Testimony of Katharine G. Abraham Commissioner of Labor
Statistics Before the Subcommittee on Human Resources House
Committee on Government Reform and Oversight April 29, 1998
I very much appreciate the opportunity to appear once again
before the Subcommittee to discuss the Consumer Price Index
(CPI). As I indicated in my testimony a year ago, the Bureau of
Labor Statistics (BLS) has a long tradition of being in the
forefront of price measurement research and operational
innovation. In that testimony, I chronicled a lengthy list of CPI
improvements introduced by the BLS over the years, emphasizing
especially important improvements introduced in 1995, 1996 and
the first part of 1997. My understanding is that the Subcommittee
wishes to review both the further changes we have made since last
year and our plans for future improvements. I am proud of the
progress we have made, and welcome this opportunity to review our
accomplishments and discuss our future plans.
As the Subcommittees invitation of April 17 notes, the
BLS introduced a number of very important changes in the CPI
effective with the data for January 1998, released two months ago
on February 24. To be more precise, the CPI now is based on
consumer spending patterns for the 1993 through 1995 time period,
a new sample of geographic areas that better represents the
current distribution of the U.S. population, and a new
classification structure that better reflects the categories of
goods and services consumers buy. These important changes
represent the attainment of some of the more critical objectives
of the large-scale effort to revise the CPI for which funds were
first received in Fiscal Year 1995. Although major components of
the CPI Revision project continue through Fiscal Year 2000, the
changes introduced in February complete the expenditure weight
and geographic area sample updating that also have been core
components of every prior CPI Revision.
Although the updated market basket introduced in January is
considerably more current than that it replaces, it still was 3½
years old as of its introduction date. Our proposed CPI
improvement initiative, which I will discuss in a few moments,
will allow us to introduce future expenditure weight updates in a
more timely fashion. Further, it has become apparent that
shortening the roughly 10-year interval between comprehensive
updates to the CPI would be desirable, and we have work underway
to determine a more appropriate time schedule for such updates. I
should note, however, that the availability of new Census of
Population information likely always will necessitate some form
of periodic CPI updating to account for shifts in the
distribution of the population. Thus, I suspect that updating of
the CPI geographic sample may well, of necessity, remain a
decennial statistical activity.
Finally, both for the sake of completeness and because it is
important in its own right, I should call the Subcommittees
attention to an improvement in the treatment of quality changes
in personal computers also introduced with data for January 1998.
The revised item classification structure for the CPI that I
already have mentioned includes a new stratum called Personal
Computers and Peripheral Equipment. Analysts in the BLS Producer
Price Index (PPI) program earlier developed and implemented a
regression procedure, called a hedonic model, that attaches an
implicit price to each important feature and component of the
computer. Starting with the CPI for January 1998, when a personal
computer or selected item of peripheral equipment, such as a
modem, in the CPI sample improves in some way, a regression-based
quality adjustment to its price will be made. The value of the
improvement, as derived from the regression estimates, will be
deducted from the observed price for the product. Conversely, if
a model deteriorates, the value of the difference will be added
to the price. Application of this method will improve our
measurement of price change for personal computers and peripheral
equipment.
Up to this point, I have focused principally on the changes to
the CPI introduced with the data for this past January. Now, as
your invitation to testify requests, I would like to turn to the
announcement we made earlier this month concerning the use of the
geometric mean formula in the monthly CPI. We expect to
incorporate this change into the CPI for the month of January
1999, to be published next February.
When I testified before the Subcommittee last year, I spent a
considerable amount of time discussing the issue of substitution
bias. At that time, I noted that BLS had begun publication of an
experimental index utilizing a formula which allowed for some
degree of consumer reaction to change in the relative prices of
the goods and services being purchased. Prior to my arrival at
the BLS in 1993, CPI staff had raised the possibility of adopting
a geometric mean formula in some or all index components,
possibly in conjunction with the CPI Revision then being
launched. By 1996, the issue had become a key element of the
report of the Senate Finance Committees Advisory Commission
to Study the CPI. In April of last year, we reported that we
would conclude our review of the issue and announce our decisions
by the end of 1997, and that we would incorporate whatever
changes we believed prudent in the CPI to be published for the
month of January 1999.
As I indicated in my March 31 letter to you, and as we also
indicated in making our announcement the week before last, we
found the evidence relevant to assessing the use of the geometric
mean to be much sparser than we had expected. Taken in its
entirety, however, the evidence unambiguously supports the
proposition that consumers can, and do, alter their purchasing
behavior in response to changes in the array of prices that they
confront in the market place. Because the geometric mean
estimator can better reflect the effects of such changes in
consumer spending than can the current CPI formula, we have
decided to implement the geometric mean estimator in most CPI
basic indexes. The changes we will be making were announced on
April 16, and, as originally planned, will become effective with
data for January 1999.
The geometric mean estimator will be used in index categories
that comprise approximately 61 percent of total consumer spending
represented by the CPI-U and 64 percent of that represented by
the CPI-W. The remaining index categories, which are shown on the
attached table, will continue to be calculated as they are
currently. Based upon BLS research, it is expected that adoption
of the new formula will reduce the annual rate of increase in the
CPI by approximately 0.2 percentage point per year.
Finally, Mr. Chairman, I would like to turn to the last
question posed in your letter, which I believe was aimed
principally at ascertaining our future plans and the resources
required to implement those plans. In this regard, I would like
to call your attention to the BLS Congressional Budget request
for Fiscal Year 1999, in which we are seeking to build upon an
initiative, started in 1998, to lay the groundwork for future
improvements in the timeliness and accuracy of the CPI. For
Fiscal Year 1999, our request seeks a total of $11.2 million and
47 FTE for the second phase of a three-year buildup to put this
improvement program in place. The proposal has four major pieces.
First, in order to improve the timeliness of future CPI
Revisions, we are proposing an expansion of approximately 50
percent in the size of the sample of households that supports the
Consumer Expenditure Survey. About four-fifths of the proposed
cost of the CPI Improvement proposal would be devoted to this
project. At the present time, three years of expenditure data are
required to construct a CPI market basket with an acceptable
degree of accuracy. The proposed expansion will reduce the number
of years of expenditure data required to two. Combined with a
separate project to modernize the CPI market basket update
system, this expansion will allow the BLS to reduce substantially
the lag between collection of expenditure information and the
introduction of future updated CPI market baskets. If funded,
this proposed change will greatly enhance the Bureaus
ability to adopt a policy of more frequent weight updates.
In addition to supporting more timely updates of CPI
expenditure weights, the expanded sample for the Consumer
Expenditure Survey also will support the publication of
production-quality superlative indexes which, as I mentioned in
last years testimony, will account for substitution across
CPI item categories in response to relative price changes. At
present, these measures are produced on an experimental basis,
and the CPI improvement proposal will provide us with the
additional resources needed to produce them on a regular
schedule, to a higher standard of precision and reliability.
Finally, we are requesting funds to support expanded pricing
of goods and services together with the collection of richer
information on those items characteristics, to support
additional regression-based quality adjustment applications.
While it is not possible to predict the outcome of this work in
advance, we do know that application of these methods has proven
to be of value in improving adjustments for changes in product
characteristics. In addition, for particular item groups that are
greatly affected by the introduction of new goods, BLS will be
able to replace or augment existing samples of priced items, the
objective being to bring new goods and services into the index on
a more timely basis.
In summary, Mr. Chairman, with the support of Congress and the
Administration, the BLS has been able to make substantial
progress in improving the accuracy and relevance of the Consumer
Price Index. Certainly, more can and will be accomplished. But,
just as certainly, further progress depends critically on the
continued support of those making decisions about the Federal
budget. Our success in improving the CPI also depends on the
publics continued belief that our decisions about the
construction of the index are arrived at independently and
impartially, using the most appropriate techniques available to
us. Preservation of that independence and impartiality should be
important to all who depend upon the CPI.
Attachment A: Components Retaining the Arithmetic Mean (Laspeyres) Formula