Consumer Price Index: Impact of Commodity Analysts' Decisionmaking Needs
to Be Assessed (Letter Report, 06/15/99, GAO/GGD-99-84).

Pursuant to a congressional request, GAO examined the Bureau of Labor
Statistics' (BLS) decisions to substitute one product for another in its
computation of the Consumer Price Index (CPI), focusing on: (1) how
commodity analysts decide whether to make adjustments to the CPI; (2)
the adjustment methods they use; and (3) how supervisors of commodity
analysts review the analysts' decisions.

GAO noted that: (1) commodity analysts use a combination of professional
judgment, general procedures, specific methods, and limited written
guidance in deciding whether and how to make adjustments for
substitutions; (2) in making adjustments, BLS' objective is to include
only pure price change in the calculation of the CPI and eliminate price
change that is the result of other factors; (3) 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; (4) for items that are judged to be
comparable, no adjustment is made; (5) 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 the BLS computer programs make
the adjustments; (6) direct adjustments 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;
(7) when a direct adjustment's warranted but cannot be made, commodity
analysts apply one of two indirect adjustment methods that impute a rate
of price change; (8) both methods impute the pure price change by
averaging the rates of price changes experienced by the same type of
items in the CPI; (9) the class-mean method is generally used for
products where new models or product lines are introduced fairly
regularly; (10) it is based on the rate of price changes experienced by
other substitutions in the particular geographic area; (11) the linking
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; (12) according to BLS, there are no guidelines or
policies in writing for supervisors to follow in selecting and reviewing
the substitution decisions of the commodity analysts; (13) there is an
unwritten policy that supervisors are to review substitution decisions
when they consider the price increase or decrease to be too large; (14)
BLS has no policy to randomly or otherwise select and review
substitution decisions; and (15) GAO found no evidence to indicate
whether errors or inconsistencies in commodity analysts' decisions or
lack of comprehensive reviews has had a material effect on the
calculation of the CPI.

--------------------------- Indexing Terms -----------------------------

 REPORTNUM:  GGD-99-84
     TITLE:  Consumer Price Index: Impact of Commodity Analysts'
	     Decisionmaking Needs to Be Assessed
      DATE:  06/15/99
   SUBJECT:  Statistical data
	     Price indexes
	     Econometric modeling
	     Prices and pricing
	     Internal controls
	     Statistical methods
	     Economic analysis
IDENTIFIER:  Consumer Price Index

******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO report.  Delineations within the text indicating chapter **
** titles, headings, and bullets are preserved.  Major          **
** divisions and subdivisions of the text, such as Chapters,    **
** Sections, and Appendixes, are identified by double and       **
** single lines.  The numbers on the right end of these lines   **
** indicate the position of each of the subsections in the      **
** document outline.  These numbers do NOT correspond with the  **
** page numbers of the printed product.                         **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
** A printed copy of this report may be obtained from the GAO   **
** Document Distribution Center.  For further details, please   **
** send an e-mail message to:                                   **
**                                                              **
**                                            **
**                                                              **
** with the message 'info' in the body.                         **
******************************************************************

Cover
================================================================ COVER

Congressional Requesters

June 1999

CONSUMER PRICE INDEX - IMPACT OF
COMMODITY ANALYSTS' DECISIONMAKING
NEEDS TO BE ASSESSED

GAO/GGD-99-84

Consumer Price Index

(410197)

Abbreviations
=============================================================== ABBREV

  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

Letter
=============================================================== LETTER

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 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. 

--------------------
\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. 

   RESULTS IN BRIEF
------------------------------------------------------------ Letter :1

Commodity analysts use a combination of professional judgment,
general 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, 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 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. 

   BACKGROUND
------------------------------------------------------------ Letter :2

Every month, usually by the middle of the month, BLS, which is a part
of 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 area.\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 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. 

--------------------
\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. 

   SCOPE AND METHODOLOGY
------------------------------------------------------------ Letter :3

For our first and second objectives--to describe how commodity
analysts decide whether to make adjustments and the adjustment
methods they 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
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 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. 

--------------------
\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). 

\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. 

   DESCRIPTION OF HOW ANALYSTS
   DECIDE TO MAKE ADJUSTMENTS AND
   THE ADJUSTMENT METHODS
------------------------------------------------------------ Letter :4

BLS has two processes for reviewing changes in products and services
that are in the CPI:  one for residential rental units and another
for all other items in the CPI.  Although BLS does not substitute one
residential unit for 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 washer and dryer to the unit
since BLS last collected data, BLS would adjust 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. 

BLS has not documented the process it uses to review and adjust most
nonrent substitutions; but BLS officials have agreed that, although
there are probably some exceptions, the figure 1 flowchart reasonably
illustrates that process. 

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

   (See figure in printed
   edition.)

   (See figure in printed
   edition.)

\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. 

      COMMODITY REVIEW LISTINGS
      PROVIDE ANALYSTS WITH
      INFORMATION FOR
      DECISIONMAKING
---------------------------------------------------------- Letter :4.1

Commodity review listings (CRL) provide commodity analysts with
information for making various judgments about substitutions and are
the primary tool used by analysts in the process of reviewing changes
in products and services.  CRLs are generated each month from the
data that 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
      SUBSTITUTION IS ACCEPTABLE
---------------------------------------------------------- Letter :4.2

After the pool of substitutions has been identified, then the
commodity 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 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 NEW
      AND OLD VERSIONS ARE
      COMPARABLE
---------------------------------------------------------- Letter :4.3

For substitutions that are accepted, the next step commodity analysts
take is to judge whether the price change resulting from the
substitution can be 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 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
      DETERMINING WHICH METHOD TO
      USE
---------------------------------------------------------- Letter :4.4

When a substitution is not comparable with the item it replaced,
commodity analysts make a direct adjustment or cause an indirect
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, 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
                                               ---------------------------------
                                   Substituti
                        Number of     ons not
Major       Number of  substituti    adjusted      Direct      Class-
component       price         ons  (comparabl  adjustment        mean   Linking
s          quotations  (accepted)          e)           s    method\a   method
---------  ----------  ----------  ----------  ----------  ----------  ---------
Food and      459,635       6,485       3,640         107          31    2,707
 beverages
Housing\b     136,430       4,047       2,453         132       1,023     439
Apparel        76,736       9,797       6,598       1,223       1,796     180
 and
 upkeep
Transport      94,336       5,660       2,699       1,837         823     301
 ation
Medical        50,237       1,116         355         336           3     422
 care
Entertain      32,985       1,327         763          95         321     148
 ment
Other          22,440         422         215          40          52     115
 goods
 and
 services
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. 

         DIRECT ADJUSTMENTS
-------------------------------------------------------- Letter :4.4.1

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

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 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. 

--------------------
\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. 

         INDIRECT ADJUSTMENTS
-------------------------------------------------------- Letter :4.4.2

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
-------------------------------------------------------- Letter :4.4.3

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. 

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. 

         LINKING METHOD
         ADJUSTMENTS
-------------------------------------------------------- Letter :4.4.4

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. 

   DESCRIPTION OF HOW SUPERVISORS
   REVIEW COMMODITY ANALYSTS'
   DECISIONS
------------------------------------------------------------ Letter :5

BLS relies on supervisors to review the substitutions decisions made
by commodity analysts.  According to BLS, there are no guidelines or
policies in writing for supervisors to follow in selecting and
reviewing the decisions made by analysts.  In practice, according to
the BLS Branch Chief for Consumer Prices, there is an unwritten
policy that supervisors each 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 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. 

   CONCLUSIONS
------------------------------------------------------------ Letter :6

BLS faces a difficult task each month of collecting tens of thousands
of 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. 

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. 

   RECOMMENDATION
------------------------------------------------------------ Letter :7

To help ensure that the CPI is protected from potential effects of
errors 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. 

   AGENCY COMMENTS AND OUR
   EVALUATION
------------------------------------------------------------ Letter :8

In a letter dated April 7, 1999, the Commissioner of BLS provided
comments on a draft of this report for the Department of Labor.  The
Commissioner 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 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. 

---------------------------------------------------------- Letter :8.1

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. 

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

ADDITIONAL INFORMATION ON OUR
METHODOLOGY
=========================================================== Appendix I

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 analysts\1 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. 

--------------------
\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. 

   METHODOLOGY USED TO SELECT
   EXAMPLES, SUBSTITUTION
   DECISIONS, AND COMMODITY
   ANALYSTS
--------------------------------------------------------- Appendix I:1

We judgmentally selected a small number of examples to illustrate the
use of BLS' quality adjustment methods across the major components of
the Consumer Price Index (CPI).  We discussed these examples with the
commodity analysts who had reviewed them and requested that the
analysts consult their supervisors about our selection.  The analysts
and their supervisors agreed that the examples we had selected were
broadly 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 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 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. 

--------------------
\2 According to BLS data, there were 204 nonrent item strata, of
which 181 were priced. 

\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. 

   SURVEY METHODOLOGY
--------------------------------------------------------- Appendix I:2

To gain a better understanding of the processes the commodity
analysts 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 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. 

BACKGROUND INFORMATION ON THE
CONSUMER PRICE INDEX
========================================================== Appendix II

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 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. 

--------------------
\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. 

\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. 

   CONSTRUCTION OF THE CPI
-------------------------------------------------------- Appendix II:1

Construction of the CPI begins by selecting a collection of goods and
services that is usually bought by the reference population in the
index.  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, 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. 

--------------------
\3 The 1998 CPI revision was based on CEX data collected from 1993
through 1995, from about 36,000 individuals and families. 

      CLASSIFICATION OF MARKET
      BASKET ITEMS
------------------------------------------------------ Appendix II:1.1

BLS classified all CEX expenditure items for the 1987 CPI revision
into 206 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. 

--------------------
\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. 

      EXPENDITURE WEIGHTS OF
      MARKET BASKET ITEMS
------------------------------------------------------ Appendix II:1.2

Expenditure weights are used to give proportionate emphasis for price
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, 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. 

--------------------
\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. 

      RELATIVE IMPORTANCE OF
      MARKET BASKET ITEMS
------------------------------------------------------ Appendix II:1.3

Relative importance is related to, but not the same as, expenditure
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 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 1997

   (See figure in printed
   edition.)

Source:  BLS. 

      COLLECTING PRICES OF MARKET
      BASKET ITEMS
------------------------------------------------------ Appendix II:1.4

Each month, BLS price takers visit or call thousands of retail
stores, service 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). 

--------------------
\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. 

      REPLACEMENT OF MARKET BASKET
      ITEMS NO LONGER AVAILABLE
      FOR PRICING
------------------------------------------------------ Appendix II:1.5

Because the CPI uses a fixed market basket of goods and services, BLS
price takers are instructed to collect price information for the same
item (e.g., 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 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. 

HOW COMMODITY ANALYSTS DECIDE TO
MAKE ADJUSTMENTS
========================================================= Appendix III

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.)

   ROLE OF COMMODITY ANALYSTS
------------------------------------------------------- Appendix III:1

According to BLS, when receiving the price information that was
collected by the price takers and entered into BLS' computer system,
the commodity 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 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. 

--------------------
\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. 

   USE OF COMMODITY REVIEW
   LISTINGS
------------------------------------------------------- Appendix III:2

The primary tool used by commodity analysts in the process for
reviewing changes in products and services is the commodity review
listing (CRL).  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 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. 

   PROCESS OF MAKING DECISIONS
   ABOUT SUBSTITUTIONS
------------------------------------------------------- Appendix III:3

According to BLS officials, the process through which commodity
analysts make their decisions about substitutions begins when the
commodity analyst receives CRLs.  CRLs are generated for substitution
review usually 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 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. 

--------------------
\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. 

\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. 

      DETERMINING WHETHER
      SUBSTITUTIONS WERE
      IDENTIFIED
----------------------------------------------------- Appendix III:3.1

According to BLS officials, the commodity analysts review CRLs.  For
the price quotations that the price takers identified as
substitutions, the 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
      SUBSTITUTION IS ACCEPTABLE
----------------------------------------------------- Appendix III:3.2

According to BLS officials, when the commodity analysts review the
CRLs, they determine if the price that the price taker identified as
a 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. 

      DETERMINING THE NEED TO MAKE
      AN ADJUSTMENT
----------------------------------------------------- Appendix III:3.3

According to BLS officials, after substitutions are accepted, they go
through 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
         SUBSTITUTION IS
         COMPARABLE
--------------------------------------------------- Appendix III:3.3.1

When the analysts determine that the new version is comparable to the
old 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 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. 

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

         COMPARABLE SUBSTITUTION
         DECISION
--------------------------------------------------- Appendix III:3.3.2

The following example of a comparable substitution decision comes
from our interview with a commodity analyst's supervisor.  According
to the supervisor, this example is illustrative of an analyst's
comparability 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
   (UNDER THE CLUB MEMBERSHIP DUES
   AND FEES ITEM STRATUM)
------------------------------------------------------- Appendix III:4

A new version of club membership was substituted for the old version
and, as shown in table III.1, the "level of membership" differed. 
The analyst concluded that the two versions were comparable even with
the difference in type of membership. 

   Figure III.1:  Characteristics
   of Two Tennis Club Memberships

   (See figure in printed
   edition.)

Source:  BLS. 

      ANALYST'S REASONS FOR
      COMPARABLE DECISION
----------------------------------------------------- Appendix III:4.1

A field message from the price taker informed the commodity analyst
that 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
----------------------------------------------------- Appendix III:4.2

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
      DIFFERENCES
----------------------------------------------------- Appendix III:4.3

According to BLS officials, the commodity analyst, after deciding
that the substitution is acceptable and in conjunction with the
comparability determination, will determine if information is
available to do a direct 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 INCREASE COMPARABILITY
--------------------------------------------------- Appendix III:4.3.1

Previously, we said commodity analysts compared old and new versions
of 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
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. 

--------------------
\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. 

      USING CLASS-MEAN AND LINKING
      METHODS TO ADJUST FOR
      DIFFERENCES
----------------------------------------------------- Appendix III:4.4

In 1997, 29 percent of the nonrent substitutions were not comparable
and could not be directly adjusted for differences between the old
and new 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 is needed.  However, the
commodity analysts can overrule these computer-generated codes.  When
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. 

--------------------
\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. 

         APPLYING THE CLASS-MEAN
         METHOD
--------------------------------------------------- Appendix III:4.4.1

If the commodity analyst decides that the class mean is the
appropriate 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
--------------------------------------------------- Appendix III:4.4.2

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. 

   BLS PROCESSES AND PROCEDURES
   FOR REVIEWING SUBSTITUTION
   DECISIONS
------------------------------------------------------- Appendix III:5

BLS relies on supervisors to review the substitution decisions made
by commodity analysts.  Although there are no written guidelines for
supervisors to follow in selecting and reviewing the decisions made
by analysts, BLS officials stated that the supervisors follow
unwritten policies.  There is no unique set of procedures for
selecting and reviewing these 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
      SUBSTITUTIONS THAT RESULT IN
      LARGE PRICE CHANGES
----------------------------------------------------- Appendix III:5.1

According to BLS, supervisors are to review large price changes going
into the CPI each month, with the definition of large varying among
and within 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
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

--------------------
\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. 

      SUPERVISORY REVIEWS OCCUR
      AFTER DATA COLLECTION
      PERIODS END
----------------------------------------------------- Appendix III:5.2

For the purpose of collecting CPI prices, BLS divides each month into
three collection periods, which it calls pricing periods.  At the end
of each 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. 

--------------------
\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. 

      REVIEW ALSO OCCURS AFTER
      PRELIMINARY INDEXES ARE
      PRODUCED
----------------------------------------------------- Appendix III:5.3

After the supervisory review of price changes is complete, BLS
computer routines produce preliminary basic indexes for each item
stratum and 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 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. 

      NORMAL SCOPE OF SUPERVISORY
      REVIEW
----------------------------------------------------- Appendix III:5.4

The Chief of the Consumer Prices Branch stated that the supervisors
are 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. 

DIRECT ADJUSTMENTS
========================================================== Appendix IV

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. 

   MANUFACTURERS' COSTS
-------------------------------------------------------- Appendix IV:1

BLS uses manufacturers' cost information to make direct adjustments
for 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 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 MANUFACTURERS'
      COSTS
------------------------------------------------------ Appendix IV:1.1

Since 1967, BLS has asked the major domestic automobile and truck
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' 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. 

--------------------
\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. 

\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. 

      HOW QUALITY CHANGES ARE
      CALCULATED USING
      MANUFACTURERS' COSTS
------------------------------------------------------ Appendix IV:1.2

The CPI's commodity analyst for new cars uses the lists of quality
changes 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 model under review.  The analyst may find, for
example, that the engine of 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 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. 

--------------------
\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. 

      HOW PRICE CHANGES ARE
      CALCULATED AFTER MAKING AN
      ADJUSTMENT USING
      MANUFACTURERS' COSTS
------------------------------------------------------ Appendix IV:1.3

In the example above, an adjustment of $120 would be added to the
price of the previous year's model to make the new and old year's
models 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 ($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'
      COSTS
------------------------------------------------------ Appendix IV:1.4

The following example results from our interviews with a BLS
commodity 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. 

   EXAMPLE 1 - NEW CAR (UNDER THE
   NEW CARS ITEM STRATUM)
-------------------------------------------------------- Appendix IV:2

Although the specifications for the 1998 and 1997 models were alike,
the commodity analyst made an adjustment because the analyst had been
provided with information by the automaker that showed improvements
had been made to the 1998 model. 

   Figure IV.1:  Characteristics
   of Two Versions of a New Car

   (See figure in printed
   edition.)

Source:  BLS. 

      ANALYST'S REASONS FOR
      DECIDING THE ITEMS WERE
      COMPARABLE
------------------------------------------------------ Appendix IV:2.1

Even though the specifications of the old model and the new model
were unchanged, the analyst knew that changes had been made to the
new 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
      METHOD OF ADJUSTMENT
------------------------------------------------------ Appendix IV:2.2

To make the adjustment, the analyst had to override another
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 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 CHANGE
------------------------------------------------------ Appendix IV:2.3

The automaker had reported that the wholesale cost of the engine
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. 

   HEDONIC REGRESSION ESTIMATES
-------------------------------------------------------- Appendix IV:3

BLS first introduced the use of adjustments based on hedonic
regression estimates into the CPI in 1988 to adjust for the effect of
aging on housing 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 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. 

The 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
      REGRESSION ESTIMATES FOR
      ADJUSTMENTS
------------------------------------------------------ Appendix IV:3.1

Hedonic regression estimates are used to make direct adjustments to
the price of an item when it is replaced by a substitute item so that
BLS can 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. 

      HOW QUALITY CHANGES ARE
      CALCULATED USING HEDONIC
      REGRESSION MODELS
------------------------------------------------------ Appendix IV:3.2

For the period reviewed in our report (October and November 1997),
the application of hedonic techniques among nonrent items was limited
to 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. 

                  Table 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, 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. 

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

      HOW BLS ACCOUNTS FOR QUALITY
      CHANGE USING HEDONIC
      REGRESSION ESTIMATES
------------------------------------------------------ Appendix IV:3.3

BLS' exclusion of quality differences using the hedonic regression
estimates can be demonstrated with a hypothetical case based on a
woman's trench 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 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 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 PRICE
      CHANGE AFTER MAKING A
      HEDONIC REGRESSION-BASED
      ADJUSTMENT
------------------------------------------------------ Appendix IV:3.4

Following the hedonic regression adjustment described above, the old
item's adjusted price of $114.42 is compared with the new item's
price of $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 actual price difference between the two
versions of the coat, $14.42 was allocated to a quality difference;
and the remaining $10.58 was determined to represent a pure price
increase. 

      EXAMPLE OF BLS' USE OF THE
      HEDONIC METHOD
------------------------------------------------------ Appendix IV:3.5

The following example results from our interviews with a BLS
commodity 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. 

   EXAMPLE 2 - WOMAN'S COAT (UNDER
   THE WOMEN'S COATS AND JACKETS
   ITEM STRATUM)
-------------------------------------------------------- Appendix IV:4

A new version of heavyweight woman's coat was substituted for the old
version and, as shown in table IV.2, several important
characteristics of the two versions differed.  The analyst decided
that the two versions were not comparable, and made a direct
adjustment by using values from a hedonic regression. 

   Figure IV.2:  Characteristics
   of Two Versions of a Woman's
   Coat

   (See figure in printed
   edition.)

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

Source:  BLS. 

      ANALYST'S REASONS FOR
      DECIDING THE ITEMS WERE NOT
      COMPARABLE
------------------------------------------------------ Appendix IV:4.1

The commodity analyst noted that there were several major differences
between the old and new versions in the design, body fiber, and
method of cleaning.  The analyst said that BLS research had shown
that these differences had a significant effect on the price of
women's coats and 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
      METHOD OF ADJUSTMENT
------------------------------------------------------ Appendix IV:4.2

To make the adjustment, the analyst had to override another
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 CHANGE
------------------------------------------------------ Appendix IV:4.3

The analyst used the results of the hedonic regression to calculate
the 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. 

   OTHER DIRECT ADJUSTMENTS
-------------------------------------------------------- Appendix IV:5

A third method of direct price adjustments includes adjustments for
an item's size or the number of units it includes.  BLS considers
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 entry-level items. 

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 QUALITY
      CHANGES ASSOCIATED WITH
      DIFFERENCES IN UNIT COUNT OR
      SIZE, AND CORRECTIONS
------------------------------------------------------ Appendix IV:5.1

For purposes of measuring the price change of an item of constant
quality, the item should remain the same in unit count or size
between collection periods.  Before calculating the price change
between two versions of an item that have experienced a size or unit
count change, BLS adjusts the price of the old version to the level
it would have been if the old version had the same unit count or size
as the new version. 

      HOW BLS CALCULATES PRICE
      CHANGES AFTER DIRECT
      ADJUSTMENTS FOR DIFFERENCES
      IN UNIT COUNT OR SIZE, AND
      CORRECTIONS
------------------------------------------------------ Appendix IV:5.2

To illustrate a price calculation with an "other" direct adjustment
method, suppose for example, that a package of gum has five sticks in
one period 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 difference in unit count would make this
substitution not comparable.  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 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
      DIRECT ADJUSTMENTS FOR SIZE
      OR UNIT COUNT, AND
      CORRECTIONS
------------------------------------------------------ Appendix IV:5.3

The following two examples of BLS' use of direct adjustments for unit
count or size, or corrections come from our interviews with BLS
commodity analysts.  According to the commodity analysts and
supervisors, these examples are illustrative of the substitutions in
which 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. 

   EXAMPLE 3 - SEASON TICKET
   (UNDER THE ADMISSION TO
   SPORTING EVENTS ITEM STRATUM)
-------------------------------------------------------- Appendix IV:6

The new version of the season ticket included 44 hockey games, while
the old version included 43 games.  The analyst decided to make a
direct adjustment to account for the difference in the number of
games. 

   Figure IV.3:  Characteristics
   of Two Versions of a Season
   Ticket Package

   (See figure in printed
   edition.)

Source:  BLS. 

      ANALYST'S REASONS FOR
      DECIDING THE ITEMS WERE NOT
      COMPARABLE
------------------------------------------------------ Appendix IV:6.1

The commodity analyst decided that, because the number of tickets in
the season ticket package increased from 43 to 44 hockey games, a
direct 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
      METHOD OF ADJUSTMENT
------------------------------------------------------ Appendix IV:6.2

To make a direct adjustment in this case, the analyst overrode the
default 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 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 CHANGE
------------------------------------------------------ Appendix IV:6.3

The price calculation for the direct quality adjustment method is
done by 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. 

--------------------
\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. 

   EXAMPLE 4 - CANNED SOUP (UNDER
   THE CANNED AND PACKAGED SOUP
   ITEM STRATUM)
-------------------------------------------------------- Appendix IV:7

The analyst concluded that two soups were not comparable and made a
direct adjustment to correct for the size of the container. 

   Figure IV.4:  Characteristics
   of Two Versions of a Canned
   Soup

   (See figure in printed
   edition.)

Source:  BLS. 

      ANALYST'S REASONS FOR
      DECIDING THE ITEMS WERE NOT
      COMPARABLE
------------------------------------------------------ Appendix IV:7.1

The commodity analyst decided that the two versions were the same
soup, but a message from the price taker indicated that the size of
the container 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
      METHOD OF ADJUSTMENT
------------------------------------------------------ Appendix IV:7.2

To make a direct adjustment in this case, the analyst had to override
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. 

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 CHANGE
------------------------------------------------------ Appendix IV:7.3

The price calculation for the direct quality adjustment is done by
computer 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. 

CLASS-MEAN METHOD OF ADJUSTMENT
=========================================================== Appendix V

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

--------------------
\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. 

   BACKGROUND
--------------------------------------------------------- Appendix V:1

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

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. 

--------------------
\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. 

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

   HOW PRICE CHANGES ARE
   CALCULATED WITH THE CLASS-MEAN
   METHOD
--------------------------------------------------------- Appendix V:2

In the class-mean method, the replacement version is put aside in
calculating the price change.  Instead, a computerized procedure is
followed to impute a rate of change that is based on items similar to
the 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 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. 

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
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. 

--------------------
\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. 

   HOW QUALITY IMPROVEMENTS ARE
   ACCOUNTED FOR IN THE CLASS-MEAN
   METHOD
--------------------------------------------------------- Appendix V:3

Quality improvements are accounted for in the class-mean method
exactly as they are in the linking method (see app.  VI). 
Conceptually, under the class-mean method, BLS divides any difference
in price between the old 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 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. 

   EXAMPLES OF BLS' USE OF THE
   CLASS-MEAN METHOD
--------------------------------------------------------- Appendix V:4

The following three examples of BLS' use of the class-mean method
come from our interviews with three BLS commodity analysts.  See
appendix I for a detailed discussion of how these examples were
selected.  According to 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. 

   EXAMPLE 1 - BEDROOM LINENS
   (UNDER THE LINENS, CURTAINS,
   DRAPES, AND SEWING MATERIALS
   ITEM STRATUM)
--------------------------------------------------------- Appendix V:5

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

   Figure V.1:  Characteristics of
   Two Versions of a Dust Ruffle

   (See figure in printed
   edition.)

Source:  BLS. 

      ANALYST'S REASONS FOR
      DECIDING THE ITEMS WERE NOT
      COMPARABLE
------------------------------------------------------- Appendix V:5.1

The analyst said that the new version of the ruffle was different in
size from the old version because it was a queen-size dust ruffle, as
compared 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. 

--------------------
\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. 

      RATIONALE FOR SELECTING THE
      METHOD OF ADJUSTMENT
------------------------------------------------------- Appendix V:5.2

For not-comparable substitutions in this item stratum, BLS designated
the 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. 

      CALCULATION OF PRICE CHANGE
------------------------------------------------------- Appendix V:5.3

The price calculation for the class-mean method is done entirely by
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. 

--------------------
\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. 

   EXAMPLE 2 - WOMAN'S PARKA
   (UNDER THE WOMEN'S COATS AND
   JACKETS ITEM STRATUM)
--------------------------------------------------------- Appendix V:6

The analyst decided that the new version of a woman's parka was not
comparable with the old version.  The substitution was adjusted using
the class-mean method, which is the designated method of adjustment
for this item stratum when direct adjustments cannot be made. 

   Figure V.2:  Characteristics of
   Two Versions of a Woman's Parka

   (See figure in printed
   edition.)

Source:  BLS. 

      ANALYST'S REASONS FOR
      DECIDING THE ITEMS WERE NOT
      COMPARABLE
------------------------------------------------------- Appendix V:6.1

The analyst noted that there was a big difference in the prices of
the old and the new versions ($153.45 compared with $79.99).  The
analyst also 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
      METHOD OF ADJUSTMENT
------------------------------------------------------- Appendix V:6.2

For not-comparable substitutions in this item stratum, BLS has
designated the class-mean method as the standard method of adjustment
if a direct 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 CHANGE
------------------------------------------------------- Appendix V:6.3

The price calculation for the class-mean method is done entirely by
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. 

   EXAMPLE 3 - NEW CAR (UNDER THE
   NEW CARS ITEM STRATUM)
--------------------------------------------------------- Appendix V:7

The analyst decided that the new version of a new car was not
comparable with the old version.  The substitution was adjusted using
the class-mean method, which was the designated method for
not-comparable substitutions in this item stratum when direct
adjustments could not be made. 

   Figure V.3:  Characteristics of
   Two Versions of a New Car

   (See figure in printed
   edition.)

\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
      DECIDING THE ITEMS WERE NOT
      COMPARABLE
------------------------------------------------------- Appendix V:7.1

A 1998 model of a car had replaced the 1997 model.  According to the
commodity analyst, the characteristics of the new version, and
publicly 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. 

      RATIONALE FOR SELECTING THE
      METHOD OF ADJUSTMENT
------------------------------------------------------- Appendix V:7.2

In this case, the computer-generated default comparison code was
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 CHANGE
------------------------------------------------------- Appendix V:7.3

The price calculation for the class-mean method is done entirely by
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. 

LINKING METHOD OF ADJUSTMENT
========================================================== Appendix VI

BLS' linking method is an indirect adjustment method.  It uses
imputations\1 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. 

--------------------
\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. 

   BACKGROUND
-------------------------------------------------------- Appendix VI:1

According to BLS, the linking method is one of several adjustments
that can 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. 

   HOW PRICE CHANGES ARE
   CALCULATED WITH THE LINKING
   METHOD
-------------------------------------------------------- Appendix VI:2

In the linking method, the new version is put aside in calculating
the price change.  Instead, a procedure is followed to impute a rate
of change that is based on the remaining items priced in the
appropriate item stratum.  A 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 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. 

--------------------
\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. 

   HOW BLS ACCOUNTS FOR QUALITY
   CHANGE IN USING THE LINKING
   METHOD
-------------------------------------------------------- Appendix VI:3

When using the linking method, BLS makes the implicit assumption that
the pure price change is an amount that corresponds to the rate of
price change that occurred for the same type of items in the same
geographic area.  In other words, BLS assumes that the pure price
change is the price relative calculated through the linking method. 
BLS assumes that any 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 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. 

--------------------
\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). 

   BLS' USE OF THE LINKING METHOD
-------------------------------------------------------- Appendix VI:4

The following two examples of BLS' use of the linking method come
from our interviews with two BLS commodity analysts.  According to
the 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 change that was made.  Within each table,
differences in specifications between the old and new versions are
highlighted by shading. 

   EXAMPLE 1 - PACKAGED SOUP
   (UNDER CANNED AND PACKAGED SOUP
   ITEM STRATUM)
-------------------------------------------------------- Appendix VI:5

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

   Figure VI.1:  Characteristics
   of Two Versions of a Packaged
   Soup

   (See figure in printed
   edition.)

\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
      DECIDING ITEMS WERE NOT
      COMPARABLE
------------------------------------------------------ Appendix VI:5.1

The analyst said that the change in "other" major ingredients from
beans to herbs indicated a change in the item.  According to the
analyst, this was 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
      METHOD OF ADJUSTMENT
------------------------------------------------------ Appendix VI:5.2

Because the class-mean method had not been designated for the item
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 CHANGE
------------------------------------------------------ Appendix VI:5.3

The calculation of price change for the linking method is done
entirely by 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. 

   EXAMPLE 2 - MULTIVITAMIN
   TABLETS (UNDER INTERNAL,
   RESPIRATORY, AND
   OVER-THE-COUNTER DRUGS ITEM
   STRATUM)
-------------------------------------------------------- Appendix VI:6

A new version of multivitamin tablets was substituted for the old
version and, as shown in table VI.2, certain characteristics of the
two versions differed.  The analyst concluded that the old and new
version were not comparable and designated the linking method to
adjust the price of the old version. 

   Figure VI.2:  Characteristics
   of Two Versions of Multivitamin
   Tablets

   (See figure in printed
   edition.)

Source:  BLS. 

      ANALYST'S REASONS FOR
      DECIDING ITEMS WERE NOT
      COMPARABLE
------------------------------------------------------ Appendix VI:6.1

The commodity analyst decided that the two versions were not
comparable because there were so many important differences between
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
      METHOD OF ADJUSTMENT
------------------------------------------------------ Appendix VI:6.2

For substitutions that are not comparable in this item stratum, BLS
designated the linking method as the standard method of adjustment
when other methods are not usable.  In accordance with this
procedure, the 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 CHANGE
------------------------------------------------------ Appendix VI:6.3

The calculation of price change for the linking method is done
entirely by 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. 

MEASUREMENT OF RESIDENTIAL AND
HOMEOWNERS' EQUIVALENT RENTS
========================================================= Appendix VII

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. 

--------------------
\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). 

   BACKGROUND
------------------------------------------------------- Appendix VII:1

The CPI housing survey is used to estimate changes in rent for both
renters 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 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;

  -- 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. 

--------------------
\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. 

   ADJUSTMENTS FOR CHANGES IN
   UNITS
------------------------------------------------------- Appendix VII:2

Adjustments are made so that the same unit with the same features is
priced each time (every 6 months).  If a feature of the rental unit
has 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 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. 

   ADJUSTMENTS ARE MADE BY
   COMPUTER AND BY ANALYSTS
------------------------------------------------------- Appendix VII:3

In calendar year 1997, approximately 6,180 adjustments were made
because of changes in rental units.  Nearly all (91 percent) of the
adjustments were made automatically by computer routine without
direct 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
      COMPUTER
----------------------------------------------------- Appendix VII:3.1

The rent adjustments that are automatically identified and made by
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
----------------------------------------------------- Appendix VII:3.2

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. 

      CHANGES IN UTILITIES
----------------------------------------------------- Appendix VII:3.3

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

BLS calculates 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. 

--------------------
\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. 

      DEPRECIATION AND CHANGES IN
      STRUCTURE
----------------------------------------------------- Appendix VII:3.4

The dollar amounts for various structural changes (e.g., the addition
of a 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 COMMODITY
      ANALYSTS
----------------------------------------------------- Appendix VII:3.5

The housing commodity analysts make three types of adjustments: 
"pricing-links-cancel," "links-pause," and direct.  Of the three
types, the links-pause adjustment is by far the most frequently used. 
About 80 percent of the adjustments made by commodity analysts in
1997 were links-pause adjustments. 

         PRICING-LINKS-CANCEL
         ADJUSTMENT
--------------------------------------------------- Appendix VII:3.5.1

In the pricing-links-cancel adjustment, the analyst determines that
the unit 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
--------------------------------------------------- Appendix VII:3.5.2

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. 

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. 

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
--------------------------------------------------- Appendix VII:3.5.3

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. 

   ADDITION AND DELETION OF
   HOUSING UNITS
------------------------------------------------------- Appendix VII:4

The sample of housing units that BLS uses in calculating the CPI is
not static and represents the housing stock of the urban population. 
Rental units are 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. 

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. 

   ACCOUNTING FOR QUALITY CHANGE
------------------------------------------------------- Appendix VII:5

Whenever a substitution is not comparable to the CPI item it
replaced, BLS will make an adjustment to separate pure price change
from price changes 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 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. 

   EXAMPLES OF BLS' ADJUSTMENTS
   FOR RESIDENTIAL RENT AND REQ
------------------------------------------------------- Appendix VII:6

The following three examples of BLS' adjustments to residential rent
and REQ come from our interviews with three BLS housing commodity
analysts.  According to the commodity analysts and their section
chief, these are examples of the adjustments that are made in
residential rent and REQ.  One example is of an adjustment made by
computer routine without direct 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. 

   EXAMPLE 1 - AUTOMATIC
   ADJUSTMENT
------------------------------------------------------- Appendix VII:7

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

   Figure VII.1:  Characteristics
   of the Rental Unit

   (See figure in printed
   edition.)

Source:  BLS. 

      ANALYST'S REASONS FOR
      NOT-COMPARABLE DECISION
----------------------------------------------------- Appendix VII:7.1

BLS has developed computer routines to identify changes in a rental
unit's 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
      METHOD OF ADJUSTMENT
----------------------------------------------------- Appendix VII:7.2

Selecting a method of adjustment was not necessary. 

      CALCULATION OF PRICE CHANGE
----------------------------------------------------- Appendix VII:7.3

According to BLS, a computer routine automatically added 50 cents to
the 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. 

   EXAMPLE 2 -
   PRICING-LINKS-CANCEL ADJUSTMENT
------------------------------------------------------- Appendix VII:8

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

   Figure VII.2:  Characteristics
   of the Rental Unit

   (See figure in printed
   edition.)

Source:  BLS. 

      ANALYST'S REASONS FOR
      NOT-COMPARABLE DECISION
----------------------------------------------------- Appendix VII:8.1

A lengthy period of vacancy for the unit suggested to the commodity
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
      METHOD OF ADJUSTMENT
----------------------------------------------------- Appendix VII:8.2

The analyst first looked at the percentage change in rent to see if
it 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, 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 CHANGE
----------------------------------------------------- Appendix VII:8.3

The calculation of price change for the pricing-links-cancel method
is done 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. 

   EXAMPLE 3 - LINKS-PAUSE
   ADJUSTMENT
------------------------------------------------------- Appendix VII:9

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

   Figure VII.3:  Characteristics
   of the Rental Unit

   (See figure in printed
   edition.)

Source:  BLS. 

      ANALYST'S REASONS FOR
      NOT-COMPARABLE DECISION
----------------------------------------------------- Appendix VII:9.1

A rent decrease of 69 percent suggested to the commodity analyst that
the 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
      METHOD OF ADJUSTMENT
----------------------------------------------------- Appendix VII:9.2

The analyst used two criteria to determine which linking method to
use.  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 CHANGE
----------------------------------------------------- Appendix VII:9.3

BLS took two actions, 6 months apart, that affected the calculations
of 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 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. 

CPI PRICE QUOTATIONS,
SUBSTITUTIONS, AND METHODS OF
ADJUSTMENT FOR 1997
======================================================== Appendix VIII

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 change 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
quality 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., 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. 

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

                                                                                                                Substitutions not
                                                                                                              adjusted (comparable)          Adjusted substitutions by method
                                                                                                              ----------------------  ----------------------------------------------
Categories (name of                                            Percent of                   Percent of price                 Percent
component, expenditure           Relative       Number of      quotations                    quotations that      Number          of   Number of              Class mean
class, item stratum, and       importance           price      with price       Number of               were         not  substituti  adjustment                  and/or
ELI)                         for December      quotations         changes   substitutions      substitutions    adjusted         ons           s      Direct     overlap     Linking
-------------------------  --------------  --------------  --------------  --------------  -----------------  ----------  ----------  ----------  ----------  ----------  ----------
All items (with                  100.000%         918,561             n/a          35,061               3.8%         n/a         n/a      18,311       9,411       4,049       4,851
 residential rent units)
All items (without                 74.309         872,829             n/a          28,881                3.3      16,750       58.0%      12,131       3,770       4,049       4,312
 residential rent units)
====================================================================================================================================================================================
Food and beverages                 17.465         459,635             n/a           6,485                1.4       3,640        56.1       2,845         107          31       2,707
 component total
====================================================================================================================================================================================
Cereal and cereal                   0.449          19,574           25.0%             279                1.4         163        58.4         116           6           2         108
 products
Flour and prepared flour            0.075           6,507            27.0              90                1.4          57        63.3          33           2           1          30
 mixes
Flour                                  \a           2,027            29.0              14                0.7           9        64.3           5           0           1           4
Prepared flour mixes                   \a           4,480            26.1              76                1.7          48        63.2          28           2           0          26
Cereal                              0.272           6,516            23.8             116                1.8          65        56.0          51           3           1          47
Rice, pasta, and cornmeal           0.102           6,551            24.4              73                1.1          41        56.2          32           1           0          31
Rice                                   \a           3,085            23.2              37                1.2          23        62.2          14           0           0          14
Macaroni, similar                      \a           3,466            25.4              36                1.0          18        50.0          18           1           0          17
 products, and cornmeal
====================================================================================================================================================================================
Bakery products                     1.027          24,431            26.2             494                2.0         229        46.4       265 <        b>15           1         249
White bread                         0.260           6,249            23.3              78                1.3          45        57.7          33           2           0          31
Other breads, rolls,                0.239           6,124            24.3             142                2.3          67        47.2          75           2           0          73
 biscuits, and muffins
Bread other than white                 \a           3,538            27.8              70                2.0          29        41.4          41           0           0          41
Rolls, biscuits, and                   \a           2,586            19.5              72                2.8          38        52.8          34           2           0          32
 muffins, excluding
 frozen
Cakes, cupcakes, and                0.249           6,274            27.7             141                2.3          48        34.0          93           4           0          89
 cookies
Cakes and cupcakes,                    \a           2,433            19.5              64                2.6          17        26.6          47           0           0          47
 excluding frozen
Cookies                                \a           3,841            32.9              77                2.0          31        40.3          46           4           0          42
Other bakery products               0.279           5,784            29.9             133                2.3          69        51.9          64           7           1          56
Crackers                               \a           2,279            39.7              27                1.2          17        63.0          10           1           0           9
Bread and cracker                      \a             140            25.7               3                2.1           0         0.0           3           0           0           3
 products
Sweetrolls, coffee cake,               \a           1,404            18.1              47                3.4          19        40.4          28           2           1          25
 and doughnuts, excluding
 frozen
Frozen bakery products,                \a           1,241            26.8              22                1.8          10        45.5          12           2           0          10
 frozen and refrigerated
 doughs, and batters
Pies, tarts, and                       \a             720            27.9              34                4.7          23        67.6          11           2           0           9
 turnovers, excluding
 frozen
====================================================================================================================================================================================
Beef and veal                       0.933          47,442            46.4             312                0.7          87        27.9         225           0         7 <       b>218
Ground beef                         0.309           6,302            43.0              35                0.6          19        54.3          16           0           0          16
Chuck roast                         0.083           5,910            53.5              40                0.7           6        15.0          34           0           1          33
Round roast                         0.048           5,824            51.5              31                0.5           6        19.4          25           0           2          23
Round steak                         0.076           5,989            46.6              41                0.7           9        22.0          32           0           1          31
Sirloin steak                       0.072           5,924            47.3              33                0.6          12        36.4          21           0           1          20
Other steak, roast, and             0.345          17,493            43.0             132                0.8          35        26.5          97           0           2          95
 other beef
Other roasts, excluding                \a           2,521            40.1              23                0.9           5        21.7          18           0           1          17
 chuck and round
Other steak, excluding                 \a          10,900            46.4              74                0.7          21        28.4          53           0           1          52
 round and sirloin
Other beef                             \a           4,072            35.8              35                0.9           9        25.7          26           0           0          26
====================================================================================================================================================================================
Pork                                0.595          29,378            42.3             446                1.5         247        55.4         199           0           5         194
Bacon                               0.112           6,056            38.6              93                1.5          62        66.7          31           0           0          31
Pork chops                          0.137           6,249            46.9              18                0.3           5        27.8          13           0           2          11
Ham                                 0.138           5,298            44.6             179                3.4          95        53.1          84           0           0          84
Ham, excluding canned                  \a           5,013            46.3             173                3.5          93        53.8          80           0           0          80
Canned ham                             \a             285            14.0               6                2.1           2        33.3           4           0           0           4
Other pork, including               0.208          11,775            40.7             156                1.3          85        54.5          71           0           3          68
 sausage
Pork roasts, picnics, and              \a           6,718            44.8              88                1.3          45        51.1          43           0           3          40
 other pork
Pork sausage                           \a           5,057            35.1              68                1.3          40        58.8          28           0           0          28
====================================================================================================================================================================================
Other meats                         0.393          12,020            27.6             261                2.2         160        61.3         101           0           0         101
Frankfurters                           \a           2,627            35.4              55                2.1          35        63.6          20           0           0          20
Bologna, liverwurst, and               \a           2,852            24.5              43                1.5          20        46.5          23           0           0          23
 salami
Other lunchmeats,                      \a           5,521            26.3             143                2.6          90        62.9          53           0           0          53
 excluding bologna,
 liverwurst, and salami
Lamb, organ meats, and                 \a           1,020            23.1              20                2.0          15        75.0           5           0           0           5
 game
====================================================================================================================================================================================
Poultry                             0.439          17,348            36.9             463                2.7         398        86.0          65           0           2          63
Fresh whole chicken                 0.148           5,996            37.0             118                2.0         109        92.4           9           0           0           9
Fresh or frozen chicken             0.205           6,092            39.3              78                1.3          59        75.6          19           0           1          18
 parts
Other poultry                       0.086           5,260            34.0             267                5.1         230        86.1          37           0           1          36
====================================================================================================================================================================================
Fish and seafood                    0.373          12,036            32.1             215                1.8         124        57.7          91           3           1          87
Canned fish and seafood             0.072           6,469            26.4             102                1.6          67        65.7          35           3           0          32
Fresh or frozen fish and            0.301           5,567            38.7             113                2.0          57        50.4          56           0           1          55
 seafood
Shellfish, excluding                   \a           1,646            32.6              34                2.1          18        52.9          16           0           0          16
 canned
Fish, excluding canned                 \a           3,921            41.3              79                2.0          39        49.4          40           0           1          39
====================================================================================================================================================================================
Eggs                                0.187           6,500            63.1              38                0.6          22        57.9          16           0           0          16
====================================================================================================================================================================================
Fresh milk and cream                 0.61          12,970            34.5             125                1.0         102        81.6          23           0           0          23
Fresh whole milk                    0.352           6,545            35.1              56                0.9          46        82.1          10           0           0          10
Other fresh milk and                0.257           6,425            34.0              69                1.1          56        81.2          13           0           0          13
 cream
====================================================================================================================================================================================
Processed dairy products            0.608          19,066            31.6             308                1.6         180        58.4         128           0           3         125
Cheese                              0.338           6,380            31.5             123                1.9          72        58.5          51           0           1          50
Ice cream and related               0.157           6,396            29.0              96                1.5          56        58.3          40           0           2          38
 products
Other dairy products,               0.113           6,290            34.4              89                1.4          52        58.4          37           0           0          37
 including butter
Butter                                 \a           2,152            50.0              21                1.0          12        57.1           9           0           0           9
Other dairy products                   \a           4,138            26.3              68                1.6          40        58.8          28           0           0          28
====================================================================================================================================================================================
Fresh fruits                        0.740          43,203            51.3             118                0.3          80        67.8          38           0           0          38
Apples                              0.116           8,460            41.5              21                0.3          14        66.7           7           0           0           7
Bananas                             0.067           6,493            37.8               1                0.0           0         0.0           1           0           0           1
Oranges, including                  0.084           7,531            53.9              40                0.5          32        80.0           8           0           0           8
 tangerines
Other fresh fruits                  0.474          20,719            58.6              56                0.3          34        60.7          22           0           0          22
====================================================================================================================================================================================
Fresh vegetables                    0.631          31,291            53.4              62                0.2          43        69.4          19           0           0          19
Potatoes                            0.100           6,320            46.3              23                0.4          15        65.2           8           0           0           8
Lettuce                             0.077           6,396            55.3               1                0.0           0         0.0           1           0           0           1
Tomatoes                            0.115           6,204            66.4              18                0.3          16        88.9           2           0           0           2
Other fresh vegetables              0.339          12,371            49.5              20                0.2          12        60.0           8           0           0           8
====================================================================================================================================================================================
Processed fruits                    0.349          19,567            28.3             354                1.8         242        68.4         112           0           1         111
Fruit juices and frozen             0.273          12,963            33.0             229                1.8         145        63.3          84           0           1          83
 fruit
Frozen orange juice                    \a           2,415            33.0              21                0.9          13        61.9           8           0           0           8
Other frozen fruits and                \a           1,154            26.8              12                1.0           6        50.0           6           0           0           6
 fruit juices
Fresh, canned, and/or                  \a           9,394            31.5             196                2.1         126        64.3          70           0           1          69
 bottled fruit juices
Canned and dried fruits             0.077           6,604            22.2             125                1.9          97        77.6          28           0           0          28
====================================================================================================================================================================================
Processed vegetables                0.261          13,020            26.6             200                1.5         146        73.0          54           0           3          51
Frozen vegetables                   0.087           6,434            29.5             104                1.6          76        73.1          28           0           0          28
Processed vegetables,               0.174           6,586            23.9              96                1.5          70        72.9          26           0           3          23
 excluding frozen
Canned beans other than                \a             926            24.3              10                1.1           7        70.0           3           0           0           3
 lima
Canned cut corn                        \a             854            28.9               9                1.1           6        66.7           3           0           0           3
Other processed                        \a           4,806            22.9              77                1.6          57        74.0          20           0           3          17
 vegetables
====================================================================================================================================================================================
Sugar and sweets                    0.332          11,935            19.2             187                1.6         104        55.6          83           0           0          83
Sugar and artificial                0.085           5,810            21.8              74                1.3          38        51.4          36           0           0          36
 sweeteners
Sweets, including candy             0.247           6,125            16.7             113                1.8          66        58.4          47           0           0          47
Candy and chewing gum                  \a           4,675            14.8              91                2.0          54        59.3          37           0           0          37
Other sweets, excluding                \a           1,450            22.8              22                1.5          12        54.6          10           0           0          10
 candy and gum
====================================================================================================================================================================================
Fats and oils                       0.241          13,191            24.6             137                1.0          97        70.8          40           0           0          40
Margarine                              \a           1,999            25.7              22                1.1          18        81.8           4           0           0           4
Other fats and oils                    \a           7,583            23.3              83                1.1          55        66.3          28           0           0          28
Nondairy cream                         \a           1,571            22.3              13                0.8          11        84.6           2           0           0           2
 substitutes
Peanut butter                          \a           2,038            29.9              19                0.9          13        68.4           6           0           0           6
====================================================================================================================================================================================
Nonalcoholic beverages              0.747          19,262            31.7             267                1.4         169        63.3          98           9           2          87
Carbonated drinks                   0.348           6,639            35.3              69                1.0          46        66.7          23           3           0          20
Cola drinks                            \a           4,046            38.5              26                0.6          17        65.4           9           0           0           9
Carbonated drinks other                \a           2,593            30.2              43                1.7          29        67.4          14           3           0          11
 than cola
Coffee                              0.263           6,314            37.8              83                1.3          50        60.2          33           3           1          29
Roasted coffee                         \a           3,685            44.3              50                1.4          34        68.0          16           1           0          15
Instant and freeze dried               \a           2,629            28.8              33                1.3          16        48.5          17           2           1          14
 coffee
Other noncarbonated                 0.136           6,309            21.9             115                1.8          73        63.5          42           3           1          38
 drinks
Noncarbonated fruit-                   \a           2,079            21.8              34                1.6          20        58.8          14           0           0          14
 flavored drinks
Tea                                    \a           1,856            22.0              23                1.2          15        65.2           8           1           1           6
Other noncarbonated                    \a           2,374            22.0              58                2.4          38        65.5          20           2           0          18
 drinks
====================================================================================================================================================================================
Other prepared foods                1.046          31,997            26.9             506                1.6         273        54.0         233          11           3         219
Canned and packaged soup            0.096           6,504            25.5              90                1.4          47        52.2          43           1           1          41
Frozen prepared food                0.165           6,185            36.1             146                2.4          79        54.1          67           4           1          62
Frozen prepared meals                  \a           2,213            37.1              69                3.1          39        56.5          30           3           0          27
Frozen prepared foods                  \a           3,972            35.5              77                1.9          40        52.0          37           1           1          35
 other than meals
Snacks                              0.205           6,321            24.7             124                2.0          62        50.0          62           0           0          62
Potato chips and other                 \a           5,098            25.2             101                2.0          50        49.5          51           0           0          51
 snacks
Nuts                                   \a           1,223            22.7              23                1.9          12        52.2          11           0           0          11
Seasonings, condiments,             0.282           6,539            22.6              77                1.2          35        45.5          42           3           0          39
 sauces, and spices
Salt, other seasoning,                 \a           1,447            15.6              16                1.1           8        50.0           8           2           0           6
 and spices
Olives, pickles, and                   \a             713            26.0              11                1.5           6        54.6           5           0           0           5
 relishes
Sauces and gravies                     \a           3,269            26.2              41                1.3          15        36.6          26           1           0          25
Other condiments,                      \a           1,110            19.2               9                0.8           6        66.7           3           0           0           3
 excluding olives,
 pickles, and relishes
Miscellaneous prepared              0.299           6,448            25.9              69                1.1          50        72.5          19           3           1          15
 food, including baby
 food
Canned or packaged salads              \a             638            27.4              12                1.9          11        91.7           1           0           0           1
 and desserts
Baby food                              \a           1,389            22.0              10                0.7           8        80.0           2           1           0           1
Other canned or packaged               \a           4,421            26.8              47                1.1          31        66.0          16           2           1          13
 prepared foods
====================================================================================================================================================================================
Food away from home                 5.923          60,536            13.9           1,336                2.2         533        39.9         803          41           1         761
Lunch                               2.097          21,256            13.6             431                2.0         175        40.6         256          10           0         246
Dinner                              2.512          27,850            14.0             665                2.4         279        42.0         386          19           0         367
Other meals and snacks              1.004          11,430            14.0             240                2.1          79        32.9         161          12           1         148
Snacks and nonalcoholic                \a           8,413            13.9             209                2.5          68        32.5         141          10           1         130
 beverages
Breakfast or brunch                    \a           3,017            14.2              31                1.0          11        35.5          20           2           0          18
Unpriced board and                  0.310               0             0.0               0                0.0           0         0.0           0           0           0           0
 catered affairs\b
====================================================================================================================================================================================
Alcoholic beverages                 0.813          14,868            23.7             377                2.5         241        63.9         136          22           0         114
Beer, ale, and other                0.415           3,919            30.3              64                1.6          36        56.3          28           5           0          23
 alcoholic malt beverages
 at home
Distilled spirits                   0.209           3,803            23.1              27                0.7          17        63.0          10           2           0           8
Whiskey at home                        \a           1,553            25.3               9                0.6           5        55.6           4           1           0           3
Distilled spirits at                   \a           2,250            21.5              18                0.8          12        66.7           6           1           0           5
 home, excluding whiskey
Wine at home                        0.189           3,844            28.6             194                5.1         145        74.7          49           2           0          47
Alcoholic beverages away            0.765           3,302            10.8              92                2.8          43        46.7          49          13           0          36
 from home
Beer, ale, and other                   \a           1,333             9.7              30                2.3          12        40.0          18           6           0          12
 alcoholic malt beverages
 away from home
Wine away from home                    \a             834            10.4              42                5.0          21        50.0          21           4           0          17
Distilled spirits away                 \a           1,135            12.3              20                1.8          10        50.0          10           3           0           7
 from home
====================================================================================================================================================================================
Housing component total            41.469         182,162             n/a          10,254                5.6         n/a         n/a       7,774       5,773       1,023         978
====================================================================================================================================================================================
Renters' costs                      8.169          70,792             n/a           6,524                9.8         n/a         n/a       6,228       5,658           0         570
Rent, residential                   5.810          45,732             n/a           6,180               13.5         n/a         n/a       6,180     5,641\c           0         539
Lodging while out of town           2.089          22,889            39.0             297                1.3         264        88.9          33          12           0          21
Lodging while at school             0.236             390            47.4              24                6.2          12        50.0          12           5           0           7
Tenants' insurance                  0.034           1,781            10.1              23                1.3          20        87.0           3           0           0           3
====================================================================================================================================================================================
Rental equivalence and             20.269              \d              \d              \d                 \d          \d          \d          \d          \d          \d          \d
 household insurance
Owners' equivalent rent            19.881              \e              \e              \e                 \e          \e          \e          \e          \e          \e          \e
Household insurance                 0.388              \f              \f              \f                 \f          \f          \f          \f          \f          \f          \f
====================================================================================================================================================================================
Maintenance and repair              0.202             457            19.3              17                3.7           2        11.8          15           1           0          14
 services
Property maintenance and            0.126             457            19.3              17                3.7           2        11.8          15           1           0          14
 repair services
Inside home maintenance                \a             381            17.3               8                2.1           2        25.0           6           0           0           6
 and repair services
Repair and/or replacement              \a               0             0.0               0                0.0           0         0.0           0           0           0           0
 of hard surface flooring
Replacement of installed               \a              76            29.0               9               11.8           0         0.0           9           1           0           8
 wall-to-wall carpet
Repair of disposal,                    \a               0             0.0               0                0.0           0         0.0           0           0           0           0
 built-in dishwasher, and
 range hood
====================================================================================================================================================================================
Maintenance and repair              0.076           1,637            15.6              57                3.5          37        64.9          20           2           0          18
 commodities
Materials, supplies, and            0.034             825            19.5              29                3.5          20        69.0           9           1           0           8
 equipment for home
 repairs
Paint, wallpaper, and                  \a             408            20.1              13                3.2           8        61.5           5           0           0           5
 supplies
Tools and equipment for                \a              61            19.7               4                6.6           3        75.0           1           0           0           1
 painting
Lumber, paneling, wall                 \a             179            25.1               4                2.2           3        75.0           1           1           0           0
 and ceiling tile,
 awnings, and glass
Blacktop and masonry                   \a              23            17.4               0                0.0           0         0.0           0           0           0           0
 materials
Plumbing supplies and                  \a              67             6.0               1                1.5           1       100.0           0           0           0           0
 equipment
Electrical supplies and                \a              87            16.1               7                8.1           5        71.4           2           0           0           2
 heating and cooling
 equipment
Other maintenance and               0.042             812            11.7              28                3.4          17        60.7          11           1           0          10
 repair commodities
Miscellaneous supplies                 \a             609             9.2              22                3.6          15        68.2           7           1           0           6
 and equipment
Hard surface floor                     \a             136            16.2               4                2.9           1        25.0           3           0           0           3
 covering
Landscaping items                      \a              67            25.4               2                3.0           1        50.0           1           0           0           1
====================================================================================================================================================================================
Fuel oil and other fuels            0.368           5,297            43.0              23                0.4          12        52.2          11           6           0           5
Fuel oil                            0.254           2,383            55.4               9                0.4           4        44.4           5           2           0           3
Other household fuel                0.113           2,914            32.9              14                0.5           8        57.1           6           4           0           2
 commodities
Bottled or tank gas                    \a           2,073            35.4               8                0.4           4        50.0           4           2           0           2
Coal                                   \a               0             0.0               0                0.0           0         0.0           0           0           0           0
Other fuels                            \a             841            26.8               6                0.7           4        66.7           2           2           0           0
====================================================================================================================================================================================
Gas (piped) and                     3.401          28,152            51.4             123                0.4          98        79.7          25          14           0          11
 electricity
Electricity                         2.265          15,586            40.6              78                0.5          63        80.8          15          14           0           1
Utility natural gas                 1.136          12,566            64.9              45                0.4          35        77.8          10           0           0          10
 service
====================================================================================================================================================================================
Other utilities and                 3.246          28,653            13.4             260                0.9         173        66.5          87          54           0          33
 public services
Telephone services, local           1.115           6,450            12.7              26                0.4          17        65.4           9           0           0           9
 charges
Interstate toll calls               0.305           5,265            18.5              34                0.7           2         5.9          32          32           0           0
Intrastate toll calls               0.227           5,674             6.9              23                0.4          16        69.6           7           0           0           7
Water and sewerage                  0.795           4,112            10.9              42                1.0          34        81.0           8           6           0           2
 maintenance
Community antenna and               0.582           3,867            25.2              93                2.4          85        91.4           8           2           0           6
 cable television
Garbage and trash                   0.221           3,285             7.4              42                1.3          19        45.2          23          14           0           9
 collection
====================================================================================================================================================================================
Textile house furnishings           0.313           3,898            34.7             289                7.4         187        64.7         102           0          99           3
Bathroom linens                        \a             679            35.2              55                8.1          39        70.9          16           0        16\g           0
Bedroom linens                         \a           1,638            39.1             130                7.9          86        66.2          44           0        42\g           2
Kitchen and dining room                \a             220            23.2              22            10 \a 0          12        54.6          10           0         9\g           1
 linens
Curtains and drapes                    \a             648            36.6              42                6.5          24        57.1          18           0        18\g           0
Slipcovers and decorative              \a             106            23.6              12               11.3           6        50.0           6           0         6\g           0
 pillows
Sewing materials for                   \a             607            26.4              28                4.6          20        71.4           8           0         8\g           0
 household items
====================================================================================================================================================================================
Furniture and bedding               1.089           7,469            32.9             566                7.6         238        42.1         328           0         320           8
Bedroom furniture                   0.361           2,154            31.4             142                6.6          60        42.3          82           0          80           2
Mattress and springs                   \a             941            33.1              65                6.9          32        49.2          33           0        32\g           1
Bedroom furniture other                \a           1,213            30.2              77                6.4          28        36.4          49           0        48\g           1
 than mattress and
 springs
Sofas                               0.225           1,330            33.6             103                7.7          51        49.5          52           0        49\g           3
Living room chairs and              0.179           1,341            33.4             105                7.8          42        40.0          63           0          62           1
 tables
Living room chairs                     \a             871            36.6              84                9.6          35        41.7          49           0        49\g           0
Living room tables                     \a             470            27.5              21                4.5           7        33.3          14           0        13\g           1
Other furniture                     0.324           2,644            33.6             216                8.2          85        39.4         131           0         129           2
Kitchen and dining room                \a           1,193            34.5              84                7.0          27        32.1          57           0        56\g           1
 furniture
Infants' furniture                     \a             253            23.3              19                7.5          11        57.9           8           0         7\g           1
Outdoor furniture                      \a             201            43.8              39               19.4          24        61.5          15           0        15\g           0
Occasional furniture                   \a             997            33.0              74                7.4          23        31.1          51           0        51\g           0
====================================================================================================================================================================================
Television and sound                0.385           5,327            36.9             711               13.4         438        61.6         273           5         264           4
 equipment
Television sets                     0.128           2,170            43.3             306               14.1         191        62.4         115           1       113\g           1
Video products other than           0.065             683            32.9             108               15.8          76        70.4          32           3          28           1
 televisions
Video cassette recorders,              \a             376            41.0              63               16.8          44        69.8          19           0        19\g           0
 disc players, cameras,
 and accessories
Video cassettes and                    \a             236            18.2              28               11.9          22        78.6           6           3         3\g           0
 discs, blank and
 prerecorded
Video game hardware,                   \a              71            39.4              17               23.9          10        58.8           7           0         6\g           1
 software, and
 accessories
Audio components, radios,           0.192           2,474            32.3             297               12.0         171        57.6         126           1         123           2
 recordings, and other
 audio equipment
Radios, phonographs, and               \a             293            39.9              43               14.7          18        41.9          25           0        25\g           0
 tape recorders and
 players
Components and other                   \a             961            32.3             148               15.4          60        40.5          88           0        87\g           1
 sound equipment
Records and tapes,                     \a           1,220            17.8             106                8.7          93        87.7          13           1        11\g           1
 prerecorded and blank
Unpriced accessories for            0.000               0             0.0               0                0.0           0         0.0           0           0           0           0
 electronic equipment\b
====================================================================================================================================================================================
Household appliances                0.274           3,124            46.4             287                9.2         263        91.6          24           0          23           1
Refrigerators and home              0.084             819            47.6              79                9.7          72        91.1           7           0         7\g           0
 freezers
Laundry equipment                   0.092             820            47.0              60                7.3          59        98.3           1           0           1           0
Washers                                \a             507            48.3              43                8.5          43       100.0           0           0         0\g           0
Dryers                                 \a             313            44.7              17                5.4          16        94.1           1           0         1\g           0
Stoves, ovens, portable             0.098           1,485            45.5             148               10.0         132        89.2          16           0          15           1
 dishwashers, and window
 air conditioners
Stoves and ovens,                      \a             576            47.2              55                9.6          49        89.1           6           0         6\g           0
 excluding microwave
 ovens
Microwave ovens                        \a             658            45.7              80               12.2          74        92.5           6           0         5\g           1
Portable dishwashers                   \a              24            25.0               1                4.2           0         0.0           1           0         1\g           0
Window air conditioners                \a             227            42.3              12                5.3           9        75.0           3           0         3\g           0
====================================================================================================================================================================================
Information processing              0.064             649            42.4             119               18.3          67        56.3          52           0           0          52
 equipment
Personal computers and                 \a             375            57.6             103               27.5          59        57.3          44           0           0          44
 peripheral equipment
Computer software and                  \a              97            21.7               9                9.3           5        55.6           4           0           0           4
 accessories
Telephone, peripheral                  \a             125            26.4               6                4.8           2        33.3           4           0           0           4
 equipment, and
 accessories
Calculators, adding                    \a              41             4.9               1                2.4           1       100.0           0           0           0           0
 machines, and
 typewriters
Other information                      \a              11            27.3               0                0.0           0         0.0           0           0           0           0
 processing equipment
====================================================================================================================================================================================
Other household equipment           1.107           9,383            26.8             784                8.4         443        56.5         341           5         317          19
 and furnishings
Floor and window                    0.176           1,533            25.2             101                6.6          56        55.5          45           1          43           1
 coverings, and outdoor,
 infants', laundry, and
 cleaning equipment
Floor coverings                        \a             544            21.7              38                7.0          22        57.9          16           1        14\g           1
Window coverings                       \a             519            29.7              13                2.5          10        76.9           3           0         3\g           0
Infants' equipment                     \a             111            23.4              10                9.0           7        70.0           3           0         3\g           0
Laundry and cleaning                   \a             271            18.8              29               10.7          14        48.3          15           0        15\g           0
 equipment
Outdoor equipment                      \a              88            42.1              11               12.5           3        27.3           8           0         8\g           0
Clocks, lamps, and                  0.218           1,570            26.1             181               11.5          97        53.6          84           0          80           4
 decorator items
Clocks                                 \a              80            21.3               9               11.3           4        44.4           5           0         5\g           0
Lamps and lighting                     \a             353            29.8              37               10.5          23        62.2          14           0        13\g           1
 fixtures
Household decorative                   \a           1,137            25.2             135               11.9          70        51.9          65           0        62\g           3
 items
Tableware, serving                  0.198           1,790            24.3             162                9.1          74        45.7          88           1          83           4
 pieces, and nonelectric
 kitchenware
Plastic dinnerware                     \a              30            10.0               1                3.3           1       100.0           0           0         0\g           0
China and other                        \a             389            31.6              42               10.8          19        45.2          23           1        20\g           2
 dinnerware
Flatware                               \a             223            29.6              18                8.1          10        55.6           8           0         8\g           0
Glassware                              \a             228            24.1              17                7.5           7        41.2          10           0         9\g           1
Silver serving pieces                  \a               0             0.0               0                0.0           0         0.0           0           0           0           0
Serving pieces other than              \a              57            33.3               4                7.0           1        25.0           3           0         3\g           0
 silver or glass
Nonelectric cookware                   \a             236            24.6              19                8.1           9        47.4          10           0        10\g           0
Tableware and nonelectric              \a             627            17.7              61                9.7          27        44.3          34           0        33\g           1
 kitchenware
Lawn and garden                     0.173           1,665            24.1             109                6.6          88        80.7          21           1          20           0
 equipment, tools, and
 hardware
Lawn and garden equipment              \a             819            33.7              80                9.8          69        86.3          11           1        10\g           0
Power tools                            \a             339            16.2              11                3.2           7        63.6           4           0         4\g           0
Other hardware                         \a             263            11.8               9                3.4           7        77.8           2           0         2\g           0
Nonpowered hand tools                  \a             244            16.0               9                3.7           5        55.6           4           0         4\g           0
Small kitchen appliances,           0.113           1,615            32.8             144                8.9          93        64.6          51           1          45           5
 sewing machines, and
 portable heating and
 cooling equipment
Floor cleaning equipment               \a             597            38.5              63               10.6          48        76.2          15           1        14\g           0
 and sewing machines
Portable heating and                   \a           1,018            29.5              81                8.0          45        55.6          36           0        31\g           5
 cooling equipment, and
 small electric kitchen
 appliances
Indoor plants and fresh             0.150           1,210            29.1              87                7.2          35        40.2          52           1        46\g           5
 cut flowers
Unpriced household                  0.079               0             0.0               0                0.0           0         0.0           0           0           0           0
 equipment parts and
 small furnishings\b
====================================================================================================================================================================================
Housekeeping supplies               1.090          10,013            23.8             446                4.5         206        46.2         240          19           0         221
Laundry and cleaning                0.380           4,170            25.1             144                3.5          58        40.3          86           5           0          81
 products
Soaps and detergents                   \a           2,512            26.6              97                3.9          35        36.1          62           3           0          59
Other laundry and                      \a           1,658            23.0              47                2.8          23        48.9          24           2           0          22
 cleaning products
Household paper products,           0.364           2,974            22.1             195                6.6          95        48.7         100          12           0          88
 including stationery
Cleansing and toilet                   \a           1,531            27.3              53                3.5          16        30.2          37           7           0          30
 tissue, paper towels,
 and napkins
Stationery, stationery                 \a           1,443            16.5             142                9.8          79        55.6          63           5           0          58
 supplies, and gift wrap
Other household products,           0.346           2,869            23.6             107                3.7          53        49.5          54           2           0          52
 lawn, and garden
 supplies
Miscellaneous household                \a           2,111            23.4              61                2.9          31        50.8          30           2           0          28
 products
Lawn and garden supplies               \a             758            24.0              46                6.1          22        47.8          24           0           0          24
====================================================================================================================================================================================
Housekeeping services               1.492           7,311             6.6              48                0.7          20        41.7          28           9           0          19
Postage                             0.254           1,578             0.0               0                0.0           0         0.0           0           0           0           0
Appliance and furniture             0.182             817             8.6               3                0.4           1        33.3           2           1           0           1
 repair
Repair of television,                  \a             415             4.6               0                0.0           0         0.0           0           0           0           0
 radio, and sound
 equipment
Repair of household                    \a             229            18.8               1                0.4           0         0.0           1           1           0           0
 appliances
Reupholstery of furniture              \a             173             4.6               2                1.2           1        50.0           1           0           0           1
Gardening and other                 0.378           3,765             8.6              32                0.9          17        53.1          15           3           0          12
 household services
Gardening and lawn care                \a           1,535            11.1              20                1.3          10        50.0          10           2           0           8
 services
Water softening service                \a              84             8.3               1                1.2           0         0.0           1           1           0           0
Moving, storage, and                   \a           1,033            10.1              10                1.0           7        70.0           3           0           0           3
 freight expense
Household laundry and                  \a             946             4.0               1                0.1           0         0.0           1           0           0           1
 drycleaning, excluding
 coin operated
Coin-operated household                \a             167             1.2               0                0.0           0         0.0           0           0           0           0
 laundry and drycleaning
Babysitting                         0.271              \h              \h              \h                 \h          \h          \h          \h          \h          \h          \h
Domestic services                   0.237           1,010             7.5              10                1.0           1        10.0           9           4           0           5
Care of invalids,                   0.054             141             7.8               3                2.1           1        33.3           2           1           0           1
 elderly, and
 convalescents in the
 home
Unpriced rent and/or                0.116               0             0.0               0                0.0           0         0.0           0           0           0           0
 repair of household
 equipment, and sound
 equipment\b
====================================================================================================================================================================================
Apparel and upkeep                  5.291          76,736             n/a           9,797               12.8       6,598        67.3       3,199       1,223       1,796         180
 component total
====================================================================================================================================================================================
Men's apparel                       1.061          17,016            40.6           1,391                8.2         989        71.1         402         192         183          27
Men's suits, coats,                 0.312           6,221            40.6             466                7.5         285        61.2         181         101          69          11
 sportcoats, and jackets
Men's suits                            \a           3,841            38.4             181                4.7          93        51.4          88          71        14\g           3
Men's sportcoats and                   \a             619            47.0              40                6.5          24        60.0          16           7         7\g           2
 tailored jackets
Men's coats and jackets                \a           1,761            43.3             245               13.9         168        68.6          77          23        48\g           6
Men's furnishings                   0.257           3,335            37.2             336               10.1         254        75.6          82          18          57           7
Men's underwear and                    \a             968            29.1              43                4.4          39        90.7           4           0         3\g           1
 hosiery
Men's nightwear                        \a             287            24.0              17                5.9          16        94.1           1           0         1\g           0
Men's accessories                      \a           1,067            29.6              88                8.3          71        80.7          17           0        15\g           2
Men's sweaters                         \a             474            58.0              93               19.6          51        54.8          42          18        22\g           2
Men's active sportswear                \a             539            55.5              95               17.6          77        81.1          18           0        16\g           2
Men's shirts                        0.266           4,100            45.1             407                9.9         306        75.2         101          53        42\g           6
Men's pants and shorts              0.212           3,360            38.5             182                5.4         144        79.1          38          20        15\g           3
Unpriced men's uniforms             0.014               0             0.0               0                0.0           0         0.0           0           0           0           0
 and other clothing\b
====================================================================================================================================================================================
Boys' apparel                       0.230           3,457            43.5             436               12.6         305        70.0         131           1         125           5
Boys' coats and jackets                \a             264            66.3              58               22.0          44        75.9          14           0        14\g           0
Boys' sweaters                         \a              96            62.5              22               22.9          11        50.0          11           0         9\g           2
Boys' shirts                           \a           1,008            48.5             159               15.8         104        65.4          55           0        53\g           2
Boys' underwear,                       \a             332            25.0              16                4.8          12        75.0           4           1         3\g           0
 nightwear, and hosiery
Boys' accessories                      \a             336            30.7              27                8.0          21        77.8           6           0         6\g           0
Boys' suits, sportcoats,               \a           1,200            37.7              89                7.4          67        75.3          22           0        21\g           1
 and pants
Boys' active sportswear                \a             221            64.7              65               29.4          46        70.8          19           0        19\g           0
Unpriced boys' uniforms                \a               0             0.0               0                0.0           0         0.0           0           0           0           0
 and other clothing\b
====================================================================================================================================================================================
Women's apparel                     1.765          23,328            52.1           5,040               21.6       3,199        63.5       1,841         844         897         100
Women's coats and jackets           0.181           1,960            55.6             503               25.7         275        54.7         228          95       117\g          16
Women's dresses                     0.246           2,299            65.5             853               37.1         486        57.0         367         176       175\g          16
Women's separates and               0.821          12,452            55.3           2,653               21.3       1,689        63.7         964         484         425          55
 sportswear
Women's tops                           \a           5,472            61.7           1,514               27.7         922        60.9         592         299       256\g          37
Women's skirts                         \a             704            64.6             208               29.6         103        49.5         105          62        39\g           4
Women's pants and shorts               \a           5,272            46.6             653               12.4         442        67.7         211         120        79\g          12
Women's active sportswear              \a           1,004            60.2             278               27.7         222        79.9          56           3        51\g           2
Women's underwear,                  0.323           5,449            35.2             653               12.0         535        81.9         118           2         110           6
 nightwear, and
 accessories
Women's nightwear                      \a           1,155            56.0             288               24.9         237        82.3          51           0        50\g           1
Women's underwear                      \a           1,510            33.3              91                6.0          71        78.0          20           1        19\g           0
Women's hosiery                        \a           1,631            19.2              79                4.8          63        79.8          16           1        13\g           2
Women's accessories                    \a           1,153            39.4             195               16.9         164        84.1          31           0        28\g           3
Women's suits                       0.168           1,168            64.6             378               32.4         214        56.6         164          87        70\g           7
Unpriced women's uniforms           0.026               0             0.0               0                0.0           0         0.0           0           0           0           0
 and other clothing\b
====================================================================================================================================================================================
Girls' apparel                      0.307           4,414            49.2             918               20.8         584        63.6         334         129         183          22
Girls' coats and jackets               \a             160            60.6              55               34.4          34        61.8          21           4        15\g           2
Girls' dresses and suits               \a             547            75.1             238               43.5         130        54.6         108          51        55\g           2
Girls' tops                            \a             804            60.5             212               26.4         148        69.8          64          30        32\g           2
Girls' skirts and pants                \a           1,334            45.4             185               13.9         116        62.7          69          19        44\g           6
Girls' active sportswear               \a             283            63.3              97               34.3          58        59.8          39          22        11\g           6
Girls' underwear and                   \a             599            33.4              55                9.2          40        72.7          15           1        13\g           1
 nightwear
Girls' hosiery and                     \a             687            27.8              76               11.1          58        76.3          18           2        13\g           3
 accessories
Unpriced girls' uniforms               \a               0             0.0               0                0.0           0         0.0           0           0           0           0
 and other clothing\b
====================================================================================================================================================================================
Infants' and toddlers'              0.182           1,819            28.5             167                9.2         126        75.5          41           0          39           2
 apparel
Infants' and toddlers'                 \a              37            59.5              11               29.7           8        72.7           3           0         3\g           0
 outerwear
Infants' and toddlers'                 \a             358            50.0              82               22.9          65        79.3          17           0        17\g           0
 play and dresswear
Infants' and toddlers'                 \a           1,268            20.4              52                4.1          35        67.3          17           0        15\g           2
 underwear
Infants' and toddlers'                 \a             156            38.5              22               14.1          18        81.8           4           0         4\g           0
 sleepwear
Unpriced infants'                      \a               0             0.0               0                0.0           0         0.0           0           0           0           0
 accessories and other
 clothing\b
====================================================================================================================================================================================
Sewing materials,                   0.083           1,464            32.3             121                8.3          79        65.3          42           0          41           1
 notions, and luggage
Fabric for making clothes              \a             656            26.1              43                6.6          31        72.1          12           0        11\g           1
Sewing notions and                     \a             185            22.7              10                5.4           7        70.0           3           0         3\g           0
 patterns
Luggage                                \a             623            41.7              68               10.9          41        60.3          27           0        27\g           0
====================================================================================================================================================================================
Jewelry                             0.401           6,428            35.9             495                7.7         338        68.3         157          12         139           6
Watches                             0.078           1,223            30.3             106                8.7          82        77.4          24           0        23\g           1
Jewelry                             0.323           5,205            37.2             389                7.5         256        65.8         133          12       116\g           5
====================================================================================================================================================================================
Footwear                            0.719          11,159            39.7           1,193               10.7         962        80.6         231          30         189          12
Men's footwear                      0.224           2,972            35.6             235                7.9         193        82.1          42          10        27\g           5
Boys' and girls' footwear           0.154           1,537            38.2             203               13.2         169        83.3          34           0          31           3
Boy's footwear                         \a             760            44.1             115               15.1          96        83.5          19           0        17\g           2
Girl's footwear                        \a             777            32.4              88               11.3          73        83.0          15           0        14\g           1
Women's footwear                    0.341           6,650            41.8             755               11.4         600        79.5         155          20       131\g           4
====================================================================================================================================================================================
Apparel services                    0.543           7,651             7.1              36                0.5          16        44.4          20          15           0           5
Apparel laundry and dry             0.288           4,012             9.6              11                0.3           3        27.3           8           7           0           1
 cleaning, excluding coin
 operated
Other apparel services              0.255           3,639             4.4              25                0.7          13        52.0          12           8           0           4
Shoe repair and other                  \a             379             7.7               3                0.8           1        33.3           2           2           0           0
 shoe services
Coin-operated apparel                  \a           2,467             2.3              12                0.5          10        83.3           2           2           0           0
 laundry and dry cleaning
Alterations and repairs                \a             444             4.7               1                0.2           0         0.0           1           1           0           0
Clothing rental                        \a             224            19.2               9                4.0           2        22.2           7           3           0           4
Watch and jewelry repair               \a             125             8.0               0                0.0           0         0.0           0           0           0           0
====================================================================================================================================================================================
Transportation component           16.620          94,366             n/a           5,660                6.2       2,699        47.6       2,961       1,837         823         301
 total
====================================================================================================================================================================================
New vehicles                        4.829          15,176            53.9           2,420               16.0         805        33.3       1,615         838         670         107
New cars                            3.842          11,617            56.4           1,917               16.5         584        30.5       1,333         667       580\g          86
New trucks                          0.894           2,192            52.2             341               15.6          80        23.5         261         171        75\g          15
New motorcycles                     0.093           1,367            34.6             162               11.9         141        87.0          21           0        15\g           6
====================================================================================================================================================================================
Used vehicles                       1.195           5,067           100.0           1,097               21.7           0         0.0       1,097         990           0         107
Unpriced used cars\b                   \a               0             0.0               0                0.0           0         0.0           0           0           0           0
====================================================================================================================================================================================
Motor fuel                          2.925          29,524            75.0             172                0.6         159        92.4          13           2           0          11
Regular unleaded gasoline              \a           9,509            79.3              50                0.5          47          94           3           0           0           3
Mid-grade unleaded                     \a           8,370            77.5              47                0.6          45        95.7           2           0           0           2
 gasoline
Premium unleaded gasoline              \a           9,186            76.3              64                0.7          59        92.2           5           2           0           3
Diesel                                 \a           2,285            45.4               9                0.4           7        77.8           2           0           0           2
Other motor fuel                       \a             174            63.8               2                1.2           1        50.0           1           0           0           1
====================================================================================================================================================================================
Automobile maintenance              1.546           8,401            25.6           1,303               15.5       1,169        89.7         134           3         117          14
 and repair
Automotive body work                0.167           2,073            28.1             334               16.1         287        85.9          47           1        38\g           8
Automobile drive train              0.453           1,990            27.5             306               15.4         273        89.2          33           1          31           1
 and front end repair
Automotive drive-train                 \a             803            28.1             129               16.1         114        88.4          15           0        15\g           0
 repair
Automotive brake work                  \a             545            25.3              81               14.9          73        90.1           8           1         6\g           1
Repair to steering, front              \a             642            28.7              96               15.0          86        89.6          10           0        10\g           0
 end, cooling system, and
 air conditioning
Automotive maintenance              0.490           2,376            17.6             352               14.8         339        96.3          13           1        10\g           2
 and servicing
Power plant repair                  0.412           1,962            30.5             311               15.9         270        86.8          41           0        38\g           3
Unpriced automotive                 0.023               0             0.0               0                0.0           0         0.0           0           0           0           0
 repair service policy\b
====================================================================================================================================================================================
Motor oil, coolant, and             0.058           3,745            11.7              36                1.0          27        75.0           9           0           0           9
 other fluids
Motor oil                              \a           2,676            11.2              18                0.7          12        66.7           6           0           0           6
Coolant brake fluid,                   \a           1,069            12.9              18                1.7          15        83.3           3           0           0           3
 transmission fluid, and
 additives
====================================================================================================================================================================================
Automobile parts and                0.516           7,964            20.9             255                3.2         211        82.8          44           2          36           6
 equipment
Tires                               0.256           4,037            26.8             102                2.5          85        83.3          17           0        17\g           0
Vehicle parts and                   0.260           3,927            14.9             153                3.9         126        82.4          27           2        19\g           6
 equipment other than
 tires
====================================================================================================================================================================================
Automobile insurance                2.647           5,111            25.0             147                2.9         133        90.5          14           0           0          14
====================================================================================================================================================================================
Vehicle finance charges             0.571           2,373            36.1              68                2.9          66        97.1           2           0           0           2
Unpriced other vehicle                 \a               0             0.0               0                0.0           0         0.0           0           0           0           0
 finance charges\b
====================================================================================================================================================================================
Vehicle rental,                     0.768           5,757            25.1              89                1.6          77        86.5          12           2           0          10
 registration, and
 inspection
State and local                     0.368           1,874             4.5              15                0.8          12        80.0           3           2           0           1
 automobile registration,
 license, and inspection
State automobile                       \a           1,354             5.5              10                0.7           8        80.0           2           1           0           1
 registration
Local automobile                       \a             249             2.4               5                2.0           4        80.0           1           1           0           0
 registration
Driver's license                       \a             227             1.8               0                0.0           0         0.0           0           0           0           0
Vehicle inspection                     \a              44             0.0               0                0.0           0         0.0           0           0           0           0
Other automobile-related            0.372           3,883            35.1              74                1.9          65        87.8           9           0           0           9
 fees
Automobile rental                      \a           2,141            54.3              62                2.9          57        91.9           5           0           0           5
Truck rental                           \a             521            29.8               6                1.2           4        66.7           2           0           0           2
Parking fees                           \a             697             4.5               4                0.6           2        50.0           2           0           0           2
Vehicle tolls                          \a             464             2.6               1                0.2           1       100.0           0           0           0           0
Automobile towing charges              \a              60             0.0               1                1.7           1       100.0           0           0           0           0
Other vehicle rental                   \a               0             0.0               0                0.0           0         0.0           0           0           0           0
Unpriced docking and                0.027               0             0.0               0                0.0           0         0.0           0           0           0           0
 landing fees\b
====================================================================================================================================================================================
Public transportation               1.566          11,248            45.5              73                0.7          52        71.2          21           0           0          21
Airline fare                        1.037           6,366            69.5               0                0.0           0         0.0           0           0           0           0
Other intercity                     0.139           2,318            26.2              59                2.6          40        67.8          19           0           0          19
 transportation
Intercity bus fare                     \a             407            35.9               4                1.0           2        50.0           2           0           0           2
Intercity train fare                   \a           1,154            12.0               1                0.1           0         0.0           1           0           0           1
Ship fares                             \a             757            42.7              54                7.1          38        70.4          16           0           0          16
Intracity transportation            0.378           2,564             3.6              14                0.6          12        85.7           2           0           0           2
Intracity mass transit                 \a           2,040             2.2              13                0.6          12        92.3           1           0           0           1
Taxi fare                              \a             524             9.2               1                0.2           0           0           1           0           0           1
Car and van pools                      \a               0             0.0               0                0.0           0         0.0           0           0           0           0
Unpriced school bus\b               0.012               0             0.0               0                0.0           0         0.0           0           0           0           0
====================================================================================================================================================================================
Medical care component              7.426          50,237             n/a           1,116                2.2         355        31.8         761         336           3         422
 total
====================================================================================================================================================================================
Prescription drugs and              0.897           4,471            26.0             108                2.4          49        45.4          59          11           0          48
 medical supplies
====================================================================================================================================================================================
Nonprescription drugs and           0.383           4,488            19.3             173                3.9          41        23.7         132          43           0          89
 medical
Internal and respiratory            0.245           2,392            19.9              73                3.1          18        24.7          55          25           0          30
 over-the-counter drugs
Nonprescription medical             0.138           2,096            18.5             100                4.8          23        23.0          77          18           0          59
 equipment and supplies
Topicals and dressings                 \a           1,282            21.8              62                4.8          19        30.7          43          13           0          30
Medical equipment for                  \a             181            17.7              16                8.8           0         0.0          16           1           0          15
 general use
Supportive and                         \a             195            14.9              11                5.6           4        36.4           7           0           0           7
 convalescent medical
 equipment
Hearing aids                           \a             438            10.7              11                2.5           0         0.0          11           4           0           7
Unpriced drugs\b                       \a               0             0.0               0                0.0           0         0.0           0           0           0           0
====================================================================================================================================================================================
Professional services               3.518          18,430             9.4             210                1.1          68        32.4         142          49           0          93
Physicians services                 1.904           8,664             8.7             106                1.2          22        20.8          84          28           0          56
Dental services                     1.107           5,973             9.6              18                0.3           7        38.9          11           7           0           4
Eyeglasses and eye care             0.335           2,081            13.7              60                2.9          30        50.0          30           8           0          22
Services by other medical           0.172           1,712             7.2              26                1.5           9        34.6          17           6           0          11
 professionals
====================================================================================================================================================================================
Hospital and related                2.310          22,848            18.9             625                2.7         197        31.5         428         233           3         192
 services
Hospital services                   2.159          19,913            19.5             581                2.9         175        30.1         406         218           3         185
Nursing home services               0.145           2,935            14.9              44                1.5          22        50.0          22          15           0           7
Unpriced items\b                    0.006               0             0.0               0                0.0           0         0.0           0           0           0           0
====================================================================================================================================================================================
Health insurance\i                  0.318               0             n/a             n/a                n/a         n/a         n/a         n/a         n/a         n/a         n/a
====================================================================================================================================================================================
Entertainment component             4.339          32,985             n/a           1,327                4.0         763        57.5         564          95         321         148
 total
====================================================================================================================================================================================
Reading materials                   0.730           8,891             5.6             195                2.2         126        64.6          69           9          55           5
Newspapers                          0.376           4,964             1.8              11                0.2           6        54.6           5           2         2\g           1
Magazines, periodicals,             0.354           3,927            10.4             184                4.7         120        65.2          64           7          53           4
 and books
Magazines                              \a           1,840             7.6              23                1.3          13        56.5          10           3         7\g           0
Books purchased through                \a             362            14.4              34                9.4          17        50.0          17           0        16\g           1
 book clubs
Books not purchased                    \a           1,725            12.5             127                7.4          90        70.9          37           4        30\g           3
 through book clubs
Unpriced newsletters\b              0.000               0             0.0               0                0.0           0         0.0           0           0           0           0
====================================================================================================================================================================================
Sporting goods and                  0.391           4,303            26.3             384                8.9         247        64.3         137           7         114          16
 equipment
Sport vehicles, including           0.181           1,658            30.6             170               10.3         146        85.9          24           3          15           6
 bicycles
Outboard motors and                    \a           1,173            31.6             123               10.5         108        87.8          15           3         8\g           4
 powered sports vehicles
Unpowered boats and                    \a             136            19.1              11                8.1          10        90.9           1           0         1\g           0
 trailers
Bicycles                               \a             349            31.8              36               10.3          28        77.8           8           0         6\g           2
Sports equipment                    0.210           2,645            23.6             214                8.1         101        47.2         113           4          99          10
Indoor, warm weather, and              \a           2,057            23.0             170                8.3          77        45.3          93           3        83\g           7
 winter sports equipment
Hunting, fishing, and                  \a             588            25.5              44                7.5          24        54.6          20           1        16\g           3
 camping equipment
====================================================================================================================================================================================
Toys, hobbies, and other            0.803           6,793            22.1             436                6.4         265        60.8         171           8         152          11
 entertainment
 commodities
Toys, hobbies, and music            0.360           2,997            21.6             269                9.0         160        59.5         109           4          99           6
 equipment
Toys, games, and hobbies               \a           2,477            20.4             239                9.7         145        60.7          94           3        86\g           5
Playground equipment                   \a               8            75.0               3               37.5           0         0.0           3           0         3\g           0
Music instruments and                  \a             512            26.6              27                5.3          15        55.6          12           1        10\g           1
 accessories
Photographic supplies and           0.111           1,162            20.0              62                5.3          46        74.2          16           1          13           2
 equipment
Film                                   \a             636            20.4              27                4.3          20        74.1           7           1         5\g           1
Photographic and darkroom              \a              24             8.3               0                0.0           0         0.0           0           0         0\g           0
 supplies
Photographic equipment                 \a             502            19.9              35                7.0          26        74.3           9           0         8\g           1
Pets and pet products               0.323           2,634            23.7             105                4.0          59        56.2          46           3          40           3
Pet food                               \a           1,726            26.7              56                3.2          29        51.8          27           2        22\g           3
Purchase of pets, pet                  \a             908            18.1              49                5.4          30        61.2          19           1        18\g           0
 supplies, and
 accessories
Unpriced souvenirs,                 0.010               0             0.0               0                0.0           0         0.0           0           0           0           0
 fireworks, and optic
 goods\b
====================================================================================================================================================================================
Entertainment services              2.415          12,998            14.4             312                2.4         125        40.1         187          71           0         116
Club memberships dues and           0.346           2,363            12.1              35                1.5          10        28.6          25           6           0          19
 fees
Fees for participant                0.400           2,211            17.2              38                1.7          19        50.0          19           4           0          15
 sports
Admissions                          0.726           4,014            17.9             149                3.7          79        53.0          70          28           0          42
Admission to movies,                   \a           3,323            14.1              98                3.0          68        69.4          30           6           0          24
 theaters, and concerts
Admission to sporting                  \a             691            36.5              51                7.4          11        21.6          40          22           0          18
 events
Fees for lessons or                 0.256             888            11.5              38                4.3           3         7.9          35          22           0          13
 instructions
Photographers, film                 0.665           3,522            10.9              52                1.5          14        26.9          38          11           0          27
 procession, and pet
 services
Photographer fees                      \a             197            18.8               8                4.1           2        25.0           6           0           0           6
Film processing                        \a             710            10.6               9                1.3           1        11.1           8           2           0           6
Pet services                           \a             291             8.6               0                0.0           0         0.0           0           0           0           0
Veterinarian services                  \a           1,106            11.8              15                1.4           2        13.3          13           4           0           9
Other entertainment                    \a           1,218             9.7              20                1.6           9        45.0          11           5           0           6
 services
Unpriced rental of                  0.023               0             0.0               0                0.0           0         0.0           0           0           0           0
 recreational vehicles\b
====================================================================================================================================================================================
Other goods and services            7.390          22,440             n/a             422                1.9         215        50.9         207          40          52         115
 component total
====================================================================================================================================================================================
Tobacco products                    1.688           4,120            27.3              15                0.4           5        33.3          10           2           0           8
Cigarettes                             \a           3,710            28.8              10                0.3           2        20.0           8           2           0           6
Tobacco products other                 \a             356            14.3               3                0.8           1        33.3           2           0           0           2
 than cigarettes
Smoking accessories                    \a              54            11.1               2                3.7           2       100.0           0           0           0           0
Unpriced smoking                       \a               0             0.0               0                0.0           0         0.0           0           0           0           0
 products\b
====================================================================================================================================================================================
Toilet goods and personal           0.589           3,653            19.9             129                3.5          63        48.8          66           5           0          61
 care appliances
Cosmetics, bath, and nail           0.263           1,214            16.8              50                4.1          26        52.0          24           2           0          22
 preparations and
 implements
Hair, dental, shaving,              0.325           2,439            21.4              79                3.2          37        46.8          42           3           0          39
 and miscellaneous
 personal care products
Products for the hair                  \a           1,022            21.8              29                2.8          11        37.9          18           0           0          18
Nonelectric articles for               \a              72            12.5               5                6.9           2        40.0           3           0           0           3
 the hair
Women's hair pieces and                \a               0             0.0               0                0.0           0         0.0           0           0           0           0
 wigs
Dental products and                    \a             528            23.5              17                3.2           9        52.9           8           2           0           6
 nonelectric dental
 articles
Shaving products and                   \a             158            24.1               6                3.8           1        16.7           5           1           0           4
 nonelectric shaving
 articles
Deodorant and suntan                   \a             580            19.1              15                2.6           9        60.0           6           0           0           6
 preparations and
 sanitary and footcare
 products
Electric personal care                 \a              79            22.8               7                8.9           5        71.4           2           0           0           2
 appliances
====================================================================================================================================================================================
Personal care services              0.564           5,576             4.1              16                0.3           8        50.0           8           0           0           8
Beauty parlor services              0.447           3,679             4.2              15                0.4           7        46.7           8           0           0           8
 for females
Haircuts and other barber           0.116           1,897             4.0               1                0.1           1         100           0           0           0           0
 shop services for men
Unpriced repair of                  0.000               0             0.0               0                0.0           0         0.0           0           0           0           0
 personal care
 appliances\b
====================================================================================================================================================================================
School books and supplies           0.273           1,176            25.7              93                7.9          35        37.6          58           2          52           4
School books and supplies           0.194             637            37.1              69               10.8          19        27.5          50           2        45\g           3
 for college
Reference books and                 0.064             539            12.2              24                4.5          16        66.7           8           0           7           1
 elementary and high
 school books
Elementary and high                    \a             124            32.3              10                8.1           2        20.0           8           0         7\g           1
 school books and
 supplies
Encyclopedias and other                \a             415             6.3              14                3.4          14       100.0           0           0         0\g           0
 sets of reference books
Unpriced miscellaneous              0.014               0             0.0               0                0.0           0         0.0           0           0           0           0
 school purchases\b
====================================================================================================================================================================================
Daycare, tuition, and               2.863           4,685            23.7             104                2.2          70        67.3          34          16           0          18
 other school fees
College tuition and fees            1.685           1,334            48.8              69                5.2          57        82.6          12           6           0           6
Elementary and high                 0.519             356            48.0               7                2.0           5        71.4           2           1           0           1
 school tuition and fees
Child daycare and nursery           0.388           1,665             9.1              10                0.6           6        60.0           4           2           0           2
 school
Other tuition and fees              0.155           1,330            10.3              18                1.4           2        11.1          16           7           0           9
Unpriced miscellaneous              0.116               0             0.0               0                0.0           0         0.0           0           0           0           0
 school items, rentals,
 and other services\b
====================================================================================================================================================================================
Legal, financial, and               1.415           3,230            12.4              65                  2          34        52.3          31          15           0          16
 funeral services
Legal fees                          0.496             943             4.9               7                0.7           2        28.6           5           1           0           4
Personal financial                  0.407           1,047            12.2              23                2.2          11        47.8          12           6           0           6
 services
Safe deposit box rental                \a             113             6.2               0                  0           0         0.0           0           0           0           0
Checking accounts and                  \a             390            12.3              15                3.9           9        60.0           6           2           0           4
 special check services
Tax return preparation                 \a             544            13.4               8                1.5           2        25.0           6           4           0           2
 and other accounting
 fees
Cemetery lots and funeral           0.403           1,240            18.3              35                2.8          21        60.0          14           8           0           6
 expenses
Funeral expenses                       \a             905            20.8              29                3.2          18        62.1          11           5           0           6
Cemetery lots and crypts               \a             335            11.6               6                1.8           3        50.0           3           3           0           0
Unpriced miscellaneous              0.109               0             0.0               0                0.0           0         0.0           0           0           0           0
 personal services\b
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
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 owner's 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
entry-level item. 

\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. 

(See figure in printed edition.)Appendix IX

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

COMMENTS FROM THE BUREAU OF LABOR
STATISTICS
======================================================== Appendix VIII

(See figure in printed edition.)

(See figure in printed edition.)

(See figure in printed edition.)

The following are GAO's comments on the Bureau of Labor Statistics'
letter dated April 7, 1999. 

GAO COMMENTS

1.  The Commissioner listed several measures BLS had taken to improve
the 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. 

MAJOR CONTRIBUTORS TO THIS REPORT
=========================================================== Appendix X

   GENERAL GOVERNMENT DIVISION,
   WASHINGTON, D.C. 
--------------------------------------------------------- Appendix X:1

Martin de Alteriis, Evaluator-in-Charge
Kathleen Scholl, Supervisory Economist
Tony Assia, Senior Evaluator
Amber Roos, Student Intern

   OFFICE OF THE CHIEF ECONOMIST
--------------------------------------------------------- Appendix X:2

Loren Yager, Acting Chief Economist
Richard Krashevski, Supervisory Economist

   DALLAS OFFICE
--------------------------------------------------------- Appendix X:3

James Turkett, Senior Evaluator

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). 

*** End of document. ***