in 1997 involving nonrent
              items. The use of linking was also widespread. It was used in nearly all of
              the nonrent item strata in which an adjustment was made in 1997. In
              addition, a majority of the adjustments for two CPI components—food and
              beverages and medical care—were calculated using the linking method.




              1
               Imputation is a term used by BLS to indicate that the actual price of the substitution is not used.
              Instead, an average is calculated from the price change experienced that month by the same type of
              items in the CPI to handle a missing or unusable price quotation.



              Page 88                                                     GAO/GGD-99-84 Consumer Price Index
                      Appendix VI
                      Linking Method of Adjustment




                      In the linking method, the new version is put aside in calculating the price
How Price Changes     change. Instead, a procedure is followed to impute a rate of change that is
Are Calculated With   based on the remaining items priced in the appropriate item stratum. A
the Linking Method    first step in this procedure is to identify the item strata category, which
                      contains items that are similar by definition to the old version. There are
                      over 200 of these categories, and they are replicated for each geographic
                      area in which BLS collects CPI data. The item stratum that is used must be
                      for the same geographic area and month in which the replacement
                      occurred. For example, if a price taker in Urbantown made a
                      not-comparable replacement for a soup in November, the linking method
                      would use the canned and packaged soups item stratum for Urbantown for
                      November.

                      Next, an average of price change rates is calculated for the applicable item
                      stratum. In essence, BLS excludes from this computation all items that
                      entered the item stratum during the month through the linking method, as
                      well as those items that for various reasons do not have a usable price,
                      such as a seasonal item that is temporarily unavailable. BLS includes all
                      other items in that stratum in the computation—those with a price change
                      and those with no change in price. For these items a price relative, which
                      BLS refers to as a weighted-average price change, is determined for the
                      item-stratum index for that geographic area.2 This weighted-average price
                      change, expressed as a percentage, is assigned to all linked replacement
                      items for that item stratum, month, and geographic location.

                      The calculations for a weighted-average price change are complex but
                      basically involve two concepts: determining an average and weighting the
                      items that are averaged. For example, assume an item stratum in a
                      geographic area contained six items with equal importance in the CPI. Of
                      those 6 items, 4 had no change in price during the month, 1 had a price
                      increase of 20 percent, and 1 was a replacement that was going to be
                      adjusted by the linking method. The replacement item would be put aside,
                      and the total of the price change for the 5 items (20 percent) would be
                      divided by 5, the number of items remaining with usable prices. The result
                      would be an average increase of 4 percent—the price relative for the
                      geographic area’s item-stratum index for the replacement item, which
                      would be used only for that month.

                      Before calculating the average price change, BLS assigns a weight to each
                      item in an item stratum to give proportionate emphasis to it in relation to

                      2
                       Each month BLS calculates indexes for each item stratum and geographic area, which total to more
                      than 8,000 indexes. BLS refers to these as “item stratum-index areas,” which are aggregated into a U.S.
                      city average index for all items.



                      Page 89                                                     GAO/GGD-99-84 Consumer Price Index
                        Appendix VI
                        Linking Method of Adjustment




                        other items in the item stratum. (The weights are based on consumer
                        spending patterns with the larger weights assigned to items on which
                        consumers spend the most.) BLS uses these weights in calculating the
                        overall weighted-average price change (price relative) for the linking
                        method. Because the change in the price of each item is multiplied by the
                        item’s weight, price changes in items with large weights are likely to have
                        greater impact on the overall net price change than price changes in items
                        with smaller weights.

                        Price relative calculations for the linking method are made by computer
                        routine after all of the prices for the month have been collected and the
                        replacement items have been reviewed by the commodity analysts. A
                        single price relative is computed for each item stratum as appropriate for
                        the month. Normally, this price relative is most heavily influenced by items
                        that were not substitutions in the item stratum, which generally are most
                        items.

                        According to BLS, the use of the linking method is the same as setting aside
                        the individual price quotation from the CPI calculations for the period. That
                        is, although the collected price of the item is not used in the CPI that
                        month, it is still represented in the CPI through a weighted average of the
                        same type of items that are in the CPI for that month. Then in the following
                        pricing period the price that was previously set aside is used for the price
                        comparison in the next month.


                        When using the linking method, BLS makes the implicit assumption that the
How BLS Accounts        pure price change is an amount that corresponds to the rate of price
for Quality Change in   change that occurred for the same type of items in the same geographic
Using the Linking       area. In other words, BLS assumes that the pure price change is the price
                        relative calculated through the linking method. BLS assumes that any
Method                  remaining difference in price—the residual—reflects differences in quality
                        between the old version and its new version. This residual is excluded
                        because the CPI is designed to reflect only true changes in price.

                        BLS’exclusion of quality in the linking method can be demonstrated using
                        example 1 at the end of this appendix. There was a difference of $0.128 in
                        the effective (per ounce) price of the two soups.3 If the commodity analyst
                        had determined that the two soups were comparable, a 34.5 percent

                        3
                         To make reading the appendix easier, we rounded the effective price per ounce amounts. The
                        numbers we used for rounding appeared on the listing that the commodity analysts use to review the
                        substitutions. For instance, the difference in effective price for these two soups was $0.12760—the
                        price per ounce of the new version ($0.49737) minus the price per ounce of the old version ($0.36977).



                        Page 90                                                     GAO/GGD-99-84 Consumer Price Index
                        Appendix VI
                        Linking Method of Adjustment




                        increase would have entered into the CPI as a pure price change. Or, if the
                        commodity analyst had determined that the soups were not comparable
                        and the $0.128 difference between the soups was entirely due to better
                        ingredients (quality improvement), a direct adjustment would be made. If
                        under this adjustment the entire difference in price was deemed to have
                        resulted from better ingredients, no price increase would have entered
                        into the CPI. But neither of these events occurred.

                        Instead, the two soups were considered by the commodity analyst to be
                        not comparable because there was a weight change and a change in
                        ingredients between the two versions and through the linking method a
                        price relative increase of 0.58 percent was computed. To conceptually
                        allocate the $0.128 difference to pure price and quality, BLS first applies the
                        rate of change for this soup (0.58 percent) to the effective price of the old
                        version ($0.370) to obtain an imputed price ($0.372). Then by subtracting
                        the price of the old version from its imputed price BLS obtains the change
                        in pure price ($0.372 - $0.370 = $0.002). BLS then assumes that the residual
                        ($0.128 –$0.002 = $0.126) is a quality increase. According to BLS, $0.002 is
                        included in the CPI as a price increase; whereas the $0.126 is excluded from
                        the CPI because it is the quality difference between the two soups.


                        The following two examples of BLS’ use of the linking method come from
Examples of BLS’ Use    our interviews with two BLS commodity analysts. According to the
of the Linking Method   commodity analysts and supervisors, these examples are illustrative of the
                        substitutions in which they use the linking method in their respective CPI
                        components. These examples show analysts’ decisions when the
                        replacement is extremely different from the old version and is viewed as a
                        dissimilar item.

                        Each example includes a price calculation that shows the percentage of
                        changes in price that went into the CPI for that month as a result of using
                        the linking method. They are applicable only to the particular cases
                        illustrated. They are not intended to be representative of the percentage of
                        changes that occurred for similar kinds of substitutions in which the
                        linking method was used. Nor is the percentage change intended to be
                        representative of the impact that the linking method has on the CPI.

                        Each example includes (1) a table showing the characteristics of the old
                        version and the new version that replaced it, (2) the analyst’s reasons for
                        judging the two versions to be not comparable, (3) the rationale for
                        selecting the method of adjustment, and (4) the calculation of price



                        Page 91                                       GAO/GGD-99-84 Consumer Price Index
                                     Appendix VI
                                     Linking Method of Adjustment




                                     change that was made. Within each table, differences in specifications
                                     between the old and new versions are highlighted by shading.


                                     A new version of packaged soup was substituted for the old version and,
Example 1 - Packaged                 as shown in table VI.1, certain characteristics of the two versions differed.
Soup (Under Canned                   The analyst concluded that the two versions were not comparable and
and Packaged Soup                    designated linking as the method of adjustment.

Item Stratum)
Table VI.1: Characteristics of Two
Versions of a Packaged Soup          Description                        Old version                      New version
                                     Specifications
                                      Where purchased (in               Shelf                            Shelf
                                      store)
                                      Type                              Dried                            Dried
                                      Brand classification              Nationally marketed brand        Nationally marketed brand
                                      Flavor                            Chicken                          Chicken
                                      Other major ingredients           Pasta and beans                  Pasta and herbs
                                      Packaging                         Envelope                         Envelope
                                      Weight                            4.3 oz.                          3.8 oz.
                                      Other features                    Just add water                   Just add water
                                                                        Cooks in about 15 minutes        Cooks in about 15 minutes
                                      ZZ99a                             Chicken flavor with pasta        New country chicken with
                                                                        and beans                        pasta and herbs
                                     Price information
                                      Amount                            $1.59                            $1.89
                                      As of                             October                          November
                                      Collected/imputed                 Collected                        Collected
                                      Effective price per ounce         $0.36977                         $0.49737

                                              Specifications that differ between old and new versions


                                     a
                                      A designated line for the price taker to put information that may not belong elsewhere on the
                                     checklist.


                                     Source: BLS.




Analyst’s Reasons for                The analyst said that the change in “other” major ingredients from beans to
Deciding Items Were Not              herbs indicated a change in the item. According to the analyst, this was
Comparable

                                     Page 92                                                     GAO/GGD-99-84 Consumer Price Index
                              Appendix VI
                              Linking Method of Adjustment




                              confirmed in the “ZZ99” specification, which reiterated that the ingredients
                              had changed from beans to herbs.

                              A weight change also indicated to the analyst that these versions were not
                              comparable. The commodity analyst said that a decrease from 4.3 to 3.8
                              ounces suggested a change from one standard size to another and that
                              made the versions not comparable. The comparability criteria that BLS
                              established for this item stratum allow the analyst to exercise judgment in
                              determining if the change in weight is substantial enough to make the
                              version not comparable.

                              The comparability criteria for the item stratum, to which the analyst
                              referred in making the comparability decision, indicated whether a change
                              in various specifications between the old and the versions would make the
                              versions not comparable. According to the criteria, a substantial change in
                              either weight or “other” major ingredients would make the versions not
                              comparable.


Rationale for Selecting the   Because the class-mean method had not been designated for the item
Method of Adjustment          strata and adequate information needed to make a direct adjustment was
                              not available, the analyst chose the linking method. To designate the
                              linking method in this case, the analyst had to override another adjustment
                              code that was computer generated. A computer routine automatically
                              calculates any change in the effective prices (the unit prices) between the
                              old and new versions. When there is a change that meets or exceeds
                              predetermined amounts, the computer is programmed to use particular
                              comparison codes. In this case, the computer defaulted to a comparable
                              comparison code signifying that the old and new versions were alike, and
                              the computer generated a price increase of 34.5 percent. However,
                              because the analyst judged the old and new versions to be not comparable,
                              the analyst overrode this comparison code.


Calculation of Price          The calculation of price change for the linking method is done entirely by
Change                        computer routine without direct involvement by the commodity analyst.
                              The method calculates a rate of price change based on the price movement
                              of usable collected prices in the item-stratum index for the area. For the
                              canned and packaged soup item stratum in this soup’s geographic
                              location, a 0.58 percent rate of change was calculated for November 1997.
                              This means that the CPI in November 1997 reflected the same imputed
                              increase (0.58 percent) for this package of chicken soup.



                              Page 93                                     GAO/GGD-99-84 Consumer Price Index
                                     Appendix VI
                                     Linking Method of Adjustment




                                     A new version of multivitamin tablets was substituted for the old version
Example 2 -                          and, as shown in table VI.2, certain characteristics of the two versions
Multivitamin Tablets                 differed. The analyst concluded that the old and new version were not
(Under Internal,                     comparable and designated the linking method to adjust the price of the
                                     old version.
Respiratory, and
Over-the-Counter
Drugs Item Stratum)

Table VI.2: Characteristics of Two
Versions of Multivitamin Tablets     Description                           Old version                  New version
                                     Specifications
                                      Type                                 Vitamins                     Vitamins
                                      Form                                 Plain tablets                Plain tablets
                                      Brand                                Store brand                  Store brand
                                      Size                                 130 tablets                  100 tablets
                                      Units of vitamin C, per tablet       60                           60
                                      Units of vitamin D, per tablet       400                          400
                                      Units of vitamin E, per tablet       30 international units       45 international units
                                      Labeling                             "one daily plus minerals"    "for mature adults"
                                     Price information
                                      Amount                               $6.49                        $6.49
                                      As of                                September                    October
                                      Collected/imputed                    Collected                    Collected

                                              Specifications that differ between old and new versions



                                     Source: BLS.




Analyst’s Reasons for                The commodity analyst decided that the two versions were not
Deciding Items Were Not              comparable because there were so many important differences between
Comparable                           their specifications. These differences included the number of tablets in
                                     the bottles, the potency of vitamin E, and the labeling of the bottles.


Rationale for Selecting the          For substitutions that are not comparable in this item stratum, BLS
Method of Adjustment                 designated the linking method as the standard method of adjustment when
                                     other methods are not usable. In accordance with this procedure, the




                                     Page 94                                                     GAO/GGD-99-84 Consumer Price Index
                       Appendix VI
                       Linking Method of Adjustment




                       analyst marked this case for adjustment by the linking method. To do so,
                       the analyst had to override a comparable comparison code that generated
                       a price change of 0.0 percent (no price change).

                       The analyst considered the possibility of using an adjustment method to
                       directly adjust the price of the old version rather than using the linking
                       method. According to the analyst, a direct adjustment would have been
                       made to the price of the old version if the only change had been the
                       number of tablets. (This would have been done by computing the cost per
                       tablet of 130 tablets and then multiplying that cost by 100 tablets.)
                       However, according to the analyst, the changes in vitamin E potency and
                       label information prevented a direct adjustment because there was no
                       sensible way to directly adjust the price of the old version for these
                       changes. Also, because a class-mean method had not been designated for
                       the item stratum, this method of adjustment could not be used. The analyst
                       was therefore left with only the linking method.


Calculation of Price   The calculation of price change for the linking method is done entirely by
Change                 computer routine without direct involvement by the commodity analyst.
                       The method calculates a rate of price change based on other changes in
                       the item stratum in the same geographical area. For over-the-counter
                       drugs in the same geographical area, a minus 0.58 percent rate of change
                       was calculated for October 1997. The CPI in October 1997 effectively
                       reflects the same percentage decrease for this item, even though the old
                       version of vitamins has been replaced by a new, not comparable bottle of
                       vitamins.




                       Page 95                                     GAO/GGD-99-84 Consumer Price Index
Appendix VII

Measurement of Residential and
Homeowners’ Equivalent Rents

               The housing component of the CPI includes residential rent and
               homeowners’ equivalent rent. Residential rent measures the changes in
               rents paid by renters in the United States. Homeowners’ equivalent rent
               measures the changes in rental value of owner-occupied houses or
               apartments. BLS determines this value using a rental equivalence method,
               which estimates the amount of rent that would be paid for owner-occupied
               housing if it were rented.1 BLS refers to homeowners’ equivalent rent as
               REQ.


               In comparison to changes in the prices of other items in the CPI, price
               changes in residential rent and REQ have the greatest influence upon the
               CPI. For example, as of December 1997, the relative importance (a concept
               explained in app. II) of these two rent items combined to almost
               26 percent (19.9 percent for REQ and 5.8 percent for residential rent). This
               means that changes in rent directly affected slightly more than one-quarter
               of all of the price changes that were measured in the CPI in 1997. No other
               item in the CPI has as much influence as either residential or homeowners
               equivalent rent items (new cars was the next item in importance in 1997 at
               less than 4 percent).

               BLS collects and processes residential rent and REQ in several ways that are
               different from other items in the CPI, including those in the rest of the
               housing component. For example, the selection of housing units from
               which rent prices are obtained is done using a survey that is separate from
               the survey used to select other housing and CPI items. This appendix
               provides information about BLS’ procedures and practices for residential
               rent and REQ.


               The CPI housing survey is used to estimate changes in rent for both renters
Background     and home owners if they were to pay rent for the use of their homes. The
               CPI housing survey sample is made up of approximately 36,000 rental units
               and 26,000 owner units. The housing units in the sample were selected
               from two sources: the 1980 Decennial Census for units built before 1980
               and the U.S. Bureau of Census’ sample of building permits for units built




               1
                The rental equivalence method seeks to measure the costs of consuming housing services over time
               rather than the value of housing as an asset that might appreciate over time. The latter approach was
               used before 1983. For further information about this earlier method, see Consumer Price Index:
               Cost-of-Living Concepts and the Housing and Medical Care Components (GAO/GGD-96-166, Aug. 26,
               1996).



               Page 96                                                     GAO/GGD-99-84 Consumer Price Index
    Appendix VII
    Measurement of Residential and
    Homeowners’ Equivalent Rents




    after 1980.2 A sample of renters has been chosen from 85 geographic areas,
    which BLS refers to as primary sampling units (PSU).

    The CPI uses the rental units in the CPI housing survey to determine the
    price change for both residential rent and REQ. In the rental equivalence
    method that is used for REQ, rental units are matched to owner units in the
    CPI housing survey, and the change in the rents reported for those rental
    units is used to adjust the REQ for the matching homeowner units. In
    matching rental units with homeowner units, BLS takes into account their
    locations (must be in the same PSU), structure type (e.g., single-family
    dwellings), and whether the units have air conditioning.

    The basis for determining REQ, however, is different from that used for
    residential rent in one important way. The rental value of homeowner
    units excludes utilities; the rents of residential renters include any utilities
    paid by landlords. A BLS official said that approximately 20 percent of the
    rental units have their utilities provided by the landlord. Therefore,
    according to the official, the cost of landlord-provided utilities is
    subtracted from the rent whenever it is used to calculate REQ.

    BLS divides the rental units in the housing survey into six groups, with the
    units in each group to be contacted twice a year. For example, units in one
    group are priced in January and again 6 months later in July; units in a
    second group are contacted in February and again in August. BLS price
    takers are to record basic information about the rental unit, any extra
    charges paid for the unit, rent subsidies, and other information, such as
    any familial relationship between the renter and the landlord.

    After the data have been collected by the price takers and entered into BLS’
    computer system, CRLs are generated for review by commodity analysts.
    CRLs are computer printouts of data for the units that have certain
    problems, inconsistencies, or changes requiring a commodity analyst’s
    attention. Current and historical data are printed for each unit as well as
    the reason why a CRL was triggered.

    The reasons that trigger the printing of a CRL include

•   rent increases or decreases of 20 percent or more;




    2
    In January 1999, BLS revised the CPI housing survey with a sample of housing units based on the 1990
    Decennial Census and will update the sample with building permits for units built after 1990.



    Page 97                                                   GAO/GGD-99-84 Consumer Price Index
                       Appendix VII
                       Measurement of Residential and
                       Homeowners’ Equivalent Rents




                   •   a message from the price taker, such as the rent increased because a new
                       tenant moved in, or the rent did not go up as much for one tenant as for
                       other tenants because of the number of years the tenant had been renting;
                   •   inconsistencies, such as a rental unit with “owner” data reported for it;
                   •   a change in housing tenure, such as from renter to owner occupied;
                   •   dollar values reported to questions in the survey instrument about free
                       rent, rent reductions, or any extra charges;
                   •   changes in utilities;
                   •   any differences in house trailers, such as a new trailer on the same lot;
                   •   structural changes, such as an addition of a bathroom; and
                   •   anything that the commodity analyst has asked to be programmed for
                       review in order to follow up on an unusual circumstance.

                       The housing commodity analysts use the CRLs to (1) review the units’
                       eligibility for use in the CPI, (2) determine whether adjustments should be
                       made for changes in the units, (3) identify problems that the price takers
                       have in locating units or answering the questions on the CPI housing survey
                       questionnaire, and (4) communicate with the price takers, (e.g., follow up
                       on a particular issue). After the analysts have reviewed the CRLs and
                       approved the information or made corrections, adjustments, or changes,
                       the units are ready to be used in calculating the CPI.

                       The CPI measures rent price changes by comparing the rents tenants pay
                       with the rent they paid 6 months earlier. Before the comparison is made,
                       any necessary adjustments to the current rent for the month are to be
                       made. Some adjustments are automatically identified and made during
                       computer processing of the data. Others are identified by commodity
                       analysts.


                       Adjustments are made so that the same unit with the same features is
Adjustments for        priced each time (every 6 months). If a feature of the rental unit has
Changes in Units       changed, BLS makes a dollar adjustment to the current rent so the unit is
                       comparable to what it was earlier. This adjustment is for that month and is
                       not retained for future comparisons.

                       BLS does not make adjustments for ordinary maintenance (e.g., painting or
                       replacing an appliance) as these mostly restore the unit to its earlier
                       condition. However, when the landlord provides an appliance or service,
                       BLS lowers the current rent because the cost to live in the unit has gone
                       down. This situation would occur, for example, if the landlord installed a




                       Page 98                                    GAO/GGD-99-84 Consumer Price Index
Appendix VII
Measurement of Residential and
Homeowners’ Equivalent Rents




clothes washer and dryer in the unit. The tenant would no longer have to
use a laundromat (or privately owned washer and dryer).

However, if the landlord stops providing an appliance or service, BLS raises
the current rent because the cost to live in the unit has gone up. This
situation would occur, for example, if the rent for a unit no longer
included the cost of electricity and the tenant had to begin paying for it
separately.

BLS makes these adjustments even when the landlord adds or removes a
feature but does not change the rent. It also makes these adjustments
when the landlord raises the rent for adding an appliance or service or
lowers the rent for removing something. In all of these cases, BLS is
attempting to make the current unit comparable to what it was 6 months
earlier when data were last collected. (This is the reverse of the
adjustments made elsewhere in the CPI where adjustments are made to the
previous version to make it comparable to the current version of the
replacement.)

Any adjustments made to the current unit to make it comparable with
what it was 6 months earlier are dropped after the comparison is made.
Whatever features the current unit now includes become the basis against
which the unit will be compared to 6 months in the future. For example, if
an appliance were newly installed, BLS would adjust current rent, in effect,
to remove the appliance. However, the data records for the unit would
include this appliance for future comparisons because the renter’s cost of
living in the unit will have decreased.

BLS makes adjustments for changes to features of the rental unit; it does
not make adjustments for changes in the tenants. For example, the
landlord may charge a pet fee if a tenant has a pet and the previous tenants
paid a fee. BLS would not show a decrease in rent if the new tenants did not
have a pet and did not pay the fee. However, if the landlord stopped
charging pet fees altogether, this decrease in rent would appear in the CPI.

In summary, if the value of the unit has gone up since data were last
collected, a subtraction to the current month’s rent is made. If it is costing
more for the tenants to live there (by getting less than before), an addition
to the current month’s rent is made.




Page 99                                       GAO/GGD-99-84 Consumer Price Index
                           Appendix VII
                           Measurement of Residential and
                           Homeowners’ Equivalent Rents




                           In calendar year 1997, approximately 6,180 adjustments were made
Adjustments Are            because of changes in rental units. Nearly all (91 percent) of the
Made by Computer           adjustments were made automatically by computer routine without direct
and by Analysts            involvement of the housing commodity analysts. The other adjustments
                           (9 percent) were made by housing commodity analysts. In addition to
                           these adjustments, every rental unit is adjusted by computer routine for
                           age depreciation.


Automatic Adjustments by   The rent adjustments that are automatically identified and made by
Computer                   computer routine involve only direct adjustments. That is, the item or
                           service in the housing unit that changed is clearly identifiable, and its cost
                           can be reasonably calculated. According to BLS, automation of these direct
                           adjustments saves the time and efforts of commodity analysts and enables
                           calculations and adjustments to be made on a consistent basis. Computer
                           routines adjust the rent automatically whenever the rental unit has a
                           change in facilities, utilities, or structure.

Changes in Facilities      Prior to January 1999, facilities adjustments included additions or
                           subtractions of dollar amounts for changes in the provision of appliances
                           and parking. To determine the dollar amounts for changes in appliances,
                           BLS used studies published in trade journals. The published average cost of
                           the appliance was divided by its published average life in months to arrive
                           at its monthly value. Example 1 at the end of this appendix provides an
                           illustration of such an automatic adjustment.

                           Off-street parking costs are based on the PSU’s average of such charges
                           paid by tenants in the CPI housing survey. This adjustment is made, for
                           example, if off-street parking is no longer provided by the landlord.
                           According to BLS, adjustments are no longer made for changes in
                           furnishings because apartments, BLS concluded, do not change in the
                           provision of furnishings. They tend to stay furnished or unfurnished.

                           According to a BLS official, the data BLS used to make appliance
                           adjustments are not up to date, and there is no cost-effective way to
                           update them. He said that since the revision of the housing component in
                           January 1999, BLS makes facilities adjustments for changes only in parking
                           and air conditioning equipment. Since the majority of automatic direct
                           adjustments are for changes in the provision of appliances, this will reduce
                           the number of direct adjustments made to rental units.




                           Page 100                                     GAO/GGD-99-84 Consumer Price Index
                              Appendix VII
                              Measurement of Residential and
                              Homeowners’ Equivalent Rents




Changes in Utilities          Utility adjustments are made when the landlord alters the provision of a
                              utility, such as installing a separate meter for an apartment and making the
                              tenant responsible for payment of the utility. BLS calculates the dollar
                              amount of utility adjustments in two steps. First, BLS uses Department of
                              Energy data to determine average consumption amounts by housing unit
                              for electricity, gas, propane, and oil. These estimated amounts are based
                              on the rental unit’s location, number of heating and cooling degree days,
                              number of rooms, and type of structure. Second, to compute the monthly
                              cost, BLS multiplies the consumption amounts by the average price of
                              electricity, gas, natural gas, or oil paid by households in the unit’s PSU. The
                              average price is based on data BLS collects monthly for the CPI.3

                              BLScalculates water and sewer utility adjustment dollar amounts by
                              determining average bills from housing units in the CPI housing survey for
                              each PSU. The survey gathers data from renters on the amount of their
                              water and sewer bills.

                              As previously noted in this appendix, utility adjustments are only made to
                              the residential rental units. No adjustment is made to REQ because it
                              already excludes the cost of utilities.

Depreciation and Changes in   The dollar amounts for various structural changes (e.g., the addition of a
Structure                     bedroom or bathroom) and depreciation are based on estimates that come
                              from regression-based formulas. The formulas account for the age of the
                              unit and a number of structural characteristics. The dollar amounts from
                              these formulas are then used to adjust residential rent and REQ for
                              structural changes in the rental units. BLS recomputes these estimates
                              annually.

                              Regression-based formulas have been used since January 1988 to account
                              for the small loss in quality as housing units age (depreciate) over time.
                              According to a BLS official, the age-bias regression is recomputed every
                              year, and one-twelfth of the annual bias is applied by computer routine to
                              every unit in the housing sample each month.


Adjustments by                The housing commodity analysts make three types of adjustments:
Commodity Analysts            “pricing-links-cancel,” “links-pause,” and direct. Of the three types, the
                              links-pause adjustment is by far the most frequently used. About


                              3
                               BLS makes approximations of average costs for wood and coal heat by converting the standard
                              amount of energy from a cord of wood or ton of coal into gallons of oil, and then using the average
                              consumption and cost estimations for oil.



                              Page 101                                                    GAO/GGD-99-84 Consumer Price Index
                         Appendix VII
                         Measurement of Residential and
                         Homeowners’ Equivalent Rents




                         80 percent of the adjustments made by commodity analysts in 1997 were
                         links-pause adjustments.

Pricing-Links-Cancel     In the pricing-links-cancel adjustment, the analyst determines that the unit
Adjustment               is no longer comparable to the unit for which the rent was collected 6
                         months earlier or that the old rent is no longer accurate for comparison.
                         Examples of units that, according to BLS, are no longer comparable
                         include: a house trailer that is replaced with a different trailer, a unit in
                         which the price taker discovers that BLS has been pricing the wrong unit,
                         or a unit in which a major structural change has occurred (e.g., the
                         addition of a swimming pool) and BLS does not have a way to make an
                         adjustment.

                         Examples cited by BLS of units for which the old rent is no longer accurate
                         for comparison include: (1) a unit that the price taker recorded as
                         government subsidized and for which the renter was not paying the full
                         rent that was recorded 6 months earlier or (2) a unit in which the renter
                         was related to the landlord and was not paying the full rent that was
                         recorded 6 months earlier. Example 2 at the end of this appendix provides
                         an illustration of a pricing-links-cancel adjustment for a rental unit whose
                         old rent had been imputed because of a long-term vacancy and was
                         therefore no longer appropriate for a price comparison.

                         The pricing-links-cancel adjustment treats the rental unit in the same way
                         that the linking adjustment treats commodities and services substitutions
                         that are not comparable (see app. VI). In the pricing-links-cancel
                         adjustment, the weight of the rental unit is redistributed to housing units
                         in the PSU of that rental unit. To do this redistribution, BLS uses the rental
                         units in the PSU of the pricing-links-cancel that have usable prices—units
                         that had a rent 6 months earlier from which a price change could be
                         calculated—and a computer routine then calculates the average
                         percentage change of these rents for that PSU. In effect, BLS assigns the
                         PSU’s average percentage change to the unit that is adjusted by the
                         pricing-links-cancel method.

Links-Pause Adjustment   In the links-pause adjustment, the commodity analyst is suspicious of the
                         accuracy of the current rent. In this adjustment, the current month’s rent is
                         not used in the CPI that month. The current month’s rent is replaced with
                         an imputed value that is based on rent changes of a subset of housing units
                         in the unit’s geographic location. In the next pricing cycle, the analyst will
                         determine the accuracy of the reported rent.




                         Page 102                                      GAO/GGD-99-84 Consumer Price Index
Appendix VII
Measurement of Residential and
Homeowners’ Equivalent Rents




Until confirmed, the unit’s rent that was collected 6 months earlier is
adjusted by computer routine. First, the computer calculates the average
change in rent for units in a subset of housing units, which BLS refers to as
a cell. All rental units are assigned to 1 of 18 cells in a PSU. These cells
contain units that have similar characteristics based on the renter-owner
ratio (e.g., the unit is in a location that has equal proportions of owners
and renters) and on the level of rent paid (high, medium, and low). Next,
the computer routine then applies the cell’s rate of change to the unit’s
rent that was collected 6 months earlier. Any difference from a
comparison of the imputed rent and the rent from 6 months earlier goes
into computing the CPI for the current month.

Six months later, the analyst will use rent reported in that period to
confirm the previous rent that the analyst thought was suspicious. If the
two rents are dissimilar (that is, the suspicious rent is not confirmed and
the rent 6 months later is similar to what it was a year earlier), the
computer routine will compare the rent that was imputed 6 months earlier
with the rent most recently collected. Alternatively, if the two rents are
similar (that is, the previous rent is confirmed), then the commodity
analyst will allow the large price change to enter the CPI at this time.

A links-pause adjustment is shown in example 3 where the unit in the
Urbantown PSU had a May 1997 rent of $513 and a November 1997 rent of
$158. The analyst questioned the accuracy of the current month’s
(November) rent because the information about the unit did not indicate a
reason for such a decrease. As a result, the analyst decided to make a
links-pause adjustment. Using the units with usable rents for November
1997 in that unit’s cell, the computer routine calculated an average price
increase of 3.3 percent. Next, the computer applied that increase to the
unit’s May 1997 rent and imputed a rent of $530.208 ($513 x 1.033). In this
example the CPI calculations for November 1997 reflected a 3.3 percent
increase for this rental unit and the additional rental and owner-occupied
units it represents.

As further explained in example 3, BLS collected rent data on the unit again
in May 1998 and generated a CRL for the unit for analyst review. The
analyst used the May 1998 information about the unit to confirm the
accuracy of the reported November 1997 rent. The information indicated
that the tenancy of the unit changed from renter occupied to owner
occupied. Since BLS does not use owner-occupied units in the CPI
calculations, the unit was excluded from CPI calculations.




Page 103                                     GAO/GGD-99-84 Consumer Price Index
                        Appendix VII
                        Measurement of Residential and
                        Homeowners’ Equivalent Rents




                        According to BLS officials, for the units where the links-pause adjustment
                        is used, the price increase (or decrease) occurs but with a delay of 6
                        months (or, as in our example, does not occur). Nonetheless, BLS said the
                        adjustment prevents large increases or decreases if the rent 6 months
                        earlier was incorrect. They noted that the size of the inaccuracies during
                        the lag is smaller than the errors that would be incorporated into the CPI if
                        erroneous rents were used.

Direct Adjustment       According to a BLS official, the housing analysts also directly adjust a
                        negligible number of units for changes in extra charges that result from
                        unusual situations. He said these instances are the only occasion in which
                        a commodity analyst directly enters in a dollar adjustment in the CPI. For
                        example, a renter may have been paying an additional monthly fee to have
                        a pet living in the apartment. If the pet dies, the renter no longer pays the
                        pet fee. In these situations, the analyst will look in the historical data for
                        the unit to determine the amount of the fee and add the extra charge to the
                        current rent to make it comparable to the rent of 6 months earlier.


                        The sample of housing units that BLS uses in calculating the CPI is not static
Addition and Deletion   and represents the housing stock of the urban population. Rental units are
of Housing Units        removed from the sample, and new units are added to the sample.
                        According to a BLS official, rental units are removed when they no longer
                        exist because they represent units that drop out of the housing stock. New
                        rental units are added through new construction. According to BLS, the
                        new units are not added to replace units that were removed; the new units
                        are not considered to be substitutions, as would be the case for
                        replacement of other nonrent items in the CPI.

                        In 1997, according to a BLS official, 14 rental units were removed from the
                        housing survey because they no longer existed. Another official
                        determined that about 15 percent of the rental units in the housing survey
                        in 1997 entered the survey from new construction permits.

                        Rental units are lost to the housing sample for various reasons, such as
                        destruction by floods, tornadoes, and urban renewal. Whenever a rental
                        unit is lost, BLS does not replace it with another unit. In these instances, BLS
                        removes the unit and, implicitly, the residential rental and REQ units it
                        represents, from the computer system. However, according to BLS, if the
                        building foundation for the lost unit still exists, the address is retained in
                        the system and price takers are periodically sent to the address to see if
                        the unit has been rebuilt.



                        Page 104                                      GAO/GGD-99-84 Consumer Price Index
                 Appendix VII
                 Measurement of Residential and
                 Homeowners’ Equivalent Rents




                 Units are added to the CPI housing survey by using information that is
                 collected by the Bureau of the Census. New construction permit
                 information is obtained by Census from the appropriate government
                 entities. Census then samples from that population of permits and sends
                 the addresses of these new permits to BLS. BLS then adds the addresses to
                 the housing sample with representation (weighting) for both residential
                 rent and REQ in the same manner that it subtracts units that have been
                 destroyed. Once the address is added, BLS continues to survey that
                 location.


                 Whenever a substitution is not comparable to the CPI item it replaced, BLS
Accounting for   will make an adjustment to separate pure price change from price changes
Quality Change   that are due to other factors, such as differences in quality. (BLS refers to
                 these other factors under the general term quality.) Although BLS has
                 accounted for this separation when making adjustments to nonrent items,
                 it has not accounted for such a separation for the types of adjustments
                 described in this appendix.

                 Nonetheless, we believe a case can be made for attributing part of a
                 change in rent to price and another part to quality. We draw this view from
                 two sets of similarities that exist between the types of adjustments BLS
                 makes for rent and the adjustment methods it uses for nonrent items. One
                 set of similarities involve rent adjustments made automatically by
                 computer routine and the direct adjustment methods used for nonrent
                 items. In order for these automatic adjustments to be made, there must be
                 sufficient information about the change to a rental unit and its associated
                 cost to directly adjust the rent. This is the case for the direct adjustments
                 used elsewhere in the CPI; there must be sufficient information to directly
                 adjust the price of the item. If BLS were to consider these adjustments to
                 rent as direct adjustments, then the assumptions used to account for price
                 and quality with the direct adjustment method would likely apply as well.
                 As such, the implicit assumption would be that the entire amount of the
                 adjustment could be accounted for as quality change. For example, in
                 example 1 at the end of this appendix, 29 cents was subtracted from the
                 unit’s rent and this could be accounted for as a change in quality.

                 The second set of similarities involves the pricing-links-cancel and the
                 links-pause adjustment methods used to adjust rent and the linking
                 method used to adjust nonrent prices. Under each method, an average rate
                 of price change is computed using other rents or other prices. Under the
                 linking method, BLS assumes that any difference between the imputed



                 Page 105                                     GAO/GGD-99-84 Consumer Price Index
                       Appendix VII
                       Measurement of Residential and
                       Homeowners’ Equivalent Rents




                       price and the price of the substitution is reflective of quality and,
                       therefore, not to be included in the CPI. If BLS were to make this same
                       assumption for the pricing-links-cancel and the links-pause adjustment
                       methods, any difference—the residual—between the unit’s recorded rent
                       and the imputed rent could reflect quality differences between the current
                       unit and what was recorded for it 6 months previously.

                       As previously stated, BLS has not made this accounting to price and quality
                       for residential rent and REQ, which represent more than one-quarter of the
                       CPI.



                       The following three examples of BLS’ adjustments to residential rent and
Examples of BLS’       REQ come from our interviews with three BLS housing commodity analysts.
Adjustments for        According to the commodity analysts and their section chief, these are
Residential Rent and   examples of the adjustments that are made in residential rent and REQ. One
                       example is of an adjustment made by computer routine without direct
REQ                    involvement of the housing commodity analysts. The other examples are
                       of analysts’ decisions to use two variations of the linking method.
                       Generally, these are used when a rental unit is no longer comparable to
                       the same unit for which the rent was collected 6 months earlier. In the
                       table for each example, changes between the current unit and its
                       characteristics 6 months earlier are shaded.

                       The examples include a calculation that shows the percentage changes in
                       price that went into the CPI for that month as a result of the adjustments.
                       These percentages are applicable only to the particular cases illustrated.
                       They are not intended to be representative of the percentage changes that
                       occurred for similar kinds of adjustments that are made. Nor is the
                       percentage change intended to be representative of the impact that these
                       adjustments have on the CPI.


                       As shown in table VII.1, the landlord made changes in the appliances
Example 1 -            furnished in the apartment. To account for the changes in the unit’s
Automatic Adjustment   appliances a direct adjustment was made by computer routine without the
                       direct involvement of a commodity analyst.




                       Page 106                                    GAO/GGD-99-84 Consumer Price Index
                                      Appendix VII
                                      Measurement of Residential and
                                      Homeowners’ Equivalent Rents




Table VII.1: Characteristics of the
Rental Unit                           Description                        6 months earlier                Current
                                      Specifications
                                       Number of bedrooms                2                               2
                                       Number of bathrooms               1                               1
                                       Utilities included in rent        No                              No
                                       Heating fuel                      Oil                             Oil
                                       Hot water fuel                    Oil                             Electricity
                                       Dishwasher                        Yes, supplied by landlord       None
                                       Washing machine                   None                            Yes, supplied by landlord
                                       Dryer                             None                            Yes, supplied by landlord
                                      Price information
                                       Amount                            $500.00                         $500.00
                                       As of                             April                           October

                                               Specifications that differ between old and new versions



                                      Source: BLS.




Analyst’s Reasons for                 BLS has developed computer routines to identify changes in a rental unit’s
Not-Comparable Decision               appliances and, when identified, to calculate with preprogrammed data the
                                      dollar value associated with those changes. Commodity analysts are not
                                      involved in making these adjustment decisions.


Rationale for Selecting the           Selecting a method of adjustment was not necessary.
Method of Adjustment
Calculation of Price                  According to BLS, a computer routine automatically added 50 cents to the
Change                                unit’s current rent for the removal of the dishwasher and subtracted 79
                                      cents for the addition of a washing machine and clothes dryer. Since the
                                      landlord did not provide the utilities, a utility adjustment was not made by
                                      the computer. The computer then compared the adjusted current rent of
                                      the unit ($499.71) to the rent for the unit 6 months earlier ($500.00). The
                                      new rent was 29 cents less than the rent 6 months earlier, a decrease of
                                      0 percent. In effect, no price change was entered for this unit into the data
                                      with which the CPI was computed for October 1997.




                                      Page 107                                                    GAO/GGD-99-84 Consumer Price Index
                                      Appendix VII
                                      Measurement of Residential and
                                      Homeowners’ Equivalent Rents




                                      The rent for the unit increased significantly, as shown in table VII.2. The
Example 2 -                           commodity analyst determined that the imputed April rent for the unit was
Pricing-Links-Cancel                  not appropriate for comparison purposes and designated the
Adjustment                            pricing-links-cancel method to calculate the rate of price increase between
                                      April 1997 and October 1997.


Table VII.2: Characteristics of the
Rental Unit                           Description                     6 months earlier          Current
                                      Specifications
                                       Number of bedrooms             3                         3
                                       Number of bathrooms            1 and a half bath         1 and a half bath
                                       Landlord furnishes utilities   No                        No
                                      Price information
                                       Amount                         $271.325                  $400.00
                                       As of                          April 1997                October 1997
                                       Collected/imputed              Imputed                   Collected



                                      Source: BLS.




Analyst’s Reasons for                 A lengthy period of vacancy for the unit suggested to the commodity
Not-Comparable Decision               analyst that the rental amounts for the unit for the two collection periods
                                      should not be compared. The CRL for the unit indicated that the rent had
                                      been imputed for more than 2 years; BLS has no standard or limit on how
                                      long rent can be imputed for a unit. The CRL did not indicate the origin of
                                      the rent on which the imputations were based. The analyst said that the
                                      price comparison for the CPI should not be based on a comparison of a
                                      collected rent with an imputed rent whose origin is unknown.

                                      The analyst said another reason for not using the April rent was that the
                                      unit was in a small geographic area and that a large price increase of about
                                      150 percent (increase from $271 to $400) could cause an upward bias in
                                      the CPI for that area.


Rationale for Selecting the           The analyst first looked at the percentage change in rent to see if it
Method of Adjustment                  exceeded what BLS terms a “flinch point,” a mathematically determined
                                      percentage change rate indicating that the change in rent is doubtful and
                                      should not be used in calculating the CPI for that month. According to BLS,




                                      Page 108                                            GAO/GGD-99-84 Consumer Price Index
                       Appendix VII
                       Measurement of Residential and
                       Homeowners’ Equivalent Rents




                       the use of the flinch point is a main criterion for making a links-pause
                       adjustment—the other linking adjustment method used for rental units.
                       The analyst noted that a percentage change of about 150 percent was
                       about 3 points above the flinch point. Although this met the criterion, the
                       amount of difference did not clearly indicate that holding the unit’s
                       October rent for 6 months was appropriate, which would occur with the
                       other linking adjustment method.

                       Another factor the analyst considered was the geographic location of the
                       unit. The two-story duplex was located in a small geographic area in which
                       rental units were difficult to keep in the CPI housing survey. The analyst
                       said that it was desirable to have units with collected rent data, rather than
                       imputed rental amounts, included in CPI calculations. Therefore, the
                       analyst chose the pricing-links-cancel method so the unit would be used in
                       the next pricing period.


Calculation of Price   The calculation of price change for the pricing-links-cancel method is done
Change                 automatically by computer routine without direct involvement of the
                       commodity analyst. The method calculates a rate of price change based on
                       other units that had usable rents and were located in the unit’s geographic
                       area. For these rental units, a 0.12 percent rate of change was calculated
                       for October 1997. This means that the CPI in October 1997 reflected this
                       same percentage of increase for this rental unit and the owner-occupied
                       units it represents.


                       As shown in table VII.3, the rent for the unit increased significantly. The
Example 3 -            commodity analyst determined that the November rent for the unit could
Links-Pause            be inaccurate and designated the links-pause method to calculate the rate
Adjustment             of price increase between May 1997 and November 1997.




                       Page 109                                     GAO/GGD-99-84 Consumer Price Index
                                      Appendix VII
                                      Measurement of Residential and
                                      Homeowners’ Equivalent Rents




Table VII.3: Characteristics of the
Rental Unit                           Description                      6 months earlier          Current
                                      Specifications
                                       Number of bedrooms              1                         1
                                       Number of bathrooms             1                         1
                                       Landlord furnishes utilities    Yes                       Yes
                                      Price information
                                       Amount                          $513.27                   $157.78
                                       As of                           May 1997                  November 1997


                                      Source: BLS.




Analyst’s Reasons for                 A rent decrease of 69 percent suggested to the commodity analyst that the
Not-Comparable Decision               unit in November had changed in some major way from what was
                                      recorded for it in May and was no longer comparable.


Rationale for Selecting the           The analyst used two criteria to determine which linking method to use.
Method of Adjustment                  The first was whether the percentage change in rent exceeded the flinch
                                      point. The rent change met this first criterion. The second consideration
                                      was whether the same tenants lived there. The analyst indicated that a
                                      change in tenants might explain a large change in rent. The information on
                                      this housing unit indicated that these were the same tenants. Since both
                                      criteria were met, the analyst indicated that the links-pause method would
                                      be used for this unit.


Calculation of Price                  BLS took two actions, 6 months apart, that affected the calculations of
Change                                price change for this rental unit. For the current month (November 1997),
                                      the links-pause designation caused a computer routine to calculate a rate
                                      of price change based on a subset of other units that were similar to this
                                      rental unit. These similar units were in Urbantown and had usable rents.
                                      For the subset of units, a 3.3 percent increase was calculated for
                                      November 1997. This means that the CPI in November 1997 reflected a
                                      3.3 percent increase for this rental unit and the owner-occupied units it
                                      represents.

                                      BLS collected information on this unit again in May 1998 and, on the basis
                                      of this information, decided that no calculation should be made. BLS




                                      Page 110                                            GAO/GGD-99-84 Consumer Price Index
Appendix VII
Measurement of Residential and
Homeowners’ Equivalent Rents




learned that the tenancy for the unit changed from renter occupied to
owner occupied. Because BLS does not use owner-occupied units in
making the monthly CPI calculations, it excluded the unit from the
May 1998 CPI calculation. According to BLS, the unit will remain excluded
from the CPI as long as it is owner occupied. However, the unit would be
returned to the CPI calculation if it again becomes a rental unit.




Page 111                                    GAO/GGD-99-84 Consumer Price Index
Appendix VIII

CPI Price Quotations, Substitutions, and
Methods of Adjustment for 1997

                Table VIII.1 contains information on the relative importance and summary
                data on CPI price quotations; substitutions; and methods of adjustment by
                major component, expenditure class, item stratum, and entry level item
                (ELI) for 1997.

                While most column headings in the table are self-explanatory, we believe
                that three require some explanation. They are the columns on relative
                importance, percent of quotations with price changes, and class mean
                and/or overlap. Relative importance shows the share of total expenditures
                that would occur if consumed quantities of the items at the stratum level
                remained constant. Relative importance is a concept that is related to
                expenditure weights, but unlike those weights, which have been
                recomputed only about every 10 years or so, BLS computes relative
                importance at least annually to reflect the effect of price changes. Relative
                importance can be used to show the direct effect an item has on the
                overall CPI price change. (See app. II for further information about relative
                importance and expenditure weights.)

                The price quotations represented by the percent of quotations with price
                changes include both those where an increase in price occurred and those
                where a decrease in price occurred. Also, these changes in prices were for
                goods and services for which a substitution was made and those for which
                no substitution was made.1

                The class-mean and/or overlap column combines the number of
                adjustments made by both methods. BLS was unable to separate the
                number of each of these methods by ELI, but officials told us that nearly all
                of the substitutions listed in this column were adjusted by using the
                class-mean method. This is because BLS has been phasing out use of the
                overlap adjustment method.

                Table VIII.1 includes two overall totals: all items with residential rent units
                and all items without residential rent units. Because the adjustments made
                by BLS to residential rent differ from those made elsewhere in the CPI, we
                present the totals separately. See appendix VII for information on how
                adjustments are made to rental units. In table VIII.1 these adjustments are
                reported in the residential rent ELI under the housing component.

                Table VIII.1 is further organized as follows: The broadest of the categories
                are the seven major components. Components are indicated as totals (e.g.,

                1
                 According to BLS, this column excludes substitutions for which there was no price change between
                the two versions of the item.



                Page 112                                                  GAO/GGD-99-84 Consumer Price Index
Appendix VIII
CPI Price Quotations, Substitutions, and
Methods of Adjustment for 1997




food and beverages component totals). Under each component is the
expenditure class category, the next broadest category. Expenditure class
appears bolded in the table (e.g., cereal and cereal products). Adding all of
the numbers in the expenditure class for a given column will equal the
major component total. Under expenditure class are the item strata, which
when totaled in a given column will equal the reported subtotals for
expenditure class. For example, under the expenditure class of cereal and
cereal products the item strata are “flour and prepared flour mixes”;
“cereal”; and “rice, pasta, and cornmeal.”

The final category, as well as the most specific, is the ELI. For example,
under the item stratum of flour and prepared flour mixes are the two ELIs,
“flour” and “prepared flour mixes.” In some cases, only one ELI existed for
a single item stratum. When the item stratum and the ELI were the same
and the stratum was not broken down into additional ELIs, we deleted the
ELI from the table. By doing so, we avoided duplication of information,
since the numbers for both the stratum and the ELI were identical. ELIs do
not have relative importance assigned to them; and, where noted in the
table, some ELIs are not priced.




Page 113                                     GAO/GGD-99-84 Consumer Price Index
                                             Appendix VIII
                                             CPI Price Quotations, Substitutions, and
                                             Methods of Adjustment for 1997




Table VIII.1: Relative Importance and Summary Data on CPI Price Quotations, Substitutions, and Methods of Adjustment by
Categories for 1997


Categories (name of component,                        Relative                                    Percent of
expenditure class, item stratum, and              importance          Number of price             quotations          Number of
ELI)                                            for December              quotations     with price changes         substitutions
All items (with residential rent units)                 100.000%               918,561                   n/a              35,061
All items (without residential rent units)              74.309                 872,829                   n/a              28,881
  Food and beverages component
  total                                                 17.465                 459,635                  n/a                6,485
     Cereal and cereal products                          0.449                  19,574                 25.0%                 279
        Flour and prepared flour mixes                   0.075                   6,507                 27.0                   90
                                                               a
          Flour                                                                  2,027                 29.0                   14
                                                               a
          Prepared flour mixes                                                   4,480                 26.1                   76
        Cereal                                           0.272                   6,516                 23.8                  116
        Rice, pasta, and cornmeal                        0.102                   6,551                 24.4                   73
                                                               a
          Rice                                                                   3,085                 23.2                   37
          Macaroni, similar products,
                                                               a
          and cornmeal                                                           3,466                 25.4                   36
     Bakery products                                     1.027                  24,431                 26.2                  494
        White bread                                      0.260                   6,249                 23.3                   78
        Other breads, rolls, biscuits, and
        muffins                                          0.239                   6,124                 24.3                  142
                                                               a
          Bread other than white                                                 3,538                 27.8                   70
          Rolls, biscuits, and muffins,
                                                               a
          excluding frozen                                                       2,586                 19.5                   72
        Cakes, cupcakes, and cookies                     0.249                   6,274                 27.7                  141
          Cakes and cupcakes,
                                                               a
          excluding frozen                                                       2,433                 19.5                   64
                                                               a
          Cookies                                                                3,841                 32.9                   77
        Other bakery products                            0.279                   5,784                 29.9                  133
                                                               a
          Crackers                                                               2,279                 39.7                   27
                                                               a
          Bread and cracker products                                               140                 25.7                    3
          Sweetrolls, coffee cake, and
                                                               a
          doughnuts, excluding frozen                                            1,404                 18.1                   47
          Frozen bakery products,
          frozen and refrigerated
                                                               a
          doughs, and batters                                                    1,241                 26.8                   22
          Pies, tarts, and turnovers,
                                                               a
          excluding frozen                                                         720                 27.9                   34
     Beef and veal                                       0.933                  47,442                 46.4                  312
        Ground beef                                      0.309                   6,302                 43.0                   35




                                             Page 114                                          GAO/GGD-99-84 Consumer Price Index
                                       Appendix VIII
                                       CPI Price Quotations, Substitutions, and
                                       Methods of Adjustment for 1997




                        Substitutions not adjusted
    Percent of price          (comparable)                                    Adjusted substitutions by method
quotations that were    Number not          Percent of          Number of                         Class mean
       substitutions      adjusted        substitutions       adjustments              Direct   and/or overlap        Linking
                 3.8%            n/a                  n/a           18,311              9,411           4,049           4,851
                 3.3          16,750                58.0%           12,131              3,770           4,049           4,312

                 1.4           3,640                56.1             2,845                107              31           2,707
                 1.4            163                 58.4               116                  6               2             108
                 1.4              57                63.3                33                  2               1              30
                 0.7               9                64.3                  5                 0               1               4
                 1.7              48                63.2                28                  2               0              26
                 1.8              65                56.0                51                  3               1              47
                 1.1              41                56.2                32                  1               0              31
                 1.2              23                62.2                14                  0               0              14

                 1.0              18                50.0                18                  1               0              17
                 2.0            229                 46.4               265                 15               1             249
                 1.3              45                57.7                33                  2               0              31

                 2.3              67                47.2                75                  2               0              73
                 2.0              29                41.4                41                  0               0              41

                 2.8              38                52.8                34                  2               0              32
                 2.3              48                34.0                93                  4               0              89

                 2.6              17                26.6                47                  0               0              47
                 2.0              31                40.3                46                  4               0              42
                 2.3              69                51.9                64                  7               1              56
                 1.2              17                63.0                10                  1               0               9
                 2.1               0                  0.0                 3                 0               0               3

                 3.4              19                40.4                28                  2               1              25


                 1.8              10                45.5                12                  2               0              10

                 4.7              23                67.6                11                  2               0               9
                 0.7              87                27.9               225                  0               7             218
                 0.6              19                54.3                16                  0               0              16
                                                                                                                   (continued)


                                    Page 115                                                GAO/GGD-99-84 Consumer Price Index
                                         Appendix VIII
                                         CPI Price Quotations, Substitutions, and
                                         Methods of Adjustment for 1997




Categories (name of component,                    Relative                                    Percent of
expenditure class, item stratum, and          importance          Number of price             quotations          Number of
ELI)                                        for December              quotations     with price changes         substitutions
      Chuck roast                                    0.083                   5,910                 53.5                   40
      Round roast                                    0.048                   5,824                 51.5                   31
      Round steak                                    0.076                   5,989                 46.6                   41
      Sirloin steak                                  0.072                   5,924                 47.3                   33
      Other steak, roast, and other
      beef                                           0.345                  17,493                 43.0                  132
         Other roasts, excluding chuck
                                                           a
         and round                                                           2,521                 40.1                   23
         Other steak, excluding round
                                                           a
         and sirloin                                                        10,900                 46.4                   74
                                                           a
         Other beef                                                          4,072                 35.8                   35
    Pork                                             0.595                  29,378                 42.3                  446
      Bacon                                          0.112                   6,056                 38.6                   93
      Pork chops                                     0.137                   6,249                 46.9                   18
      Ham                                            0.138                   5,298                 44.6                  179
                                                           a
         Ham, excluding canned                                               5,013                 46.3                  173
                                                           a
         Canned ham                                                            285                 14.0                    6
      Other pork, including sausage                  0.208                  11,775                 40.7                  156
         Pork roasts, picnics, and
                                                           a
         other pork                                                          6,718                 44.8                   88
                                                           a
         Pork sausage                                                        5,057                 35.1                   68
    Other meats                                      0.393                  12,020                 27.6                  261
                                                           a
         Frankfurters                                                        2,627                 35.4                   55
         Bologna, liverwurst, and
                                                           a
         salami                                                              2,852                 24.5                   43
         Other lunchmeats, excluding
         bologna, liverwurst, and
                                                           a
         salami                                                              5,521                 26.3                  143
         Lamb, organ meats, and
                                                           a
         game                                                                1,020                 23.1                   20
    Poultry                                          0.439                  17,348                 36.9                  463
      Fresh whole chicken                            0.148                   5,996                 37.0                  118
      Fresh or frozen chicken parts                  0.205                   6,092                 39.3                   78
      Other poultry                                  0.086                   5,260                 34.0                  267
    Fish and seafood                                 0.373                  12,036                 32.1                  215
      Canned fish and seafood                        0.072                   6,469                 26.4                  102
      Fresh or frozen fish and seafood               0.301                   5,567                 38.7                  113
                                                           a
         Shellfish, excluding canned                                         1,646                 32.6                   34




                                         Page 116                                          GAO/GGD-99-84 Consumer Price Index
                                    Appendix VIII
                                    CPI Price Quotations, Substitutions, and
                                    Methods of Adjustment for 1997




                       Substitutions not adjusted
    Percent of price         (comparable)                                  Adjusted substitutions by method
quotations that were   Number not        Percent of          Number of                         Class mean
       substitutions     adjusted      substitutions       adjustments              Direct   and/or overlap        Linking
                 0.7              6              15.0                34                  0               1              33
                 0.5              6              19.4                25                  0               2              23
                 0.7              9              22.0                32                  0               1              31
                 0.6             12              36.4                21                  0               1              20

                 0.8             35              26.5                97                  0               2              95

                 0.9              5              21.7                18                  0               1              17

                 0.7             21              28.4                53                  0               1              52
                 0.9              9              25.7                26                  0               0              26
                 1.5           247               55.4               199                  0               5             194
                 1.5             62              66.7                31                  0               0              31
                 0.3              5              27.8                13                  0               2              11
                 3.4             95              53.1                84                  0               0              84
                 3.5             93              53.8                80                  0               0              80
                 2.1              2              33.3                  4                 0               0               4
                 1.3             85              54.5                71                  0               3              68

                 1.3             45              51.1                43                  0               3              40
                 1.3             40              58.8                28                  0               0              28
                 2.2           160               61.3               101                  0               0             101
                 2.1             35              63.6                20                  0               0              20

                 1.5             20              46.5                23                  0               0              23


                 2.6             90              62.9                53                  0               0              53

                 2.0             15              75.0                  5                 0               0               5
                 2.7           398               86.0                65                  0               2              63
                 2.0           109               92.4                  9                 0               0               9
                 1.3             59              75.6                19                  0               1              18
                 5.1           230               86.1                37                  0               1              36
                 1.8           124               57.7                91                  3               1              87
                 1.6             67              65.7                35                  3               0              32
                 2.0             57              50.4                56                  0               1              55
                 2.1             18              52.9                16                  0               0              16
                                                                                                                (continued)



                                   Page 117                                              GAO/GGD-99-84 Consumer Price Index
                                         Appendix VIII
                                         CPI Price Quotations, Substitutions, and
                                         Methods of Adjustment for 1997




Categories (name of component,                    Relative                                    Percent of
expenditure class, item stratum, and          importance          Number of price             quotations          Number of
ELI)                                        for December              quotations     with price changes         substitutions
                                                           a
         Fish, excluding canned                                              3,921                 41.3                   79
    Eggs                                             0.187                   6,500                 63.1                   38
    Fresh milk and cream                              0.61                  12,970                 34.5                  125
      Fresh whole milk                               0.352                   6,545                 35.1                   56
      Other fresh milk and cream                     0.257                   6,425                 34.0                   69
    Processed dairy products                         0.608                  19,066                 31.6                  308
      Cheese                                         0.338                   6,380                 31.5                  123
      Ice cream and related products                 0.157                   6,396                 29.0                   96
      Other dairy products, including
      butter                                         0.113                   6,290                 34.4                   89
                                                           a
         Butter                                                              2,152                 50.0                   21
                                                           a
         Other dairy products                                                4,138                 26.3                   68
    Fresh fruits                                     0.740                  43,203                 51.3                  118
      Apples                                         0.116                   8,460                 41.5                   21
      Bananas                                        0.067                   6,493                 37.8                    1
      Oranges, including tangerines                  0.084                   7,531                 53.9                   40
      Other fresh fruits                             0.474                  20,719                 58.6                   56
    Fresh vegetables                                 0.631                  31,291                 53.4                   62
      Potatoes                                       0.100                   6,320                 46.3                   23
      Lettuce                                        0.077                   6,396                 55.3                    1
      Tomatoes                                       0.115                   6,204                 66.4                   18
      Other fresh vegetables                         0.339                  12,371                 49.5                   20
    Processed fruits                                 0.349                  19,567                 28.3                  354
      Fruit juices and frozen fruit                  0.273                  12,963                 33.0                  229
                                                           a
         Frozen orange juice                                                 2,415                 33.0                   21
         Other frozen fruits and fruit
                                                           a
         juices                                                              1,154                 26.8                   12
         Fresh, canned, and/or bottled
                                                           a
         fruit juices                                                        9,394                 31.5                  196
      Canned and dried fruits                        0.077                   6,604                 22.2                  125
    Processed vegetables                             0.261                  13,020                 26.6                  200
      Frozen vegetables                              0.087                   6,434                 29.5                  104
      Processed vegetables,
      excluding frozen                               0.174                   6,586                 23.9                   96
                                                           a
         Canned beans other than lima                                          926                 24.3                   10
                                                           a
         Canned cut corn                                                       854                 28.9                    9
                                                           a
         Other processed vegetables                                          4,806                 22.9                   77




                                         Page 118                                          GAO/GGD-99-84 Consumer Price Index
                                    Appendix VIII
                                    CPI Price Quotations, Substitutions, and
                                    Methods of Adjustment for 1997




                       Substitutions not adjusted
    Percent of price         (comparable)                                  Adjusted substitutions by method
quotations that were   Number not        Percent of          Number of                         Class mean
       substitutions     adjusted      substitutions       adjustments              Direct   and/or overlap        Linking
                 2.0             39              49.4                40                  0               1              39
                 0.6             22              57.9                16                  0               0              16
                 1.0           102               81.6                23                  0               0              23
                 0.9             46              82.1                10                  0               0              10
                 1.1             56              81.2                13                  0               0              13
                 1.6           180               58.4               128                  0               3             125
                 1.9             72              58.5                51                  0               1              50
                 1.5             56              58.3                40                  0               2              38

                 1.4             52              58.4                37                  0               0              37
                 1.0             12              57.1                  9                 0               0               9
                 1.6             40              58.8                28                  0               0              28
                 0.3             80              67.8                38                  0               0              38
                 0.3             14              66.7                  7                 0               0               7
                 0.0              0                0.0                 1                 0               0               1
                 0.5             32              80.0                  8                 0               0               8
                 0.3             34              60.7                22                  0               0              22
                 0.2             43              69.4                19                  0               0              19
                 0.4             15              65.2                  8                 0               0               8
                 0.0              0                0.0                 1                 0               0               1
                 0.3             16              88.9                  2                 0               0               2
                 0.2             12              60.0                  8                 0               0               8
                 1.8           242               68.4               112                  0               1             111
                 1.8           145               63.3                84                  0               1              83
                 0.9             13              61.9                  8                 0               0               8

                 1.0              6              50.0                  6                 0               0               6

                 2.1           126               64.3                70                  0               1              69
                 1.9             97              77.6                28                  0               0              28
                 1.5           146               73.0                54                  0               3              51
                 1.6             76              73.1                28                  0               0              28

                 1.5             70              72.9                26                  0               3              23
                 1.1              7              70.0                  3                 0               0               3
                 1.1              6              66.7                  3                 0               0               3
                 1.6             57              74.0                20                  0               3              17
                                                                                                                (continued)



                                   Page 119                                              GAO/GGD-99-84 Consumer Price Index
                                           Appendix VIII
                                           CPI Price Quotations, Substitutions, and
                                           Methods of Adjustment for 1997




Categories (name of component,                      Relative                                    Percent of
expenditure class, item stratum, and            importance          Number of price             quotations          Number of
ELI)                                          for December              quotations     with price changes         substitutions
    Sugar and sweets                                   0.332                  11,935                 19.2                  187
      Sugar and artificial sweeteners                  0.085                   5,810                 21.8                   74
      Sweets, including candy                          0.247                   6,125                 16.7                  113
                                                             a
         Candy and chewing gum                                                 4,675                 14.8                   91
         Other sweets, excluding
                                                             a
         candy and gum                                                         1,450                 22.8                   22
    Fats and oils                                      0.241                  13,191                 24.6                  137
                                                             a
         Margarine                                                             1,999                 25.7                   22
                                                             a
         Other fats and oils                                                   7,583                 23.3                   83
                                                             a
         Nondairy cream substitutes                                            1,571                 22.3                   13
                                                             a
         Peanut butter                                                         2,038                 29.9                   19
    Nonalcoholic beverages                             0.747                  19,262                 31.7                  267
      Carbonated drinks                                0.348                   6,639                 35.3                   69
                                                             a
         Cola drinks                                                           4,046                 38.5                   26
         Carbonated drinks other than
                                                             a
         cola                                                                  2,593                 30.2                   43
      Coffee                                           0.263                   6,314                 37.8                   83
                                                             a
         Roasted coffee                                                        3,685                 44.3                   50
                                                             a
         Instant and freeze dried coffee                                       2,629                 28.8                   33
      Other noncarbonated drinks                       0.136                   6,309                 21.9                  115
         Noncarbonated fruit-flavored
                                                             a
         drinks                                                                2,079                 21.8                   34
                                                             a
         Tea                                                                   1,856                 22.0                   23
                                                             a
         Other noncarbonated drinks                                            2,374                 22.0                   58
    Other prepared foods                               1.046                  31,997                 26.9                  506
      Canned and packaged soup                         0.096                   6,504                 25.5                   90
      Frozen prepared food                             0.165                   6,185                 36.1                  146
                                                             a
         Frozen prepared meals                                                 2,213                 37.1                   69
         Frozen prepared foods other
                                                             a
         than meals                                                            3,972                 35.5                   77
      Snacks                                           0.205                   6,321                 24.7                  124
                                                             a
         Potato chips and other snacks                                         5,098                 25.2                  101
                                                             a
         Nuts                                                                  1,223                 22.7                   23
      Seasonings, condiments,
      sauces, and spices                               0.282                   6,539                 22.6                   77
         Salt, other seasoning, and
                                                             a
         spices                                                                1,447                 15.6                   16
                                                             a
         Olives, pickles, and relishes                                           713                 26.0                   11




                                           Page 120                                          GAO/GGD-99-84 Consumer Price Index
                                    Appendix VIII
                                    CPI Price Quotations, Substitutions, and
                                    Methods of Adjustment for 1997




                       Substitutions not adjusted
    Percent of price         (comparable)                                  Adjusted substitutions by method
quotations that were   Number not        Percent of          Number of                         Class mean
       substitutions     adjusted      substitutions       adjustments              Direct   and/or overlap        Linking
                 1.6           104               55.6                83                  0               0              83
                 1.3             38              51.4                36                  0               0              36
                 1.8             66              58.4                47                  0               0              47
                 2.0             54              59.3                37                  0               0              37

                 1.5             12              54.6                10                  0               0              10
                 1.0             97              70.8                40                  0               0              40
                 1.1             18              81.8                  4                 0               0               4
                 1.1             55              66.3                28                  0               0              28
                 0.8             11              84.6                  2                 0               0               2
                 0.9             13              68.4                  6                 0               0               6
                 1.4           169               63.3                98                  9               2              87
                 1.0             46              66.7                23                  3               0              20
                 0.6             17              65.4                  9                 0               0               9

                 1.7             29              67.4                14                  3               0              11
                 1.3             50              60.2                33                  3               1              29
                 1.4             34              68.0                16                  1               0              15
                 1.3             16              48.5                17                  2               1              14
                 1.8             73              63.5                42                  3               1              38

                 1.6             20              58.8                14                  0               0              14
                 1.2             15              65.2                  8                 1               1               6
                 2.4             38              65.5                20                  2               0              18
                 1.6           273               54.0               233                 11               3             219
                 1.4             47              52.2                43                  1               1              41
                 2.4             79              54.1                67                  4               1              62
                 3.1             39              56.5                30                  3               0              27

                 1.9             40              52.0                37                  1               1              35
                 2.0             62              50.0                62                  0               0              62
                 2.0             50              49.5                51                  0               0              51
                 1.9             12              52.2                11                  0               0              11

                 1.2             35              45.5                42                  3               0              39

                 1.1              8              50.0                  8                 2               0               6
                 1.5              6              54.6                  5                 0               0               5
                                                                                                                (continued)


                                   Page 121                                              GAO/GGD-99-84 Consumer Price Index
                                          Appendix VIII
                                          CPI Price Quotations, Substitutions, and
                                          Methods of Adjustment for 1997




Categories (name of component,                     Relative                                       Percent of
expenditure class, item stratum, and           importance          Number of price                quotations          Number of
ELI)                                         for December              quotations        with price changes         substitutions
                                                            a
         Sauces and gravies                                                   3,269                    26.2                   41
         Other condiments, excluding
                                                            a
         olives, pickles, and relishes                                        1,110                    19.2                    9
      Miscellaneous prepared food,
      including baby food                             0.299                   6,448                    25.9                   69
         Canned or packaged salads
                                                            a
         and desserts                                                           638                    27.4                   12
                                                            a
         Baby food                                                            1,389                    22.0                   10
         Other canned or packaged
                                                            a
         prepared foods                                                       4,421                    26.8                   47
    Food away from home                               5.923                  60,536                    13.9                1,336
      Lunch                                           2.097                  21,256                    13.6                  431
      Dinner                                          2.512                  27,850                    14.0                  665
      Other meals and snacks                          1.004                  11,430                    14.0                  240
         Snacks and nonalcoholic
                                                            a
         beverages                                                            8,413                    13.9                  209
                                                            a
         Breakfast or brunch                                                  3,017                    14.2                   31
      Unpriced board and catered
      affairsb                                        0.310                          0                   0.0                   0
    Alcoholic beverages                               0.813                  14,868                    23.7                  377
      Beer, ale, and other alcoholic
      malt beverages at home                          0.415                   3,919                    30.3                   64
      Distilled spirits                               0.209                   3,803                    23.1                   27
                                                            a
         Whiskey at home                                                      1,553                    25.3                    9
         Distilled spirits at home,
                                                            a
         excluding whiskey                                                    2,250                    21.5                   18
      Wine at home                                    0.189                   3,844                    28.6                  194
      Alcoholic beverages away from
      home                                            0.765                   3,302                    10.8                   92
         Beer, ale, and other alcoholic
         malt beverages away from
                                                            a
         home                                                                 1,333                      9.7                  30
                                                            a
         Wine away from home                                                    834                    10.4                   42
         Distilled spirits away from
                                                            a
         home                                                                 1,135                    12.3                   20
  Housing component total                            41.469                 182,162                     n/a               10,254
    Renters’ costs                                    8.169                  70,792                     n/a                6,524
      Rent, residential                               5.810                  45,732                      n/a               6,180
      Lodging while out of town                       2.089                  22,889                    39.0                  297
      Lodging while at school                         0.236                     390                    47.4                   24



                                          Page 122                                             GAO/GGD-99-84 Consumer Price Index
                                    Appendix VIII
                                    CPI Price Quotations, Substitutions, and
                                    Methods of Adjustment for 1997




                       Substitutions not adjusted
    Percent of price         (comparable)                                  Adjusted substitutions by method
quotations that were   Number not        Percent of          Number of                          Class mean
       substitutions     adjusted      substitutions       adjustments              Direct    and/or overlap       Linking
                 1.3             15              36.6                26                  1                0             25

                 0.8              6              66.7                  3                 0                0              3

                 1.1             50              72.5                19                  3                1             15

                 1.9             11              91.7                  1                 0                0              1
                 0.7              8              80.0                  2                 1                0              1

                 1.1             31              66.0                16                  2                1             13
                 2.2           533               39.9               803                 41                1            761
                 2.0           175               40.6               256                 10                0            246
                 2.4           279               42.0               386                 19                0            367
                 2.1             79              32.9               161                 12                1            148

                 2.5             68              32.5               141                 10                1            130
                 1.0             11              35.5                20                  2                0             18

                 0.0              0                0.0                 0                 0                0              0
                 2.5           241               63.9               136                 22                0            114

                 1.6             36              56.3                28                  5                0             23
                 0.7             17              63.0                10                  2                0              8
                 0.6              5              55.6                  4                 1                0              3

                 0.8             12              66.7                  6                 1                0              5
                 5.1           145               74.7                49                  2                0             47

                 2.8             43              46.7                49                 13                0             36


                 2.3             12              40.0                18                  6                0             12
                 5.0             21              50.0                21                  4                0             17

                 1.8             10              50.0                10                  3                0              7
                 5.6            n/a               n/a             7,774              5,773            1,023            978
                 9.8            n/a               n/a             6,228              5,658                0            570
                13.5            n/a                n/a            6,180              5,641c               0            539
                 1.3           264               88.9                33                 12                0             21
                 6.2             12              50.0                12                  5                0              7
                                                                                                                (continued)


                                   Page 123                                              GAO/GGD-99-84 Consumer Price Index
                                            Appendix VIII
                                            CPI Price Quotations, Substitutions, and
                                            Methods of Adjustment for 1997




Categories (name of component,                       Relative                                       Percent of
expenditure class, item stratum, and             importance          Number of price                quotations          Number of
ELI)                                           for December              quotations        with price changes         substitutions
      Tenants’ insurance                                0.034                   1,781                    10.1                   23
    Rental equivalence and
                                                                                       d                     d                    d
    household insurance                                20.269
                                                                                       e                     e                    e
      Owners’ equivalent rent                          19.881
                                                                                       f                     f                    f
      Household insurance                               0.388
    Maintenance and repair services                     0.202                     457                    19.3                   17
      Property maintenance and
      repair services                                   0.126                     457                    19.3                   17
         Inside home maintenance
                                                              a
         and repair services                                                      381                    17.3                    8
         Repair and/or replacement of
                                                              a
         hard surface flooring                                                         0                   0.0                   0
         Replacement of installed
                                                              a
         wall-to-wall carpet                                                       76                    29.0                    9
         Repair of disposal, built-in
                                                              a
         dishwasher, and range hood                                                    0                   0.0                   0
    Maintenance and repair
    commodities                                         0.076                   1,637                    15.6                   57
      Materials, supplies, and
      equipment for home repairs                        0.034                     825                    19.5                   29
                                                              a
         Paint, wallpaper, and supplies                                           408                    20.1                   13
         Tools and equipment for
                                                              a
         painting                                                                  61                    19.7                    4
         Lumber, paneling, wall and
                                                              a
         ceiling tile, awnings, and glass                                         179                    25.1                    4
         Blacktop and masonry
                                                              a
         materials                                                                 23                    17.4                    0
         Plumbing supplies and
                                                              a
         equipment                                                                 67                      6.0                   1
         Electrical supplies and
         heating and cooling
                                                              a
         equipment                                                                 87                    16.1                    7
      Other maintenance and repair
      commodities                                       0.042                     812                    11.7                   28
         Miscellaneous supplies and
                                                              a
         equipment                                                                609                      9.2                  22
                                                              a
         Hard surface floor covering                                              136                    16.2                    4
                                                              a
         Landscaping items                                                         67                    25.4                    2
    Fuel oil and other fuels                            0.368                   5,297                    43.0                   23
      Fuel oil                                          0.254                   2,383                    55.4                    9




                                            Page 124                                             GAO/GGD-99-84 Consumer Price Index
                                    Appendix VIII
                                    CPI Price Quotations, Substitutions, and
                                    Methods of Adjustment for 1997




                       Substitutions not adjusted
    Percent of price         (comparable)                                   Adjusted substitutions by method
quotations that were   Number not        Percent of          Number of                           Class mean
       substitutions     adjusted      substitutions       adjustments               Direct    and/or overlap       Linking
                 1.3             20              87.0                  3                  0                0              3

                   d               d                  d                 d                  d                d              d

                   e               e                  e                 e                  e                e              e

                   f               f                  f                 f                  f                f              f

                 3.7              2              11.8                15                   1                0             14

                 3.7              2              11.8                15                   1                0             14

                 2.1              2              25.0                  6                  0                0              6

                 0.0              0                0.0                 0                  0                0              0

                11.8              0                0.0                 9                  1                0              8

                 0.0              0                0.0                 0                  0                0              0

                 3.5             37              64.9                20                   2                0             18

                 3.5             20              69.0                  9                  1                0              8
                 3.2              8              61.5                  5                  0                0              5

                 6.6              3              75.0                  1                  0                0              1

                 2.2              3              75.0                  1                  1                0              0

                 0.0              0                0.0                 0                  0                0              0

                 1.5              1             100.0                  0                  0                0              0


                 8.1              5              71.4                  2                  0                0              2

                 3.4             17              60.7                11                   1                0             10

                 3.6             15              68.2                  7                  1                0              6
                 2.9              1              25.0                  3                  0                0              3
                 3.0              1              50.0                  1                  0                0              1
                 0.4             12              52.2                11                   6                0              5
                 0.4              4              44.4                  5                  2                0              3
                                                                                                                 (continued)



                                   Page 125                                               GAO/GGD-99-84 Consumer Price Index
                                          Appendix VIII
                                          CPI Price Quotations, Substitutions, and
                                          Methods of Adjustment for 1997




Categories (name of component,                     Relative                                       Percent of
expenditure class, item stratum, and           importance          Number of price                quotations          Number of
ELI)                                         for December              quotations        with price changes         substitutions
      Other household fuel
      commodities                                     0.113                   2,914                    32.9                   14
                                                            a
         Bottled or tank gas                                                  2,073                    35.4                    8
                                                            a
         Coal                                                                        0                   0.0                   0
                                                            a
         Other fuels                                                            841                    26.8                    6
    Gas (piped) and electricity                       3.401                  28,152                    51.4                  123
      Electricity                                     2.265                  15,586                    40.6                   78
      Utility natural gas service                     1.136                  12,566                    64.9                   45
    Other utilities and public services               3.246                  28,653                    13.4                  260
      Telephone services, local
      charges                                         1.115                   6,450                    12.7                   26
      Interstate toll calls                           0.305                   5,265                    18.5                   34
      Intrastate toll calls                           0.227                   5,674                      6.9                  23
      Water and sewerage
      maintenance                                     0.795                   4,112                    10.9                   42
      Community antenna and cable
      television                                      0.582                   3,867                    25.2                   93
      Garbage and trash collection                    0.221                   3,285                      7.4                  42
    Textile house furnishings                         0.313                   3,898                    34.7                  289
                                                            a
         Bathroom linens                                                        679                    35.2                   55
                                                            a
         Bedroom linens                                                       1,638                    39.1                  130
         Kitchen and dining room
                                                            a
         linens                                                                 220                    23.2                   22
                                                            a
         Curtains and drapes                                                    648                    36.6                   42
         Slipcovers and decorative
                                                            a
         pillows                                                                106                    23.6                   12
         Sewing materials for
                                                            a
         household items                                                        607                    26.4                   28
    Furniture and bedding                             1.089                   7,469                    32.9                  566
      Bedroom furniture                               0.361                   2,154                    31.4                  142
                                                            a
         Mattress and springs                                                   941                    33.1                   65
         Bedroom furniture other than
                                                            a
         mattress and springs                                                 1,213                    30.2                   77
      Sofas                                           0.225                   1,330                    33.6                  103
      Living room chairs and tables                   0.179                   1,341                    33.4                  105
                                                            a
         Living room chairs                                                     871                    36.6                   84
                                                            a
         Living room tables                                                     470                    27.5                   21
      Other furniture                                 0.324                   2,644                    33.6                  216




                                          Page 126                                             GAO/GGD-99-84 Consumer Price Index
                                    Appendix VIII
                                    CPI Price Quotations, Substitutions, and
                                    Methods of Adjustment for 1997




                       Substitutions not adjusted
    Percent of price         (comparable)                                  Adjusted substitutions by method
quotations that were   Number not        Percent of          Number of                         Class mean
       substitutions     adjusted      substitutions       adjustments              Direct   and/or overlap        Linking

                 0.5              8              57.1                  6                 4               0               2
                 0.4              4              50.0                  4                 2               0               2
                 0.0              0                0.0                 0                 0               0               0
                 0.7              4              66.7                  2                 2               0               0
                 0.4             98              79.7                25                 14               0              11
                 0.5             63              80.8                15                 14               0               1
                 0.4             35              77.8                10                  0               0              10
                 0.9           173               66.5                87                 54               0              33

                 0.4             17              65.4                  9                 0               0               9
                 0.7              2                5.9               32                 32               0               0
                 0.4             16              69.6                  7                 0               0               7

                 1.0             34              81.0                  8                 6               0               2

                 2.4             85              91.4                  8                 2               0               6
                 1.3             19              45.2                23                 14               0               9
                 7.4           187               64.7               102                  0              99               3
                 8.1             39              70.9                16                  0              16g              0
                                                                                                          g
                 7.9             86              66.2                44                  0              42               2

              10 a 0             12              54.6                10                  0               9g              1
                 6.5             24              57.1                18                  0              18g              0

                11.3              6              50.0                  6                 0               6g              0

                 4.6             20              71.4                  8                 0               8g              0
                 7.6           238               42.1               328                  0             320               8
                 6.6             60              42.3                82                  0              80               2
                                                                                                          g
                 6.9             32              49.2                33                  0              32               1

                 6.4             28              36.4                49                  0              48g              1
                 7.7             51              49.5                52                  0              49g              3
                 7.8             42              40.0                63                  0              62               1
                 9.6             35              41.7                49                  0              49g              0
                                                                                                          g
                 4.5              7              33.3                14                  0              13               1
                 8.2             85              39.4               131                  0             129               2
                                                                                                                (continued)



                                   Page 127                                              GAO/GGD-99-84 Consumer Price Index
                                        Appendix VIII
                                        CPI Price Quotations, Substitutions, and
                                        Methods of Adjustment for 1997




Categories (name of component,                   Relative                                       Percent of
expenditure class, item stratum, and         importance          Number of price                quotations          Number of
ELI)                                       for December              quotations        with price changes         substitutions
         Kitchen and dining room
                                                          a
         furniture                                                          1,193                    34.5                   84
                                                          a
         Infants’ furniture                                                   253                    23.3                   19
                                                          a
         Outdoor furniture                                                    201                    43.8                   39
                                                          a
         Occasional furniture                                                 997                    33.0                   74
    Television and sound equipment                  0.385                   5,327                    36.9                  711
      Television sets                               0.128                   2,170                    43.3                  306
      Video products other than
      televisions                                   0.065                     683                    32.9                  108
         Video cassette recorders,
         disc players, cameras, and
                                                          a
         accessories                                                          376                    41.0                   63
         Video cassettes and discs,
                                                          a
         blank and prerecorded                                                236                    18.2                   28
         Video game hardware,
                                                          a
         software, and accessories                                             71                    39.4                   17
      Audio components, radios,
      recordings, and other audio
      equipment                                     0.192                   2,474                    32.3                  297
         Radios, phonographs, and
                                                          a
         tape recorders and players                                           293                    39.9                   43
         Components and other sound
                                                          a
         equipment                                                            961                    32.3                  148
         Records and tapes,
                                                          a
         prerecorded and blank                                              1,220                    17.8                  106
      Unpriced accessories for
      electronic equipmentb                         0.000                          0                   0.0                   0
    Household appliances                            0.274                   3,124                    46.4                  287
      Refrigerators and home freezers               0.084                     819                    47.6                   79
      Laundry equipment                             0.092                     820                    47.0                   60
                                                          a
         Washers                                                              507                    48.3                   43
                                                          a
         Dryers                                                               313                    44.7                   17
      Stoves, ovens, portable
      dishwashers, and window air
      conditioners                                  0.098                   1,485                    45.5                  148
         Stoves and ovens, excluding
                                                          a
         microwave ovens                                                      576                    47.2                   55
                                                          a
         Microwave ovens                                                      658                    45.7                   80
                                                          a
         Portable dishwashers                                                  24                    25.0                    1
                                                          a
         Window air conditioners                                              227                    42.3                   12




                                        Page 128                                             GAO/GGD-99-84 Consumer Price Index
                                    Appendix VIII
                                    CPI Price Quotations, Substitutions, and
                                    Methods of Adjustment for 1997




                       Substitutions not adjusted
    Percent of price         (comparable)                                  Adjusted substitutions by method
quotations that were   Number not        Percent of          Number of                         Class mean
       substitutions     adjusted      substitutions       adjustments              Direct   and/or overlap        Linking

                 7.0             27              32.1                57                  0              56g              1
                                                                                                          g
                 7.5             11              57.9                  8                 0               7               1
                19.4             24              61.5                15                  0              15g              0
                                                                                                          g
                 7.4             23              31.1                51                  0              51               0
                13.4           438               61.6               273                  5             264               4
                                                                                                          g
                14.1           191               62.4               115                  1             113               1

                15.8             76              70.4                32                  3              28               1


                16.8             44              69.8                19                  0              19g              0

                11.9             22              78.6                  6                 3               3g              0

                23.9             10              58.8                  7                 0               6g              1


                12.0           171               57.6               126                  1             123               2

                14.7             18              41.9                25                  0              25g              0

                15.4             60              40.5                88                  0              87g              1

                 8.7             93              87.7                13                  1              11g              1

                 0.0              0                0.0                 0                 0               0               0
                 9.2           263               91.6                24                  0              23               1
                                                                                                          g
                 9.7             72              91.1                  7                 0               7               0
                 7.3             59              98.3                  1                 0               1               0
                                                                                                          g
                 8.5             43             100.0                  0                 0               0               0
                 5.4             16              94.1                  1                 0               1g              0


                10.0           132               89.2                16                  0              15               1

                 9.6             49              89.1                  6                 0               6g              0
                12.2             74              92.5                  6                 0               5g              1
                                                                                                          g
                 4.2              0                0.0                 1                 0               1               0
                 5.3              9              75.0                  3                 0               3g              0
                                                                                                                (continued)



                                   Page 129                                              GAO/GGD-99-84 Consumer Price Index
                                        Appendix VIII
                                        CPI Price Quotations, Substitutions, and
                                        Methods of Adjustment for 1997




Categories (name of component,                   Relative                                       Percent of
expenditure class, item stratum, and         importance          Number of price                quotations          Number of
ELI)                                       for December              quotations        with price changes         substitutions
    Information processing
    equipment                                       0.064                     649                    42.4                  119
         Personal computers and
                                                          a
         peripheral equipment                                                 375                    57.6                  103
         Computer software and
                                                          a
         accessories                                                           97                    21.7                    9
         Telephone, peripheral
                                                          a
         equipment, and accessories                                           125                    26.4                    6
         Calculators, adding
                                                          a
         machines, and typewriters                                             41                      4.9                   1
         Other information processing
                                                          a
         equipment                                                             11                    27.3                    0
    Other household equipment and
    furnishings                                     1.107                   9,383                    26.8                  784
      Floor and window coverings,
      and outdoor, infants’, laundry,
      and cleaning equipment                        0.176                   1,533                    25.2                  101
                                                          a
         Floor coverings                                                      544                    21.7                   38
                                                          a
         Window coverings                                                     519                    29.7                   13
                                                          a
         Infants’ equipment                                                   111                    23.4                   10
         Laundry and cleaning
                                                          a
         equipment                                                            271                    18.8                   29
                                                          a
         Outdoor equipment                                                     88                    42.1                   11
      Clocks, lamps, and decorator
      items                                         0.218                   1,570                    26.1                  181
                                                          a
         Clocks                                                                80                    21.3                    9
                                                          a
         Lamps and lighting fixtures                                          353                    29.8                   37
                                                          a
         Household decorative items                                         1,137                    25.2                  135
      Tableware, serving pieces, and
      nonelectric kitchenware                       0.198                   1,790                    24.3                  162
                                                          a
         Plastic dinnerware                                                    30                    10.0                    1
                                                          a
         China and other dinnerware                                           389                    31.6                   42
                                                          a
         Flatware                                                             223                    29.6                   18
                                                          a
         Glassware                                                            228                    24.1                   17
                                                          a
         Silver serving pieces                                                     0                   0.0                   0
         Serving pieces other than
                                                          a
         silver or glass                                                       57                    33.3                    4
                                                          a
         Nonelectric cookware                                                 236                    24.6                   19
         Tableware and nonelectric
                                                          a
         kitchenware                                                          627                    17.7                   61




                                        Page 130                                             GAO/GGD-99-84 Consumer Price Index
                                    Appendix VIII
                                    CPI Price Quotations, Substitutions, and
                                    Methods of Adjustment for 1997




                       Substitutions not adjusted
    Percent of price         (comparable)                                  Adjusted substitutions by method
quotations that were   Number not        Percent of          Number of                         Class mean
       substitutions     adjusted      substitutions       adjustments              Direct   and/or overlap        Linking

                18.3             67              56.3                52                  0               0              52

                27.5             59              57.3                44                  0               0              44

                 9.3              5              55.6                  4                 0               0               4

                 4.8              2              33.3                  4                 0               0               4

                 2.4              1             100.0                  0                 0               0               0

                 0.0              0                0.0                 0                 0               0               0

                 8.4           443               56.5               341                  5             317              19


                 6.6             56              55.5                45                  1              43               1
                 7.0             22              57.9                16                  1              14g              1
                                                                                                          g
                 2.5             10              76.9                  3                 0               3               0
                 9.0              7              70.0                  3                 0               3g              0

                10.7             14              48.3                15                  0              15g              0
                                                                                                          g
                12.5              3              27.3                  8                 0               8               0

                11.5             97              53.6                84                  0              80               4
                11.3              4              44.4                  5                 0               5g              0
                                                                                                          g
                10.5             23              62.2                14                  0              13               1
                11.9             70              51.9                65                  0              62g              3

                 9.1             74              45.7                88                  1              83               4
                                                                                                          g
                 3.3              1             100.0                  0                 0               0               0
                10.8             19              45.2                23                  1              20g              2
                                                                                                          g
                 8.1             10              55.6                  8                 0               8               0
                 7.5              7              41.2                10                  0               9g              1
                 0.0              0                0.0                 0                 0               0               0

                 7.0              1              25.0                  3                 0               3g              0
                 8.1              9              47.4                10                  0              10g              0

                 9.7             27              44.3                34                  0              33g              1
                                                                                                                (continued)


                                   Page 131                                              GAO/GGD-99-84 Consumer Price Index
                                         Appendix VIII
                                         CPI Price Quotations, Substitutions, and
                                         Methods of Adjustment for 1997




Categories (name of component,                    Relative                                       Percent of
expenditure class, item stratum, and          importance          Number of price                quotations          Number of
ELI)                                        for December              quotations        with price changes         substitutions
      Lawn and garden equipment,
      tools, and hardware                            0.173                   1,665                    24.1                  109
                                                           a
         Lawn and garden equipment                                             819                    33.7                   80
                                                           a
         Power tools                                                           339                    16.2                   11
                                                           a
         Other hardware                                                        263                    11.8                    9
                                                           a
         Nonpowered hand tools                                                 244                    16.0                    9
      Small kitchen appliances,
      sewing machines, and portable
      heating and cooling equipment                  0.113                   1,615                    32.8                  144
         Floor cleaning equipment and
                                                           a
         sewing machines                                                       597                    38.5                   63
         Portable heating and cooling
         equipment, and small electric
                                                           a
         kitchen appliances                                                  1,018                    29.5                   81
      Indoor plants and fresh cut
      flowers                                        0.150                   1,210                    29.1                   87
      Unpriced household equipment
      parts and small furnishingsb                   0.079                          0                   0.0                   0
    Housekeeping supplies                            1.090                  10,013                    23.8                  446
      Laundry and cleaning products                  0.380                   4,170                    25.1                  144
                                                           a
         Soaps and detergents                                                2,512                    26.6                   97
         Other laundry and cleaning
                                                           a
         products                                                            1,658                    23.0                   47
      Household paper products,
      including stationery                           0.364                   2,974                    22.1                  195
         Cleansing and toilet tissue,
                                                           a
         paper towels, and napkins                                           1,531                    27.3                   53
         Stationery, stationery
                                                           a
         supplies, and gift wrap                                             1,443                    16.5                  142
      Other household products, lawn,
      and garden supplies                            0.346                   2,869                    23.6                  107
         Miscellaneous household
                                                           a
         products                                                            2,111                    23.4                   61
                                                           a
         Lawn and garden supplies                                              758                    24.0                   46
    Housekeeping services                            1.492                   7,311                      6.6                  48
      Postage                                        0.254                   1,578                      0.0                   0
      Appliance and furniture repair                 0.182                     817                      8.6                   3
         Repair of television, radio,
                                                           a
         and sound equipment                                                   415                      4.6                   0
         Repair of household
                                                           a
         appliances                                                            229                    18.8                    1



                                         Page 132                                             GAO/GGD-99-84 Consumer Price Index
                                    Appendix VIII
                                    CPI Price Quotations, Substitutions, and
                                    Methods of Adjustment for 1997




                       Substitutions not adjusted
    Percent of price         (comparable)                                  Adjusted substitutions by method
quotations that were   Number not        Percent of          Number of                         Class mean
       substitutions     adjusted      substitutions       adjustments              Direct   and/or overlap        Linking

                 6.6             88              80.7                21                  1              20               0
                                                                                                          g
                 9.8             69              86.3                11                  1              10               0
                 3.2              7              63.6                  4                 0               4g              0
                                                                                                          g
                 3.4              7              77.8                  2                 0               2               0
                 3.7              5              55.6                  4                 0               4g              0


                 8.9             93              64.6                51                  1              45               5

                10.6             48              76.2                15                  1              14g              0


                 8.0             45              55.6                36                  0              31g              5

                 7.2             35              40.2                52                  1              46g              5

                 0.0              0                0.0                 0                 0               0               0
                 4.5           206               46.2               240                 19               0             221
                 3.5             58              40.3                86                  5               0              81
                 3.9             35              36.1                62                  3               0              59

                 2.8             23              48.9                24                  2               0              22

                 6.6             95              48.7               100                 12               0              88

                 3.5             16              30.2                37                  7               0              30

                 9.8             79              55.6                63                  5               0              58

                 3.7             53              49.5                54                  2               0              52

                 2.9             31              50.8                30                  2               0              28
                 6.1             22              47.8                24                  0               0              24
                 0.7             20              41.7                28                  9               0              19
                 0.0              0                0.0                 0                 0               0               0
                 0.4              1              33.3                  2                 1               0               1

                 0.0              0                0.0                 0                 0               0               0

                 0.4              0                0.0                 1                 1               0               0
                                                                                                                (continued)


                                   Page 133                                              GAO/GGD-99-84 Consumer Price Index
                                         Appendix VIII
                                         CPI Price Quotations, Substitutions, and
                                         Methods of Adjustment for 1997




Categories (name of component,                    Relative                                       Percent of
expenditure class, item stratum, and          importance          Number of price                quotations          Number of
ELI)                                        for December              quotations        with price changes         substitutions
                                                           a
         Reupholstery of furniture                                             173                      4.6                   2
      Gardening and other household
      services                                       0.378                   3,765                      8.6                  32
         Gardening and lawn care
                                                           a
         services                                                            1,535                    11.1                   20
                                                           a
         Water softening service                                                84                      8.3                   1
         Moving, storage, and freight
                                                           a
         expense                                                             1,033                    10.1                   10
         Household laundry and
         drycleaning, excluding coin
                                                           a
         operated                                                              946                      4.0                   1
         Coin-operated household
                                                           a
         laundry and drycleaning                                               167                      1.2                   0
                                                                                    h                     h                    h
      Babysitting                                    0.271
      Domestic services                              0.237                   1,010                      7.5                  10
      Care of invalids, elderly, and
      convalescents in the home                      0.054                     141                      7.8                   3
      Unpriced rent and/or repair of
      household equipment, and
      sound equipmentb                               0.116                          0                   0.0                   0
  Apparel and upkeep component
  total                                              5.291                  76,736                     n/a                9,797
    Men’s apparel                                    1.061                  17,016                    40.6                1,391
      Men’s suits, coats, sportcoats,
      and jackets                                    0.312                   6,221                    40.6                  466
                                                           a
         Men’s suits                                                         3,841                    38.4                  181
         Men’s sportcoats and tailored
                                                           a
         jackets                                                               619                    47.0                   40
                                                           a
         Men’s coats and jackets                                             1,761                    43.3                  245
      Men’s furnishings                              0.257                   3,335                    37.2                  336
                                                           a
         Men’s underwear and hosiery                                           968                    29.1                   43
                                                           a
         Men’s nightwear                                                       287                    24.0                   17
                                                           a
         Men’s accessories                                                   1,067                    29.6                   88
                                                           a
         Men’s sweaters                                                        474                    58.0                   93
                                                           a
         Men’s active sportswear                                               539                    55.5                   95
      Men’s shirts                                   0.266                   4,100                    45.1                  407
      Men’s pants and shorts                         0.212                   3,360                    38.5                  182
      Unpriced men’s uniforms and
      other clothingb                                0.014                          0                   0.0                   0
    Boys’ apparel                                    0.230                   3,457                    43.5                  436




                                         Page 134                                             GAO/GGD-99-84 Consumer Price Index
                                    Appendix VIII
                                    CPI Price Quotations, Substitutions, and
                                    Methods of Adjustment for 1997




                       Substitutions not adjusted
    Percent of price         (comparable)                                  Adjusted substitutions by method
quotations that were   Number not        Percent of          Number of                          Class mean
       substitutions     adjusted      substitutions       adjustments              Direct    and/or overlap         Linking
                 1.2              1              50.0                  1                 0                0                1

                 0.9             17              53.1                15                  3                0               12

                 1.3             10              50.0                10                  2                0                8
                 1.2              0                0.0                 1                 1                0                0

                 1.0              7              70.0                  3                 0                0                3


                 0.1              0                0.0                 1                 0                0                1

                 0.0              0                0.0                 0                 0                0                0
                   h               h                  h                 h                 h                   h             h

                 1.0              1              10.0                  9                 4                0                5

                 2.1              1              33.3                  2                 1                0                1


                 0.0              0                0.0                 0                 0                0                0

                12.8          6,598              67.3             3,199              1,223            1,796              180
                 8.2           989               71.1               402                192              183               27

                 7.5           285               61.2               181                101               69               11
                 4.7             93              51.4                88                 71               14g               3

                 6.5             24              60.0                16                  7                7g               2
                                                                                                           g
                13.9           168               68.6                77                 23               48                6
                10.1           254               75.6                82                 18               57                7
                                                                                                           g
                 4.4             39              90.7                  4                 0                3                1
                 5.9             16              94.1                  1                 0                1g               0
                                                                                                           g
                 8.3             71              80.7                17                  0               15                2
                19.6             51              54.8                42                 18               22g               2
                                                                                                           g
                17.6             77              81.1                18                  0               16                2
                 9.9           306               75.2               101                 53               42g               6
                                                                                                           g
                 5.4           144               79.1                38                 20               15                3

                 0.0              0                0.0                 0                 0                0                0
                12.6           305               70.0               131                  1              125                5
                                                                                                                  (continued)


                                   Page 135                                              GAO/GGD-99-84 Consumer Price Index
                                          Appendix VIII
                                          CPI Price Quotations, Substitutions, and
                                          Methods of Adjustment for 1997




Categories (name of component,                     Relative                                       Percent of
expenditure class, item stratum, and           importance          Number of price                quotations          Number of
ELI)                                         for December              quotations        with price changes         substitutions
                                                            a
         Boys’ coats and jackets                                                264                    66.3                   58
                                                            a
         Boys’ sweaters                                                          96                    62.5                   22
                                                            a
         Boys’ shirts                                                         1,008                    48.5                  159
         Boys’ underwear, nightwear,
                                                            a
         and hosiery                                                            332                    25.0                   16
                                                            a
         Boys’ accessories                                                      336                    30.7                   27
         Boys’ suits, sportcoats, and
                                                            a
         pants                                                                1,200                    37.7                   89
                                                            a
         Boys’ active sportswear                                                221                    64.7                   65
         Unpriced boys’ uniforms and
         other clothingb                                    a
                                                                                     0                   0.0                   0
    Women’s apparel                                   1.765                  23,328                    52.1                5,040
      Women’s coats and jackets                       0.181                   1,960                    55.6                  503
      Women’s dresses                                 0.246                   2,299                    65.5                  853
      Women’s separates and
      sportswear                                      0.821                  12,452                    55.3                2,653
                                                            a
         Women’s tops                                                         5,472                    61.7                1,514
                                                            a
         Women’s skirts                                                         704                    64.6                  208
                                                            a
         Women’s pants and shorts                                             5,272                    46.6                  653
                                                            a
         Women’s active sportswear                                            1,004                    60.2                  278
      Women’s underwear, nightwear,
      and accessories                                 0.323                   5,449                    35.2                  653
                                                            a
         Women’s nightwear                                                    1,155                    56.0                  288
                                                            a
         Women’s underwear                                                    1,510                    33.3                   91
                                                            a
         Women’s hosiery                                                      1,631                    19.2                   79
                                                            a
         Women’s accessories                                                  1,153                    39.4                  195
      Women’s suits                                   0.168                   1,168                    64.6                  378
      Unpriced women’s uniforms and
      other clothingb                                 0.026                          0                   0.0                   0
    Girls’ apparel                                    0.307                   4,414                    49.2                  918
                                                            a
         Girls’ coats and jackets                                               160                    60.6                   55
                                                            a
         Girls’ dresses and suits                                               547                    75.1                  238
                                                            a
         Girls’ tops                                                            804                    60.5                  212
                                                            a
         Girls’ skirts and pants                                              1,334                    45.4                  185
                                                            a
         Girls’ active sportswear                                               283                    63.3                   97
         Girls’ underwear and
                                                            a
         nightwear                                                              599                    33.4                   55
                                                            a
         Girls’ hosiery and accessories                                         687                    27.8                   76




                                          Page 136                                             GAO/GGD-99-84 Consumer Price Index
                                    Appendix VIII
                                    CPI Price Quotations, Substitutions, and
                                    Methods of Adjustment for 1997




                       Substitutions not adjusted
    Percent of price         (comparable)                                  Adjusted substitutions by method
quotations that were   Number not        Percent of          Number of                         Class mean
       substitutions     adjusted      substitutions       adjustments              Direct   and/or overlap        Linking
                                                                                                          g
                22.0             44              75.9                14                  0              14               0
                22.9             11              50.0                11                  0               9g              2
                                                                                                          g
                15.8           104               65.4                55                  0              53               2

                 4.8             12              75.0                  4                 1               3g              0
                 8.0             21              77.8                  6                 0               6g              0

                 7.4             67              75.3                22                  0              21g              1
                                                                                                          g
                29.4             46              70.8                19                  0              19               0

                 0.0              0                0.0                 0                 0               0               0
                21.6          3,199              63.5             1,841                844             897             100
                25.7           275               54.7               228                 95             117g             16
                37.1           486               57.0               367                176             175g             16

                21.3          1,689              63.7               964                484             425              55
                                                                                                          g
                27.7           922               60.9               592                299             256              37
                29.6           103               49.5               105                 62              39g              4
                                                                                                          g
                12.4           442               67.7               211                120              79              12
                27.7           222               79.9                56                  3              51g              2

                12.0           535               81.9               118                  2             110               6
                                                                                                          g
                24.9           237               82.3                51                  0              50               1
                 6.0             71              78.0                20                  1              19g              0
                                                                                                          g
                 4.8             63              79.8                16                  1              13               2
                16.9           164               84.1                31                  0              28g              3
                                                                                                          g
                32.4           214               56.6               164                 87              70               7

                 0.0              0                0.0                 0                 0               0               0
                20.8           584               63.6               334                129             183              22
                34.4             34              61.8                21                  4              15g              2
                43.5           130               54.6               108                 51              55g              2
                                                                                                          g
                26.4           148               69.8                64                 30              32               2
                13.9           116               62.7                69                 19              44g              6
                                                                                                          g
                34.3             58              59.8                39                 22              11               6

                 9.2             40              72.7                15                  1              13g              1
                11.1             58              76.3                18                  2              13g              3
                                                                                                                (continued)


                                   Page 137                                              GAO/GGD-99-84 Consumer Price Index
                                         Appendix VIII
                                         CPI Price Quotations, Substitutions, and
                                         Methods of Adjustment for 1997




Categories (name of component,                    Relative                                       Percent of
expenditure class, item stratum, and          importance          Number of price                quotations          Number of
ELI)                                        for December              quotations        with price changes         substitutions
         Unpriced girls’ uniforms and
         other clothingb                                   a
                                                                                    0                   0.0                   0
    Infants’ and toddlers’ apparel                   0.182                   1,819                    28.5                  167
         Infants’ and toddlers’
                                                           a
         outerwear                                                              37                    59.5                   11
         Infants’ and toddlers’ play
                                                           a
         and dresswear                                                         358                    50.0                   82
         Infants’ and toddlers’
                                                           a
         underwear                                                           1,268                    20.4                   52
         Infants’ and toddlers’
                                                           a
         sleepwear                                                             156                    38.5                   22
         Unpriced infants’ accessories
         and other clothingb                               a
                                                                                    0                   0.0                   0
    Sewing materials, notions, and
    luggage                                          0.083                   1,464                    32.3                  121
                                                           a
         Fabric for making clothes                                             656                    26.1                   43
                                                           a
         Sewing notions and patterns                                           185                    22.7                   10
                                                           a
         Luggage                                                               623                    41.7                   68
    Jewelry                                          0.401                   6,428                    35.9                  495
      Watches                                        0.078                   1,223                    30.3                  106
      Jewelry                                        0.323                   5,205                    37.2                  389
    Footwear                                         0.719                  11,159                    39.7                1,193
      Men’s footwear                                 0.224                   2,972                    35.6                  235
      Boys’ and girls’ footwear                      0.154                   1,537                    38.2                  203
                                                           a
         Boys’ footwear                                                        760                    44.1                  115
                                                           a
         Girls’ footwear                                                       777                    32.4                   88
      Women’s footwear                               0.341                   6,650                    41.8                  755
    Apparel services                                 0.543                   7,651                      7.1                  36
      Apparel laundry and dry
      cleaning, excluding coin
      operated                                       0.288                   4,012                      9.6                  11
      Other apparel services                         0.255                   3,639                      4.4                  25
         Shoe repair and other shoe
                                                           a
         services                                                              379                      7.7                   3
         Coin-operated apparel
                                                           a
         laundry and dry cleaning                                            2,467                      2.3                  12
                                                           a
         Alterations and repairs                                               444                      4.7                   1
                                                           a
         Clothing rental                                                       224                    19.2                    9
                                                           a
         Watch and jewelry repair                                              125                      8.0                   0




                                         Page 138                                             GAO/GGD-99-84 Consumer Price Index
                                    Appendix VIII
                                    CPI Price Quotations, Substitutions, and
                                    Methods of Adjustment for 1997




                       Substitutions not adjusted
    Percent of price         (comparable)                                  Adjusted substitutions by method
quotations that were   Number not        Percent of          Number of                         Class mean
       substitutions     adjusted      substitutions       adjustments              Direct   and/or overlap        Linking

                 0.0              0                0.0                 0                 0               0               0
                 9.2           126               75.5                41                  0              39               2

                29.7              8              72.7                  3                 0               3g              0

                22.9             65              79.3                17                  0              17g              0

                 4.1             35              67.3                17                  0              15g              2

                14.1             18              81.8                  4                 0               4g              0

                 0.0              0                0.0                 0                 0               0               0

                 8.3             79              65.3                42                  0              41               1
                 6.6             31              72.1                12                  0              11g              1
                                                                                                          g
                 5.4              7              70.0                  3                 0               3               0
                10.9             41              60.3                27                  0              27g              0
                 7.7           338               68.3               157                 12             139               6
                 8.7             82              77.4                24                  0              23g              1
                                                                                                          g
                 7.5           256               65.8               133                 12             116               5
                10.7           962               80.6               231                 30             189              12
                 7.9           193               82.1                42                 10              27g              5
                13.2           169               83.3                34                  0              31               3
                                                                                                          g
                15.1             96              83.5                19                  0              17               2
                11.3             73              83.0                15                  0              14g              1
                                                                                                          g
                11.4           600               79.5               155                 20             131               4
                 0.5             16              44.4                20                 15               0               5


                 0.3              3              27.3                  8                 7               0               1
                 0.7             13              52.0                12                  8               0               4

                 0.8              1              33.3                  2                 2               0               0

                 0.5             10              83.3                  2                 2               0               0
                 0.2              0                0.0                 1                 1               0               0
                 4.0              2              22.2                  7                 3               0               4
                 0.0              0                0.0                 0                 0               0               0
                                                                                                                (continued)



                                   Page 139                                              GAO/GGD-99-84 Consumer Price Index
                                          Appendix VIII
                                          CPI Price Quotations, Substitutions, and
                                          Methods of Adjustment for 1997




Categories (name of component,                     Relative                                       Percent of
expenditure class, item stratum, and           importance          Number of price                quotations          Number of
ELI)                                         for December              quotations        with price changes         substitutions
  Transportation component total                     16.620                  94,366                     n/a                5,660
    New vehicles                                      4.829                  15,176                    53.9                2,420
      New cars                                        3.842                  11,617                    56.4                1,917
      New trucks                                      0.894                   2,192                    52.2                  341
      New motorcycles                                 0.093                   1,367                    34.6                  162
    Used vehicles                                     1.195                   5,067                   100.0                1,097
         Unpriced used carsb                                a
                                                                                     0                   0.0                   0
    Motor fuel                                        2.925                  29,524                    75.0                  172
                                                            a
         Regular unleaded gasoline                                            9,509                    79.3                   50
                                                            a
         Mid-grade unleaded gasoline                                          8,370                    77.5                   47
                                                            a
         Premium unleaded gasoline                                            9,186                    76.3                   64
                                                            a
         Diesel                                                               2,285                    45.4                    9
                                                            a
         Other motor fuel                                                       174                    63.8                    2
    Automobile maintenance and
    repair                                            1.546                   8,401                    25.6                1,303
      Automotive body work                            0.167                   2,073                    28.1                  334
      Automobile drive train and front
      end repair                                      0.453                   1,990                    27.5                  306
                                                            a
         Automotive drive-train repair                                          803                    28.1                  129
                                                            a
         Automotive brake work                                                  545                    25.3                   81
         Repair to steering, front end,
         cooling system, and air
                                                            a
         conditioning                                                           642                    28.7                   96
      Automotive maintenance and
      servicing                                       0.490                   2,376                    17.6                  352
      Power plant repair                              0.412                   1,962                    30.5                  311
      Unpriced automotive repair
      service policyb                                 0.023                          0                   0.0                   0
    Motor oil, coolant, and other
    fluids                                            0.058                   3,745                    11.7                   36
                                                            a
         Motor oil                                                            2,676                    11.2                   18
         Coolant brake fluid,
         transmission fluid, and
                                                            a
         additives                                                            1,069                    12.9                   18
    Automobile parts and equipment                    0.516                   7,964                    20.9                  255
      Tires                                           0.256                   4,037                    26.8                  102
      Vehicle parts and equipment
      other than tires                                0.260                   3,927                    14.9                  153
    Automobile insurance                              2.647                   5,111                    25.0                  147




                                          Page 140                                             GAO/GGD-99-84 Consumer Price Index
                                    Appendix VIII
                                    CPI Price Quotations, Substitutions, and
                                    Methods of Adjustment for 1997




                       Substitutions not adjusted
    Percent of price         (comparable)                                  Adjusted substitutions by method
quotations that were   Number not        Percent of          Number of                         Class mean
       substitutions     adjusted      substitutions       adjustments              Direct   and/or overlap        Linking
                 6.2          2,699              47.6             2,961              1,837             823             301
                16.0           805               33.3             1,615                838             670             107
                                                                                                          g
                16.5           584               30.5             1,333                667             580              86
                15.6             80              23.5               261                171              75g             15
                                                                                                          g
                11.9           141               87.0                21                  0              15               6
                21.7              0                0.0            1,097                990               0             107
                 0.0              0                0.0                 0                 0               0               0
                 0.6           159               92.4                13                  2               0              11
                 0.5             47                 94                 3                 0               0               3
                 0.6             45              95.7                  2                 0               0               2
                 0.7             59              92.2                  5                 2               0               3
                 0.4              7              77.8                  2                 0               0               2
                 1.2              1              50.0                  1                 0               0               1

                15.5          1,169              89.7               134                  3             117              14
                16.1           287               85.9                47                  1              38g              8

                15.4           273               89.2                33                  1              31               1
                                                                                                          g
                16.1           114               88.4                15                  0              15               0
                14.9             73              90.1                  8                 1               6g              1


                15.0             86              89.6                10                  0              10g              0

                14.8           339               96.3                13                  1              10g              2
                15.9           270               86.8                41                  0              38g              3

                 0.0              0                0.0                 0                 0               0               0

                 1.0             27              75.0                  9                 0               0               9
                 0.7             12              66.7                  6                 0               0               6


                 1.7             15              83.3                  3                 0               0               3
                 3.2           211               82.8                44                  2              36               6
                 2.5             85              83.3                17                  0              17g              0

                 3.9           126               82.4                27                  2              19g              6
                 2.9           133               90.5                14                  0               0              14
                                                                                                                (continued)


                                   Page 141                                              GAO/GGD-99-84 Consumer Price Index
                                         Appendix VIII
                                         CPI Price Quotations, Substitutions, and
                                         Methods of Adjustment for 1997




Categories (name of component,                    Relative                                       Percent of
expenditure class, item stratum, and          importance          Number of price                quotations          Number of
ELI)                                        for December              quotations        with price changes         substitutions
    Vehicle finance charges                          0.571                   2,373                    36.1                   68
         Unpriced other vehicle
         finance chargesb                                  a
                                                                                    0                   0.0                   0
    Vehicle rental, registration, and
    inspection                                       0.768                   5,757                    25.1                   89
      State and local automobile
      registration, license, and
      inspection                                     0.368                   1,874                      4.5                  15
                                                           a
         State automobile registration                                       1,354                      5.5                  10
                                                           a
         Local automobile registration                                         249                      2.4                   5
                                                           a
         Driver’s license                                                      227                      1.8                   0
                                                           a
         Vehicle inspection                                                     44                      0.0                   0
      Other automobile-related fees                  0.372                   3,883                    35.1                   74
                                                           a
         Automobile rental                                                   2,141                    54.3                   62
                                                           a
         Truck rental                                                          521                    29.8                    6
                                                           a
         Parking fees                                                          697                      4.5                   4
                                                           a
         Vehicle tolls                                                         464                      2.6                   1
                                                           a
         Automobile towing charges                                              60                      0.0                   1
                                                           a
         Other vehicle rental                                                       0                   0.0                   0
      Unpriced docking and landing
      feesb                                          0.027                          0                   0.0                   0
    Public transportation                            1.566                  11,248                    45.5                   73
      Airline fare                                   1.037                   6,366                    69.5                    0
      Other intercity transportation                 0.139                   2,318                    26.2                   59
                                                           a
         Intercity bus fare                                                    407                    35.9                    4
                                                           a
         Intercity train fare                                                1,154                    12.0                    1
                                                           a
         Ship fares                                                            757                    42.7                   54
      Intracity transportation                       0.378                   2,564                      3.6                  14
                                                           a
         Intracity mass transit                                              2,040                      2.2                  13
                                                           a
         Taxi fare                                                             524                      9.2                   1
                                                           a
         Car and van pools                                                          0                   0.0                   0
      Unpriced school busb                           0.012                          0                   0.0                   0
  Medical care component total                       7.426                  50,237                     n/a                1,116
    Prescription drugs and medical
    supplies                                         0.897                   4,471                    26.0                  108
    Nonprescription drugs and
    medical                                          0.383                   4,488                    19.3                  173




                                         Page 142                                             GAO/GGD-99-84 Consumer Price Index
                                    Appendix VIII
                                    CPI Price Quotations, Substitutions, and
                                    Methods of Adjustment for 1997




                       Substitutions not adjusted
    Percent of price         (comparable)                                  Adjusted substitutions by method
quotations that were   Number not        Percent of          Number of                         Class mean
       substitutions     adjusted      substitutions       adjustments              Direct   and/or overlap        Linking
                 2.9             66              97.1                  2                 0               0               2

                 0.0              0                0.0                 0                 0               0               0

                 1.6             77              86.5                12                  2               0              10


                 0.8             12              80.0                  3                 2               0               1
                 0.7              8              80.0                  2                 1               0               1
                 2.0              4              80.0                  1                 1               0               0
                 0.0              0                0.0                 0                 0               0               0
                 0.0              0                0.0                 0                 0               0               0
                 1.9             65              87.8                  9                 0               0               9
                 2.9             57              91.9                  5                 0               0               5
                 1.2              4              66.7                  2                 0               0               2
                 0.6              2              50.0                  2                 0               0               2
                 0.2              1             100.0                  0                 0               0               0
                 1.7              1             100.0                  0                 0               0               0
                 0.0              0                0.0                 0                 0               0               0

                 0.0              0                0.0                 0                 0               0               0
                 0.7             52              71.2                21                  0               0              21
                 0.0              0                0.0                 0                 0               0               0
                 2.6             40              67.8                19                  0               0              19
                 1.0              2              50.0                  2                 0               0               2
                 0.1              0                0.0                 1                 0               0               1
                 7.1             38              70.4                16                  0               0              16
                 0.6             12              85.7                  2                 0               0               2
                 0.6             12              92.3                  1                 0               0               1
                 0.2              0                 0                  1                 0               0               1
                 0.0              0                0.0                 0                 0               0               0
                 0.0              0                0.0                 0                 0               0               0
                 2.2           355               31.8               761                336               3             422

                 2.4             49              45.4                59                 11               0              48

                 3.9             41              23.7               132                 43               0              89
                                                                                                                (continued)




                                   Page 143                                              GAO/GGD-99-84 Consumer Price Index
                                        Appendix VIII
                                        CPI Price Quotations, Substitutions, and
                                        Methods of Adjustment for 1997




Categories (name of component,                   Relative                                       Percent of
expenditure class, item stratum, and         importance          Number of price                quotations          Number of
ELI)                                       for December              quotations        with price changes         substitutions
      Internal and respiratory
      over-the-counter drugs                        0.245                   2,392                    19.9                   73
      Nonprescription medical
      equipment and supplies                        0.138                   2,096                    18.5                  100
                                                          a
         Topicals and dressings                                             1,282                    21.8                   62
         Medical equipment for
                                                          a
         general use                                                          181                    17.7                   16
         Supportive and convalescent
                                                          a
         medical equipment                                                    195                    14.9                   11
                                                          a
         Hearing aids                                                         438                    10.7                   11
                          b                               a
         Unpriced drugs                                                            0                   0.0                   0
    Professional services                           3.518                  18,430                      9.4                 210
      Physicians services                           1.904                   8,664                      8.7                 106
      Dental services                               1.107                   5,973                      9.6                  18
      Eyeglasses and eye care                       0.335                   2,081                    13.7                   60
      Services by other medical
      professionals                                 0.172                   1,712                      7.2                  26
    Hospital and related services                   2.310                  22,848                    18.9                  625
      Hospital services                             2.159                  19,913                    19.5                  581
      Nursing home services                         0.145                   2,935                    14.9                   44
                     b
      Unpriced items                                0.006                          0                   0.0                   0
    Health insurancei                               0.318                          0                  n/a                   n/a
  Entertainment component total                     4.339                  32,985                     n/a                1,327
    Reading materials                               0.730                   8,891                      5.6                 195
      Newspapers                                    0.376                   4,964                      1.8                  11
      Magazines, periodicals, and
      books                                         0.354                   3,927                    10.4                  184
                                                          a
         Magazines                                                          1,840                      7.6                  23
         Books purchased through
                                                          a
         book clubs                                                           362                    14.4                   34
         Books not purchased through
                                                          a
         book clubs                                                         1,725                    12.5                  127
                              b
      Unpriced newsletters                          0.000                          0                   0.0                   0
    Sporting goods and equipment                    0.391                   4,303                    26.3                  384
      Sport vehicles, including
      bicycles                                      0.181                   1,658                    30.6                  170
         Outboard motors and
                                                          a
         powered sports vehicles                                            1,173                    31.6                  123
                                                          a
         Unpowered boats and trailers                                         136                    19.1                   11




                                        Page 144                                             GAO/GGD-99-84 Consumer Price Index
                                    Appendix VIII
                                    CPI Price Quotations, Substitutions, and
                                    Methods of Adjustment for 1997




                       Substitutions not adjusted
    Percent of price         (comparable)                                  Adjusted substitutions by method
quotations that were   Number not        Percent of          Number of                         Class mean
       substitutions     adjusted      substitutions       adjustments              Direct   and/or overlap        Linking

                 3.1             18              24.7                55                 25               0              30

                 4.8             23              23.0                77                 18               0              59
                 4.8             19              30.7                43                 13               0              30

                 8.8              0                0.0               16                  1               0              15

                 5.6              4              36.4                  7                 0               0               7
                 2.5              0                0.0               11                  4               0               7
                 0.0              0                0.0                 0                 0               0               0
                 1.1             68              32.4               142                 49               0              93
                 1.2             22              20.8                84                 28               0              56
                 0.3              7              38.9                11                  7               0               4
                 2.9             30              50.0                30                  8               0              22

                 1.5              9              34.6                17                  6               0              11
                 2.7           197               31.5               428                233               3             192
                 2.9           175               30.1               406                218               3             185
                 1.5             22              50.0                22                 15               0               7
                 0.0              0                0.0                 0                 0               0               0
                 n/a            n/a               n/a                n/a               n/a              n/a             n/a
                 4.0           763               57.5               564                 95             321             148
                 2.2           126               64.6                69                  9              55               5
                                                                                                          g
                 0.2              6              54.6                  5                 2               2               1

                 4.7           120               65.2                64                  7              53               4
                 1.3             13              56.5                10                  3               7g              0

                 9.4             17              50.0                17                  0              16g              1

                 7.4             90              70.9                37                  4              30g              3
                 0.0              0                0.0                 0                 0               0               0
                 8.9           247               64.3               137                  7             114              16

                10.3           146               85.9                24                  3              15               6

                10.5           108               87.8                15                  3               8g              4
                                                                                                          g
                 8.1             10              90.9                  1                 0               1               0
                                                                                                                (continued)


                                   Page 145                                              GAO/GGD-99-84 Consumer Price Index
                                         Appendix VIII
                                         CPI Price Quotations, Substitutions, and
                                         Methods of Adjustment for 1997




Categories (name of component,                    Relative                                       Percent of
expenditure class, item stratum, and          importance          Number of price                quotations          Number of
ELI)                                        for December              quotations        with price changes         substitutions
                                                           a
         Bicycles                                                              349                    31.8                   36
      Sports equipment                               0.210                   2,645                    23.6                  214
         Indoor, warm weather, and
                                                           a
         winter sports equipment                                             2,057                    23.0                  170
         Hunting, fishing, and camping
                                                           a
         equipment                                                             588                    25.5                   44
    Toys, hobbies, and other
    entertainment commodities                        0.803                   6,793                    22.1                  436
      Toys, hobbies, and music
      equipment                                      0.360                   2,997                    21.6                  269
                                                           a
         Toys, games, and hobbies                                            2,477                    20.4                  239
                                                           a
         Playground equipment                                                       8                 75.0                    3
         Music instruments and
                                                           a
         accessories                                                           512                    26.6                   27
      Photographic supplies and
      equipment                                      0.111                   1,162                    20.0                   62
                                                           a
         Film                                                                  636                    20.4                   27
         Photographic and darkroom
                                                           a
         supplies                                                               24                      8.3                   0
                                                           a
         Photographic equipment                                                502                    19.9                   35
      Pets and pet products                          0.323                   2,634                    23.7                  105
                                                           a
         Pet food                                                            1,726                    26.7                   56
         Purchase of pets, pet
                                                           a
         supplies, and accessories                                             908                    18.1                   49
      Unpriced souvenirs, fireworks,
      and optic goodsb                               0.010                          0                   0.0                   0
    Entertainment services                           2.415                  12,998                    14.4                  312
      Club memberships dues and
      fees                                           0.346                   2,363                    12.1                   35
      Fees for participant sports                    0.400                   2,211                    17.2                   38
      Admissions                                     0.726                   4,014                    17.9                  149
         Admission to movies,
                                                           a
         theaters, and concerts                                              3,323                    14.1                   98
                                                           a
         Admission to sporting events                                          691                    36.5                   51
      Fees for lessons or instructions               0.256                     888                    11.5                   38
      Photographers, film procession,
      and pet services                               0.665                   3,522                    10.9                   52
                                                           a
         Photographer fees                                                     197                    18.8                    8
                                                           a
         Film processing                                                       710                    10.6                    9
                                                           a
         Pet services                                                          291                      8.6                   0



                                         Page 146                                             GAO/GGD-99-84 Consumer Price Index
                                    Appendix VIII
                                    CPI Price Quotations, Substitutions, and
                                    Methods of Adjustment for 1997




                       Substitutions not adjusted
    Percent of price         (comparable)                                  Adjusted substitutions by method
quotations that were   Number not        Percent of          Number of                         Class mean
       substitutions     adjusted      substitutions       adjustments              Direct   and/or overlap        Linking
                                                                                                          g
                10.3             28              77.8                  8                 0               6               2
                 8.1           101               47.2               113                  4              99              10

                 8.3             77              45.3                93                  3              83g              7

                 7.5             24              54.6                20                  1              16g              3

                 6.4           265               60.8               171                  8             152              11

                 9.0           160               59.5               109                  4              99               6
                                                                                                          g
                 9.7           145               60.7                94                  3              86               5
                37.5              0                0.0                 3                 0               3g              0

                 5.3             15              55.6                12                  1              10g              1

                 5.3             46              74.2                16                  1              13               2
                                                                                                          g
                 4.3             20              74.1                  7                 1               5               1

                 0.0              0                0.0                 0                 0               0g              0
                 7.0             26              74.3                  9                 0               8g              1
                 4.0             59              56.2                46                  3              40               3
                 3.2             29              51.8                27                  2              22g              3

                 5.4             30              61.2                19                  1              18g              0

                 0.0              0                0.0                 0                 0               0               0
                 2.4           125               40.1               187                 71               0             116

                 1.5             10              28.6                25                  6               0              19
                 1.7             19              50.0                19                  4               0              15
                 3.7             79              53.0                70                 28               0              42

                 3.0             68              69.4                30                  6               0              24
                 7.4             11              21.6                40                 22               0              18
                 4.3              3                7.9               35                 22               0              13

                 1.5             14              26.9                38                 11               0              27
                 4.1              2              25.0                  6                 0               0               6
                 1.3              1              11.1                  8                 2               0               6
                 0.0              0                0.0                 0                 0               0               0
                                                                                                                (continued)


                                   Page 147                                              GAO/GGD-99-84 Consumer Price Index
                                             Appendix VIII
                                             CPI Price Quotations, Substitutions, and
                                             Methods of Adjustment for 1997




Categories (name of component,                        Relative                                       Percent of
expenditure class, item stratum, and              importance          Number of price                quotations          Number of
ELI)                                            for December              quotations        with price changes         substitutions
                                                               a
         Veterinarian services                                                   1,106                    11.8                   15
                                                               a
         Other entertainment services                                            1,218                      9.7                  20
      Unpriced rental of recreational
      vehiclesb                                          0.023                          0                   0.0                   0
  Other goods and services
  component total                                        7.390                  22,440                     n/a                  422
    Tobacco products                                     1.688                   4,120                    27.3                   15
                                                               a
         Cigarettes                                                              3,710                    28.8                   10
         Tobacco products other than
                                                               a
         cigarettes                                                                356                    14.3                    3
                                                               a
         Smoking accessories                                                        54                    11.1                    2
         Unpriced smoking productsb                            a
                                                                                        0                   0.0                   0
    Toilet goods and personal care
    appliances                                           0.589                   3,653                    19.9                  129
      Cosmetics, bath, and nail
      preparations and implements                        0.263                   1,214                    16.8                   50
      Hair, dental, shaving, and
      miscellaneous personal care
      products                                           0.325                   2,439                    21.4                   79
                                                               a
         Products for the hair                                                   1,022                    21.8                   29
                                                               a
         Nonelectric articles for the hair                                          72                    12.5                    5
                                                               a
         Women’s hair pieces and wigs                                                   0                   0.0                   0
         Dental products and
                                                               a
         nonelectric dental articles                                               528                    23.5                   17
         Shaving products and
                                                               a
         nonelectric shaving articles                                              158                    24.1                    6
         Deodorant and suntan
         preparations and sanitary and
                                                               a
         footcare products                                                         580                    19.1                   15
         Electric personal care
                                                               a
         appliances                                                                 79                    22.8                    7
    Personal care services                               0.564                   5,576                      4.1                  16
      Beauty parlor services for
      females                                            0.447                   3,679                      4.2                  15
      Haircuts and other barber shop
      services for men                                   0.116                   1,897                      4.0                   1
      Unpriced repair of personal care
      appliancesb                                        0.000                          0                   0.0                   0
    School books and supplies                            0.273                   1,176                    25.7                   93
      School books and supplies for
      college                                            0.194                     637                    37.1                   69



                                             Page 148                                             GAO/GGD-99-84 Consumer Price Index
                                    Appendix VIII
                                    CPI Price Quotations, Substitutions, and
                                    Methods of Adjustment for 1997




                       Substitutions not adjusted
    Percent of price         (comparable)                                  Adjusted substitutions by method
quotations that were   Number not        Percent of          Number of                         Class mean
       substitutions     adjusted      substitutions       adjustments              Direct   and/or overlap        Linking
                 1.4              2              13.3                13                  4               0               9
                 1.6              9              45.0                11                  5               0               6

                 0.0              0                0.0                 0                 0               0               0

                 1.9           215               50.9               207                 40              52             115
                 0.4              5              33.3                10                  2               0               8
                 0.3              2              20.0                  8                 2               0               6

                 0.8              1              33.3                  2                 0               0               2
                 3.7              2             100.0                  0                 0               0               0
                 0.0              0                0.0                 0                 0               0               0

                 3.5             63              48.8                66                  5               0              61

                 4.1             26              52.0                24                  2               0              22


                 3.2             37              46.8                42                  3               0              39
                 2.8             11              37.9                18                  0               0              18
                 6.9              2              40.0                  3                 0               0               3
                 0.0              0                0.0                 0                 0               0               0

                 3.2              9              52.9                  8                 2               0               6

                 3.8              1              16.7                  5                 1               0               4


                 2.6              9              60.0                  6                 0               0               6

                 8.9              5              71.4                  2                 0               0               2
                 0.3              8              50.0                  8                 0               0               8

                 0.4              7              46.7                  8                 0               0               8

                 0.1              1               100                  0                 0               0               0

                 0.0              0                0.0                 0                 0               0               0
                 7.9             35              37.6                58                  2              52               4

                10.8             19              27.5                50                  2              45g              3
                                                                                                                (continued)


                                   Page 149                                              GAO/GGD-99-84 Consumer Price Index
                                        Appendix VIII
                                        CPI Price Quotations, Substitutions, and
                                        Methods of Adjustment for 1997




Categories (name of component,                   Relative                                       Percent of
expenditure class, item stratum, and         importance          Number of price                quotations          Number of
ELI)                                       for December              quotations        with price changes         substitutions
      Reference books and
      elementary and high school
      books                                         0.064                     539                    12.2                   24
         Elementary and high school
                                                          a
         books and supplies                                                   124                    32.3                   10
         Encyclopedias and other sets
                                                          a
         of reference books                                                   415                      6.3                  14
      Unpriced miscellaneous school
      purchasesb                                    0.014                          0                   0.0                   0
    Daycare, tuition, and other
    school fees                                     2.863                   4,685                    23.7                  104
      College tuition and fees                      1.685                   1,334                    48.8                   69
      Elementary and high school
      tuition and fees                              0.519                     356                    48.0                    7
      Child daycare and nursery
      school                                        0.388                   1,665                      9.1                  10
      Other tuition and fees                        0.155                   1,330                    10.3                   18
      Unpriced miscellaneous school
      items, rentals, and other
      servicesb                                     0.116                          0                   0.0                   0
    Legal, financial, and funeral
    services                                        1.415                   3,230                    12.4                   65
      Legal fees                                    0.496                     943                      4.9                   7
      Personal financial services                   0.407                   1,047                    12.2                   23
                                                          a
         Safe deposit box rental                                              113                      6.2                   0
         Checking accounts and
                                                          a
         special check services                                               390                    12.3                   15
         Tax return preparation and
                                                          a
         other accounting fees                                                544                    13.4                    8
      Cemetery lots and funeral
      expenses                                      0.403                   1,240                    18.3                   35
                                                          a
         Funeral expenses                                                     905                    20.8                   29
                                                          a
         Cemetery lots and crypts                                             335                    11.6                    6
      Unpriced miscellaneous
      personal servicesb                            0.109                          0                   0.0                   0




                                        Page 150                                             GAO/GGD-99-84 Consumer Price Index
                                    Appendix VIII
                                    CPI Price Quotations, Substitutions, and
                                    Methods of Adjustment for 1997




                       Substitutions not adjusted
    Percent of price         (comparable)                                  Adjusted substitutions by method
quotations that were   Number not        Percent of          Number of                         Class mean
       substitutions     adjusted      substitutions       adjustments              Direct   and/or overlap        Linking


                 4.5             16              66.7                  8                 0               7               1

                 8.1              2              20.0                  8                 0               7g              1

                 3.4             14             100.0                  0                 0               0g              0

                 0.0              0                0.0                 0                 0               0               0

                 2.2             70              67.3                34                 16               0              18
                 5.2             57              82.6                12                  6               0               6

                 2.0              5              71.4                  2                 1               0               1

                 0.6              6              60.0                  4                 2               0               2
                 1.4              2              11.1                16                  7               0               9


                 0.0              0                0.0                 0                 0               0               0

                  2              34              52.3                31                 15               0              16
                 0.7              2              28.6                  5                 1               0               4
                 2.2             11              47.8                12                  6               0               6
                  0               0                0.0                 0                 0               0               0

                 3.9              9              60.0                  6                 2               0               4

                 1.5              2              25.0                  6                 4               0               2

                 2.8             21              60.0                14                  8               0               6
                 3.2             18              62.1                11                  5               0               6
                 1.8              3              50.0                  3                 3               0               0

                 0.0              0                0.0                 0                 0               0               0




                                   Page 151                                              GAO/GGD-99-84 Consumer Price Index
Appendix VIII
CPI Price Quotations, Substitutions, and
Methods of Adjustment for 1997




Legend: n/a = Not applicable
a
    ELIs do not have relative importance assigned to them.
b
    The weight for this unpriced ELI is moved by changes in its expenditure class.
c
 This represents the lower bound of the number of direct adjustments. The upper bound of direct
adjustments of 6,042 represents the number of adjustments made to residential rent units. A unit
could have as many as three types of direct adjustments per collection period. The lower bound
of 5,641 excludes multiple adjustments made to one unit in a collection period.
d
    None of these are directly priced and therefore do not experience substitution.
e
    The residential rent units are also used for owners’ equivalent rent.
f
    The price quotations collected under tenants’ insurance are also used for household insurance.
g
    The class-mean method of adjustment can be used for this ELI.
h
 The price quotations collected under child daycare and nursery school are used to determine
babysitting.
i
 This expenditure class is moved by a combination of medical care price quotations and
insurance companies’ retained earnings, which BLS collects separately.

Source: BLS.




Page 152                                                        GAO/GGD-99-84 Consumer Price Index
Appendix IX

Comments From the Bureau of Labor
Statistics

Note: GAO comments
supplementing those in the
report text appear at the
end of this appendix.




                             Page 153   GAO/GGD-99-84 Consumer Price Index
                 Appendix IX
                 Comments From the Bureau of Labor
                 Statistics




See comment 1.




                 Page 154                            GAO/GGD-99-84 Consumer Price Index
                 Appendix IX
                 Comments From the Bureau of Labor
                 Statistics




See comment 2.




                 Page 155                            GAO/GGD-99-84 Consumer Price Index
                 Appendix IX
                 Comments From the Bureau of Labor
                 Statistics




See comment 2.




                 Page 156                            GAO/GGD-99-84 Consumer Price Index
               Appendix IX
               Comments From the Bureau of Labor
               Statistics




               The following are GAO’s comments on the Bureau of Labor Statistics’ letter
               dated April 7, 1999.


               1. The Commissioner listed several measures BLS had taken to improve the
GAO Comments   substitution and quality adjustment processes in the CPI, and stated that
               these measures worked to improve the accuracy of substitution handling.
               However, because no data has been collected in periodic evaluations,
               neither BLS nor we can assess what effects these measures have had on the
               accuracy of commodity analysts’ substitution handling.

               2. The Commissioner reported that the expert system software BLS
               investigated in 1993 was limited because it only allowed for ex post
               evaluations of commodity analysts’ decisions. Our recommendation,
               however, allows for the ex post evaluation of decisions and is therefore
               not affected by this limitation.




               Page 157                                    GAO/GGD-99-84 Consumer Price Index
Appendix X

Major Contributors to This Report


                        Martin de Alteriis, Evaluator-in-Charge
General Government      Kathleen Scholl, Supervisory Economist
Division, Washington,   Tony Assia, Senior Evaluator
D.C.                    Amber Roos, Student Intern

                        Loren Yager, Acting Chief Economist
Office of the Chief     Richard Krashevski, Supervisory Economist
Economist
                        James Turkett, Senior Evaluator
Dallas Office




                        Page 158                                  GAO/GGD-99-84 Consumer Price Index
Appendix X
Major Contributors to This Report




Page 159                            GAO/GGD-99-84 Consumer Price Index
Related GAO Products


              Bureau of Labor Statistics: Making the CPI More Reflective of Current
              Consumer Spending (GAO/T-GGD-98-115, Apr. 29, 1998).

              Consumer Price Index: More Frequent Updating of Market Basket
              Expenditure Weights Is Needed (GAO/GGD/OCE-98-2, Oct. 9, 1997).

              Consumer Price Index: Cost-of-Living Concepts and the Housing and
              Medical Care Components (GAO/GGD-96-166, Aug. 26, 1996).

              Economic Statistics: Status Report on the Initiative to Improve Economic
              Statistics (GAO/GGD-95-98, July 7, 1995).

              Economic Statistics: Measurement Problems Can Affect the Budget and
              Economic Policymaking (GAO/GGD-95-99, May 2, 1995).

              Prescription Drug Prices: Official Index Overstates Producer Price
              Inflation (GAO/HEHS-95-90, Apr. 28, 1995).

              Developing a Consumer Price Index for the Elderly (GAO/T-GGD-87-22,
              June 29, 1987).

              Stabilizing Social Security—Which Wage Measure Would Best Align
              Benefit Increases With Revenue Increases? (GAO/IMTEC-85-13, Aug. 27, 1985).

              Funds Needed to Develop CPI Quality Control System (GAO/GGD-83-32, Apr. 1,
              1983).

              A CPI for Retirees Is Not Needed Now but Could Be in the Future
              (GAO/GGD-82-41, June 1, 1982).

              A Consumer Price Index for Retirees and Alternatives for Controlling
              Indexing (Testimony, Apr. 20, 1982).

              Measurement of Homeownership Costs in the Consumer Price Index
              Should Be Changed (GAO/PAD-81-12, Apr. 16, 1981).

              Alternatives for Modifying the Indexation of Federal Programs
              (Testimony, Mar. 10, 1981).




(410197)      Page 160                                    GAO/GGD-99-84 Consumer Price Index
Ordering Information

The first copy of each GAO report and testimony is free.
Additional copies are $2 each. Orders should be sent to the
following address, accompanied by a check or money order
made out to the Superintendent of Documents, when
necessary. VISA and MasterCard credit cards are accepted, also.
Orders for 100 or more copies to be mailed to a single address
are discounted 25 percent.

Orders by mail:

U.S. General Accounting Office
P.O. Box 37050
Washington, DC 20013

or visit:

Room 1100
700 4th St. NW (corner of 4th and G Sts. NW)
U.S. General Accounting Office
Washington, DC

Orders may also be placed by calling (202) 512-6000
or by using fax number (202) 512-6061, or TDD (202) 512-2537.

Each day, GAO issues a list of newly available reports and
testimony. To receive facsimile copies of the daily list or any
list from the past 30 days, please call (202) 512-6000 using a
touchtone phone. A recorded menu will provide information on
how to obtain these lists.

For information on how to access GAO reports on the INTERNET,
send an e-mail message with "info" in the body to:

info@www.gao.gov

or visit GAO’s World Wide Web Home Page at:

http://www.gao.gov




PRINTED ON    RECYCLED PAPER
United States                       Bulk Rate
General Accounting Office      Postage & Fees Paid
Washington, D.C. 20548-0001           GAO
                                 Permit No. G100
Official Business
Penalty for Private Use $300

Address Correction Requested