Content uploaded by Robert J. Gordon
Author content
All content in this area was uploaded by Robert J. Gordon on Jun 28, 2016
Content may be subject to copyright.
I
NTERNATIONAL
P
RODUCTIVITY
M
ONITOR
7
The Boskin Commission
Report: A Retrospective
One Decade Later
Robert J. Gordon1
Northwestern University and
National Bureau of Economic Research
M
ORE
THAN
TEN
YEARS
HAVE
now elapsed
since the formation of the Boskin Commission,
more formally the “Advisory Commission to
Study the Consumer Price Index,” which was
appointed by the Senate Finance Committee in
June, 1995. The Commission's final report
(Boskin et al., 1996) was issued on December 4,
1996 and concluded that the U. S. Consumer
Price Index (CPI) overstated inflation by 1.1
percentage points per year for the time period
of the Commission’s deliberations, 1995-96.
The report suggested that the CPI bias might
have been larger before 1995 and predicted that
it would be lower after 1997.
A sharp distinction must be drawn between
the technical and scientific issues which led the
Commission to its bias estimate of 1.1 percent-
age points, and the political and redistributional
implications of that conclusion. The Commis-
sion report contained not only the technical
background for its overall bias estimate but also
its estimates of the vast amount by which that
bias had increased the federal budget deficit
looking back into the past and would increase
the deficit looking forward into the future.
2
The
suggestion that the bias had caused excessive
growth in Social Security and other benefits
evoked a sharp and damning political reaction,
as the AARP (American Association of Retired
Persons) sent its lobbyists scurrying through the
corridors of Congress to throw cold water on
those senators and representatives who had ini-
tially been sympathetic to reducing the budget
deficit by adjusting the indexation formula by
some fraction of the Commission's bias esti-
mate, the so-called “CPI minus X” approach to
indexation. This paper ignores the implications
of the bias estimate for fiscal policy and instead
concentrates entirely on the issues of interest to
the worldwide measurement community.
The Boskin Commission's
Approach and Method
The Boskin Commission represented the first
extensive external evaluation of the nation’s
price statistics in more than three decades; the
previous report was the famous Stigler (1961)
Commission Report. There were several
1 The author served as one of the five members of the Boskin Commission. Portions of the summary of the
Boskin Commission findings and its reactions to its critics are adapted from Boskin et al. (1998) and Gordon
(2000). The evaluations and opinions contained herein are those of the author only and should not be taken
to represent the view of any other member of the Boskin Commission. I am grateful to Jerry Hausman for help-
ful discussions. Email: rjg@northwestern.edu
2 The Commission calculated that the 1.1 percentage points bias would contribute $1.07 trillion to the
U.S. national debt over the period 1997-2008 as compared to an alternative scenario in which Social
Security and other programs were indexed by a formula which subtracted the bias from the published CPI.
8
N
UMBER
12, S
PRING
2006
important differences between the two Com-
missions. The Boskin report concerned only the
CPI, while the Stigler report also covered the
Producer Price Index (PPI) and agricultural
price indexes. The Stigler report did not pro-
duce any numerical bias estimates, whereas the
mandate of the Boskin Commission included the
provision for a point estimate of the overall bias
in the CPI. The Stigler Commission had a sub-
stantial budget to commission new research
studies, whereas the Boskin Commission had no
research budget at all. Thus the Boskin Com-
mission report of necessity was a survey article
based in part on ongoing research, not a pro-
ducer of new research.
A novel aspect of the Commission's method
was to divide up the CPI into 27 categories and
develop a separate estimate of quality change
bias for each. This required some extrapolation
from categories for which research existed (e.g.
personal computers and TV sets) to related cat-
egories where no research existed (e.g. elec-
tronic toys). While some commentators have
complained of the subjective nature of the
Commission's assessment of quality change
bias, the Commission felt that to assume that
the bias was exactly zero in categories which
had not been the subject of previous research
was just as subjective as to extrapolate the
results of related research.
Summary of the Commission’s
Findings
Findings Related to Substitution
The Commission's findings are summarized
in Table 1. In this section we discuss the findings
related to substitution, comprising the first
three lines in the table. Next we discuss the cat-
egory of new products and quality change.
The CPI relied on fixed weight Laspeyres
indexes which did not account for consumer
substitution among commodities. These
Laspeyres measures of inflation were inherently
an upper bound, and empirical studies led the
Commission to conclude that this source of bias
amounted to about 0.4 percentage points per
year. Of this 0.4 percentage points, 0.15 came
from estimates of the effect of substitution
among the upper-level “strata” (apples vs.
bananas) and 0.25 from the effect of substitution
among the lower-level categories (Red Delicious
apples vs. Jonathan apples).
In addition to substitution bias among com-
modities, there is an outlet substitution bias, the
third line in Table 1, which refers to the practice
of the BLS in ignoring differences in prices for
the same item across outlets. Since price data are
collected
within
outlets, the shift of consumers
to purchases from discounters does not show up
in the CPI as a price decline even though con-
sumers reveal their preferences for these outlets
by their purchases, measured by the steady shift
in market share. We estimated this adds another
0.1 percentage point of upward bias.
Findings and Recommendations
Related to New Products and
Quality Change
The three types of substitution bias listed in
Table 1 account for just a little under half (0.5
percentage points) of the 1.1 percentage point
bias identified by the Boskin Commission.
Slightly over half (0.6 percentage points) results
from the difficulty of adjusting fully for quality
Table 1
Boskin Commission Estimates of Bias in
the Consumer Price Index in 1995-96
(percentage points per annum)
Source of Bias Estimate
Upper Level Substitution 0.15
Lower Level Substitution 0.25
Outlet Substitution 0.10
New Products / Quality Change 0.60
Total 1 .1 0
Plausible Range (0.80 — 1.60)
I
NTERNATIONAL
P
RODUCTIVITY
M
ONITOR
9
change and the introduction of new products.
Economists have known since Hicks (1940) that
the introduction of a new product should be
dealt with in a COL index by estimating its res-
ervation price and including the consumer sur-
plus attributable to the introduction of the
product. While accepting this framework, the
Boskin Commission did not make any of its own
estimates of the consumer-surplus value of new
products. Instead, it took the more cautious view
of primarily including estimates of explicit
dimensions of quality change and the late intro-
duction of major new products into the index.
VCRs, cellular phones, and other products were
included in the CPI a decade or more
after
they
had penetrated the market and
after
their price
had fallen by 80 per cent or more. Previous
research allowed an estimate to be made of the
bias in the CPI that occurred as a result of the
late introduction of these new products.
The Commission attempted to identify com-
modity categories in which the CPI was biased for
failing to take account of improvements in qual-
ity. To carry out this task, it examined separately
27 subcomponents of the CPI to determine the
extent of quality bias, if any, and established an
estimate for quality bias within each of the 27 cat-
egories. In eight of the categories the Commis-
sion discovered no research evidence or other
factor that would indicate a bias other than zero.
In the other 19 categories bias estimates were
assigned which ranged as high as 3.0 percentage
points per year for medical and hospital services
and 5.6 percentage points for appliances and elec-
tronic goods.
3
When these 27 bias estimates were
weighted by the relative importance of each cate-
gory in 1995 (based on 1982-84 expenditure
weights), the overall quality and new product bias
was determined to be 0.6 percentage points per
year.
4
While the Commission formulated a point
estimate of CPI bias related to quality change
and new products, it did not attempt any quanti-
tative estimate of changes in the quality of life.
Nevertheless, it did present an informal discus-
sion of changes in the quality of life and con-
cluded that the “good” outweighed the “bad.” It
cited a reduction in air and water pollution, a
decline in crime by various measures, a decline
in the suicide and infant mortality rates, and an
increase in life expectancy.
5
It also cited a wide
range of improvements in goods related to qual-
ity and new products for which the Commis-
sion's estimates made no allowance, namely “the
faster speed and reduced vibration of jet planes,
improved reliability of appliances and automo-
biles, improved sound quality of audio equip-
ment in homes and automobiles, improved
safety devices on home power tools and power
lawn mowers, reduction in the noise, weight,
and installation cost of room air conditioners,
and immeasurably better picture quality of color
TV sets.”
6
As other improvements for which no
allowance was made, it cited the spread of cable
and satellite TV and the new availability of the
World Wide Web to owners of personal
3 Table 2 of the Boskin Commission report attempted to discriminate between different periods in assigning the
estimates for quality change and the late introduction of new products. Thus, for instance, the bias estimate
for prescription drugs was 3.0 percentage points for 1970-95 and 2.0 percentage points for 1995-96, reflecting
the change in CPI methodology in 1995 which recognized the introduction of a generic version of a given drug
as a price decline.
4 The only case in which alternative weights were used was in the case of consumer appliances and elec-
tronic goods, where the 1982-84 CPI weights were deemed to be too low, and weights were taken instead
from the National Income and Product Accounts. See Boskin et al. (1996), Table 2, note a.
5 The one serious negative which was cited was the increase in births to unmarried women. Other “intangi-
ble negatives” cited were “increased job insecurity, possible increased inequality, and decreased job
opportunities for workers with only a high school education” (Boskin et al., 1996:76).
6 Although not identified as such in the Boskin report, this quote was copied directly from Gordon
(1990:560).
10
N
UMBER
12, S
PRING
2006
computers (another product which was intro-
duced into the CPI long after it began to become
common in homes). The Commission's section
on quality-of-life issues concluded that “...the
major increase in longevity... perhaps swamps
everything else. Accordingly, our estimate of the
current bias in the CPI is, if anything, probably
understated” (Boskin et al., 1996:77).
The Commission's
Recommendations
The Commission’s first and overarching rec-
ommendation was that the BLS should establish
the COL index as its objective in measuring con-
sumer prices. All of the other, more specific rec-
ommendations were aimed at achieving this
goal. The emphasis on this first recommenda-
tion may seem strange to economists, but the
BLS in its publications for decades had explicitly
stated that the CPI is not a COL index. The
Commission stated that a fundamentally sound
COL index could and should be developed.
7
As part of the short-run set of recommenda-
tions, the Commission recommended that the
single current CPI, which can never be revised
(due to its use in legal contracts and for index-
ation) be supplemented by a second “research-
based” index which is published annually rather
than monthly and is continuously revised to
incorporate new research results. The timely
monthly index would continue to be called the
CPI and would move toward a COL concept by
adopting a “superlative” index formula to
account for changing market baskets, abandon-
ing the Laspeyres formula at both the upper
level and lower level of aggregation.
At the upper level the Commission recom-
mended that the BLS should move away from the
assumption that consumers do not respond at all
to price changes in close substitutes, moving
instead to a “trailing Tornqvist” index (weighted
geometric mean of price relatives) or another
approximation to a superlative index, and also,
concurrently, to geometric means of price rela-
tives at the elementary aggregation level. These
changes would eliminate the problem of the
growing irrelevancy of market baskets based on
decade-old consumption patterns, reduce signifi-
cantly the substitution and (any remaining) for-
mula bias, and facilitate the speedier introduction
of new goods and services into the index.
The distinction between the “timely” CPI
and the new annual research-based index rests
on the fundamental proposition that the basic
monthly CPI can never be revised. The Com-
mission recommended that as subsequent data
became available, the weights were updated,
and new goods were introduced and the his-
tory of their price changes was extended back-
ward, the information incorporated in the
published CPI could undergo retroactive revi-
sion of a new annual cost-of-living index using
a compatible “superlative-index” formula that
would no longer be affected by the lag in the
availability of the required expenditure
weight data. This alternative COL index
would be published annually, with a lag of a
year or two, and would be subject to addi-
tional revisions after new information
emerges and new methodology is introduced.
Continuing with recommendations that would
require a somewhat longer term for implementa-
tion, the Commission felt that the BLS should
revise its approach to sampling. The Commission
was astonished at the number of price quotations
that the BLS collected on ordinary products like
bananas that were not subject to quality-change
or new-product bias, relative to the effort devoted
to collecting price data on new products like cel-
lular phones, personal computers, and computer
peripherals. In addition to reducing the data col-
lection effort devoted to apples and bananas, the
7 The BLS has stated (U.S. Bureau of Labor Statistics, 1997) that it has embraced the Commission’s overarching
recommendation concerning the objective of the CPI.
I
NTERNATIONAL
P
RODUCTIVITY
M
ONITOR
11
Commission believed that the data-collection
effort should be divided up between national and
local goods. In this revised scheme there would
no longer be any collection of data on appliances,
other consumer durables, and imported produce
(like tomatoes and bananas) in each separate city
— data for these national goods could be col-
lected in a much smaller sample applied to the
entire country. This would free up resources to do
a better job of collecting prices for local goods
with components that might vary across cities,
e.g., fuel costs, rent, household services, and non-
imported produce.
Criticism of the Commission's
Findings and Its Reaction
The report of the CPI Commission received
much attention. Most of the findings and rec-
ommendations have generally been accepted by
the economics profession, if we are to judge by
various symposia in which prominent academic
economists have been invited to comment on
the Commission's findings. External corrobora-
tion comes from several sources, including the
Federal Reserve’s study of productivity (Slifman
and Corrado, 1996), Nordhaus’ (1998) analysis
of survey data on changes in economic well-
being, and Diewert’s (1998) thoughtful comple-
mentary analyses of bias by type.
It is noteworthy that few if any criticisms
addressed the Commission's findings related to
substitution. Instead, most of the criticisms of
the Commission's findings centered on its treat-
ment of quality change and new products.
• The Commission did not give adequate
attention to quality deterioration (Abraham,
1997, U.S. BLS, 1997, Abraham et al.,
1998);
• The BLS already makes many quality adjust-
ments, which the Commission did not ade-
quately credit (Abraham, 1997, U.S. BLS,
1997, Moulton, 1996, Moulton and Moses,
1997);
• The Commission made too many back-of-
the-envelope calculations and was too will-
ing to generalize from research on one item
to research on related items, both with
respect to quality change and with respect to
the desirability of geometric means to deal
with lower level substitution bias (Abraham
et al.
,
1998, Nordhaus 1998);
• The Commission’s estimates, whether right
or wrong, are of limited use because they
cannot be generated from a mechanical pro-
cedure implementable in a month-to-month
price program (U. S. BLS, 1997, Abraham,
1997);
• We should have been more aggressive in our
estimates of the value of new products
(Hausman, 1997b, Nordhaus, 1998).
The Commission's Response
to the Critiques
8
Because there was so little criticism of the
Commission's estimates of substitution bias,
it is perhaps worth noting that research by
Shapiro and Wilcox (1997) suggests that the
bias could have been a tenth of a percentage
point or so higher than the estimate in our
report, which was based heavily on BLS
research.
9
The BLS response to our recom-
mendations about lower-level substitution
bias pointed out that, while the degree of sub-
stitution between some goods like white shirts
and blue shirts may be very high, that between
other goods like particular types of pharma-
8 This section combines sections of the jointly authored paper by commission members (Boskin et al., 1998)
with my own further observations.
9 Many of the best studies of substitution bias have been done by BLS researchers (see, e.g., Aizcorbe and
Jackman, 1993). The studies from systems of demand equations at high levels of aggregation generally
also come to an estimate of 0.2 to 0.25 on both U.S. data and data from other countries. See Greenlees
(1997) for further elaboration of the BLS research.
12
N
UMBER
12, S
PRING
2006
ceuticals may be very low (Abraham et al.,
1998:32). However, pharmaceuticals are the
exception, not the rule, and within the vast
majority of the 207 strata there is ample lati-
tude for substitution, whether we are talking
about types of food, shelter, apparel, cars,
tires, appliances, entertainment, or personal
care items. In fact, pharmaceuticals comprise
only 1.2 per cent of the upper-level weight in
the CPI. If anything, the unitary elasticity of
substitution embodied in the Commission's
recommendation of geometric weights at the
lower level probably understates the extent of
substitutability within the majority of the
strata.
The Debate over Quality Change
and New Product Bias
Most of the criticism focused on our extensive
analysis of quality change and new product bias
and our willingness to extrapolate a bias estimate
from one category to another, and this critique
of “guesstimates” is a major theme of Triplett’s
(2006) paper in this symposium. The notion that
assuming zero bias is scientific, whereas
attempting to generalize cautiously from related
goods or practical reasoning is not precise
enough, throws out available information. For
instance, even though we will never precisely
measure the value of the invention of the jet air-
plane, as economists we
know
that consumer sur-
plus triangles have an area that is positive rather
than zero.
Most of the Commission’s estimates of
quality change were based on the collection of
price data from independent sources and the
careful quality adjustment of those indepen-
dent data. Independent sources of price data
were employed in our bias estimates for shel-
ter, appliances, radio-TV, personal comput-
ers, apparel, public transportation,
prescription drugs, and medical care. Esti-
mates derived from these categories were
extrapolated, sometimes partially rather than
fully, to other house furnishings, nonprescrip-
tion drugs, entertainment, commodities, and
personal care. This left only a few remaining
categories where we added a bias estimate to
the CPI category in which there were already
quality adjustments, rather than computing
the bias estimate indirectly by subtracting an
independent estimate from the CPI estimate
for the same category. These categories were
food and beverages, other utilities, new and
used cars, motor fuel, and personal expenses.
The BLS did not object to our “down in the
trenches” approach to the problem. Indeed,
Moulton and Moses (1997:308) state, “This is
the first time that a systematic analysis of
quality bias has been done category by cate-
gory, which we consider to be a noteworthy
accomplishment of the Commission ...[the]
overall approach seems to us to be a sensible
and useful way to approach the problem of
coming up with an overall assessment of bias,
and we expect this type of structure will prove
to be useful in the future.”
Some outside critics of the Commission argued
that the BLS already does a great deal of quality
adjustment, and that the Commission report is
flawed for ignoring the extent of the BLS adjust-
ments.
10
However, for most categories, the extent
of current BLS quality adjustments is irrelevant
to an assessment of the Commission's treatment
of quality change. We were comparing our own
evidence to the corresponding CPI indexes —
however they are quality adjusted, in a major or
minor way — and thus our estimates of quality
change bias are a
residual
that remains after the
BLS has completed its efforts.
10 In fact most of the Moulton-Moses paper (1997) is devoted not to a direct critique of the Commission's esti-
mates, but to an explanation of how the CPI is adjusted for quality change and to an attempt to estimate the
quantitative significance of those adjustments (see pp. 322-48).
I
NTERNATIONAL
P
RODUCTIVITY
M
ONITOR
13
However, it is still instructive to discuss what
the BLS calls quality adjustment, since it illus-
trates the substantive and communication diffi-
culties in this field.
11
There is very little explicit
adjustment for quality change (Nordhaus,
1998). Most of the reported “quality adjust-
ment” by the BLS, 1.65 out of the 1.76 percent-
age points in Moulton and Moses (1997) comes
from “linking” procedures, where a missing item
is replaced by another (excluding outliers, com-
modity pairs where the implicit price-quality
differential exceeds 100 per cent, the quality
adjustment number shrinks to 0.3 per cent). No
judgment at all is made about the quality differ-
ential between the new and old item. Roughly
one out of three items disappear sometime dur-
ing the year and have to be replaced by a differ-
ent item in the same general class, such as a
larger versus a smaller package of yogurt or a
blue raincoat versus black. But this churning is
not what we had in mind by “quality change,”
which rather involves the appearance of new and
improved goods, greater speed, durability, vari-
ety, convenience, safety, and energy efficiency.
The area in which the Commission's bias esti-
mates are surely likely to be too low, rather than
too high, is in the area of new products. We did
not make explicit allowance for the late introduc-
tion into the CPI of numerous new products. We
indicated that the appropriate way to deal with
new products is to value the
consumer surplus
from their introduction, as first demonstrated by
Hicks (1940), and then elaborated and applied by
Hausman (1997a, 1999) to the case of cellular
telephones. We chose to deal with this by being
deliberately cautious, but indicating that there
was an asymmetrical bias with more potential bias
on the upside than the downside. We believed
that our overall estimates were conservative, both
by ignoring the many intangible aspects of quality
change, such as the improved safety of home
power tools or the improved quality of stereo
sound and TV pictures, and by omitting any
explicit valuation of truly new products.
Problems of Implementation
The BLS response to some of the Commis-
sion’s recommendations has been that they are
very difficult to implement in real time in a
monthly CPI program using mechanical rules
that are straightforward to implement. These
difficulties explain why we proposed that the
BLS calculate a second index, published annu-
ally, that would constantly be updated and con-
tinuously revised, and which need never be final.
This second index, if it had been implemented,
would have addressed many — not all — of the
BLS concerns about practicality.
12
Changes in the CPI
Since the Report
Despite its initial set of critical comments, the
BLS moved with surprising speed to implement
some of the Commission's recommendations. At
the same time the BLS introduced other changes
which were planned previously to the release of
the Commission report and for which the Com-
mission can take no credit. Subsequently, we will
discuss the new experimental BLS index based
on chain weights at the upper level; this is not
included in the following list of CPI improve-
ments because it has not been incorporated into
the basic CPI-U. Here we list the most impor-
tant changes changes that the CPI has intro-
duced since the December, 1996, release of the
Commission report.
13
Additional perspective on
11 Some of the material in this section comes from my published discussion of Moulton and Moses (1997).
12 The Stewart and Reed (1999) paper introduces a research-based and retrospective CPI as a one-time
project, not as an ongoing annual activity of the BLS.
13 The list of changes comes from background material supplied to Commission members with a 1999 Gen-
eral Accounting Office (GAO) survey (U.S. General Accounting Office, 2000) regarding post-Boskin CPI
measurement changes. This list has been checked against the list in Greenlees (2006).
14
N
UMBER
12, S
PRING
2006
these changes is provided by Greenlees (2006) in
this symposium.
1. Lower-level Geometric Weighting.
Effective with data for January, 1999, the BLS
introduced the geometric mean estimator for
index categories that comprise approximately 61
per cent of total consumer spending in the CPI.
This was expected to reduce the rate of increase
of the CPI by about 0.2 percentage points (Dal-
ton et al.
,
1998), and this estimate was later con-
firmed by Greenlees (2006).
2. More Rapid Change in Upper-Level
Weig hts.
In another major change that appears
to be a response to the Commission's recom-
mendations, the BLS now changes upper-level
weights much more rapidly than in the past
(U.S. BLS, 1999). Eleven years elapsed between
the initial use of 1982-84 weights in 1987 and
the switch to 1993-95 weights in 1998. More
recently the lag has dropped to a mere three
years: 1999-2000 weights were introduced in
January, 2002 and apply to CPI calculations for
the years 2002 and 2003, implying an average
three-year lag between the middle of the 1999-
2000 period used to calculate the weights and
the 2002-2003 period over which the CPI is cal-
culated using those weights. With the same
three-year lag structure, new weights were
introduced in January, 2004, January, 2006, and
so on in the future.
3. Change from Area- to Item-Based Sam-
ple Rotation Procedures.
In a change planned
before release of the Commission report, the
CPI switched its sample rotation methodology
in 1998 (Cage, 1996). It shifted its point-of-pur-
chase survey from time-consuming in-person
visits to computer-assisted telephone surveys,
which allow for an increase in sample size and
focus on specific item categories where products
turn over rapidly and where new products are
frequently introduced.
4. Changes in the Methods for Pricing of
Hospital Services.
In another change
planned before the release of the Commis-
sion's report and implemented in January,
1997, the BLS has improved its procedures for
pricing hospital services (Cardenas, 1996).
Instead of the old approach, which was a
straightforward input cost index that did not
reflect shifts in the use of inputs (e.g. shorter
hospital stays for a given procedure or a shift
from inpatient to outpatient treatment), the
new methodology obtains prices for a sample
of specified treatments for particular diseases,
rather than for a day in the hospital. This
approach had been introduced into the PPI in
1992.
14
Even though the PPI made no explicit
allowance for improvements in medical tech-
nology of the types incorporated into the
research reviewed by the Boskin Commission,
during the 1992-96 period the PPI for hospi-
tal services increased at an annual rate 2.0 to
2.5 percentage points less than the equivalent
CPI index, not far from the Commission's
estimate of a 3.0 percentage point upward bias
in the CPI for medical care services. Allowing
for the value of technological advances might
imply a CPI bias above 3.0 percentage points
for the pre-1997 period.
5. Treating Mandated Pollution Control
Measures as Price Increases.
As discussed in
Fixler (1998), the BLS switched in January, 1999
to treat changes in vehicle or motor fuel charac-
teristics arising from air pollution mandates as a
change in price rather than quality. This
reverses a BLS policy in effect since 1971 and
follows from a recommendation in the Commis-
sion report. Since most of the changes in auto-
mobile technology introduced to reduce air
pollution occurred during the 1970s and 1980s,
this change will have little future effect on the
CPI. This change provides another example of
14 This shift in methodology is discussed in Triplett (1999:3-4).
I
NTERNATIONAL
P
RODUCTIVITY
M
ONITOR
15
the need for a second, annual research-based
index that can be revised into the past, and in
this case the revisions would raise the growth
rate of the CPI.
15
6. Hedonic Price Indexes for Electronic
Products.
Again following in the footsteps of
the PPI, which had used the hedonic regres-
sion technique to adjust personal computer
prices as long ago as 1991, the CPI adopted
the hedonic regression approach for personal
computers in 1998. Similarly, the CPI intro-
duced a hedonic regression approach to
adjusting television set prices, beginning in
January, 1999. This technique was introduced
following the research of Moulton, LaFleur,
and Moses (1998). Greenlees (2006) shows
that so far this additional research on quality
change has had a very small impact on the
overall CPI, because the weights of the prod-
ucts involved are small, and that small sample
sizes often preclude developing hedonic
indexes from the regular CPI sample.
The Big Surprise: Upper-level
Substitution Bias is More
Important than We Thought
The above list of changes refers to the basic,
most-often quoted version of the CPI known as
the CPI-U. The list does not include the intro-
duction of chain weights at the upper level,
because the BLS decided not to take this step in
the CPI-U. Instead, in a history described in
detail in Cage et al.
(2003) and in Greenlees
(2006), the CPI decided to use chain weights not
in the basic CPI-U but in a new index called the
C-CPI-U that is intended as an official supple-
mental index rather than an experimental
research index.
The remarkable surprise after six years of
experience with the C-CPI-U, which the BLS
currently publishes from January 2000 to the
most recent month, is that the bias between
upper-level chain weights and Laspeyres
weights is much larger than anyone would have
guessed, including BLS staff and the Boskin
Commission. Despite the more frequent updat-
ing of weights in the Laspeyres CPI-U now than
in 1996, the difference between the C-CPI-U
and the CPI-U is very large, 0.38 percentage
points per year over the six years between Janu-
ary, 2000, and January, 2006.
16
In an amazing
coincidence, the difference for essentially the
same period, 1999:Q4 to 2005:Q4, between the
chain-weighted PCE deflator and the CPI-U is
exactly the same, 0.38 percentage points.
17
These facts lead to a reassessment of the quan-
titative importance of the Boskin evaluation and
the changes in the CPI over the past decade.
The Boskin Commission found substitution bias
at the upper level of 0.15 percentage points and
at the lower level of 0.25 points. One would have
thought that more rapid updating of weights at
the upper level and the movement to geometric
weights at the lower level would have eliminated
all but perhaps 0.1 points. But we now are faced
with six years of evidence of upper-level bias of
0.38 points, suggesting that the Boskin Com-
mission, relying primarily on previous BLS
15 Gordon (1990:351) Table 8.10 shows the time series of BLS quality adjustments for new automobiles divided
among safety, environmental mandates, and other factors. The environmental adjustments had the effect of
reducing the rate of inflation of new auto prices by 1.22 percentage points per year over the period 1967-85.
The Boskin Commission report found no net bias in the CPI for autos, taking a 0.94 percentage points per year
downward bias due to the treatment of environmental adjustments as a quality change, and cancelling that
out by a 0.95 percentage points upward bias due to the CPI’s neglect of the increased durability of autos. The
Commission’s treatment of increased auto longevity is validated by White (2006), who contrasts a median auto
lifetime of 10.5 years and 107,000 miles in 1977 with 13 years and 152,000 miles in 2001.
16 The annual growth rate of the CPI-U over this six-year period was 2.68 per cent and of the C-CPI-U was
2.31 per cent (the 0.38 percentage point difference quoted in the text allows for rounding error). It
should be noted that data after 2004 are preliminary.
17 The annual growth rate of the CPI-U over this period was 2.75 per cent and of the PCE deflator was 2.37
per cent.
16
N
UMBER
12, S
PRING
2006
research, substantially underestimated the
amount of upper-level substitution bias.
18
Implications of Research
Since the Boskin Commission
Report
The final section of this paper examines the
implications of research carried out since the 1996
release of the Commission Report. Our topics are
recent research on outlet substitution bias, long-
term historical evidence on CPI bias, and further
research on quality change and new products.
Outlet Substitution Bias
and the Wal-Mart Effect
The Commission had estimated outlet substi-
tution bias to contribute 0.1 percentage point per
year to the overall upward bias in the CPI, which
ignores changes in the level of prices between
full-priced and discount stores, assuming that the
price differential is fully offset by a service differ-
ential. But shifts in market share contradict that
assumption; when shoppers shift from full-price
to discount outlets, they are “voting with their
feet” that the price differential is worth more to
them than any service differential. In fact, much
of the shift in market share over the past two
decades has been from higher priced self-service
stores like Sears and K-Mart to lower-priced and
more efficient self-service stores like Wal-Mart
and Target. The level of service is often not an
issue, as shoppers move from one type of self-ser-
vice outlet to another.
Important new evidence on the Wal-Mart
effect for food at home is provided by Hausman
and Leibtag (2005), who study both bar-code
data on prices charged by each outlet as well as
household panel data that can track household
shopping patterns over time. Their discount
sellers of food include supercenters, warehouse
clubs, and mass merchants, and they cite sources
estimating that these outlets began selling food
in the late 1980s and by 2003 had achieved a 25
per cent market share of total food expenditures.
They find an average benefit of the introduction
of discount outlets to be 25 per cent of food
expenditures, consisting of 20.2 per cent for the
direct effect of lower prices at the supercenters,
and an additional 4.8 per cent coming from the
competitive responses of lower prices at tradi-
tional outlets. They also find that the average
rate of price change at the supercenters and the
traditional outlets is the same; that is, the benefit
of the supercenters comes from their much
lower prices when each supercenter opens for
business, a consumer benefit that is linked out
by the CPI.
If we assume that the 4.8 per cent price decline
at traditional outlets is accurately measured by
the CPI, then we can calculate the impact of the
supercenters as a 25 per cent market share, times
a 20 per cent price differential, or 5 per cent,
spread over roughly 15 years between 1988 and
2003, or an outlet substitution bias of 0.33 per-
centage points per year for food, which has a 12
per cent weight in the CPI. Thus food alone
would contribute 0.04 percentage points to the
outlet substitution bias estimated by the Boskin
Commission to be 0.1 percentage points. Pre-
sumably other durable and nondurable goods
would contribute the rest. It is doubtful that the
total of outlet substitution bias could be apprecia-
bly above the Boskin estimate of 0.1 percentage
points, because housing, medical care, and
numerous other types of consumer services are
not sold by discount stores.
19
18 The matter is more complicated than this, because there are numerous differences between the CPI-U and the
PCE deflator other than weighting schemes, whereas the difference between the CPI-U and C-CPI-U results
only from upper-level weighting differences. Thus the identical differences cited above over the past six years
may be in part a coincidence.
19 We note that discount chains now sell haircuts, and that big box retailers like Pet Smart now offer both
dog grooming and veterinarian services.
I
NTERNATIONAL
P
RODUCTIVITY
M
ONITOR
17
The Hulten-Brueghel Paradox
and its Implications for Quality
Change Bias
The Boskin Commission’s CPI bias estimate
of 1.1 percentage points per year was explicitly
applied to the period 1995-96, and the Commis-
sion suggested that the bias was 0.25 points
higher prior to 1995 and extending back to
1978, due to so-called “formula” bias in the
arithmetic mean formula used to combine indi-
vidual item prices.
20
What do we know about the
bias prior to 1978? Conjectures by Nordhaus
(1997) and responses by his discussant Hulten
(1997) led to the recognition that an upward bias
in the CPI in the order of magnitude of 1.5 per-
centage points a year cannot be extended back
for a century or two without the implication that
the standard of living in 1800 (Hulten, 1997) or
in 1569 (Gordon, 2005) was implausibly low. By
Gordon’s calculations, which extended Hulten’s
analysis back several more centuries, an annual
price index bias of 1.5 percentage points implies
for 1569 a median
annual
income in today’s prices
of only $5.60, enough to buy 0.8 ounces of pota-
toes per day, with nothing left over for food or
shelter. Gordon had chosen the year 1569 as the
year of death of Pieter Brueghel the Elder, who
had painted happy burghers “often shown as
overfed, content, well-clothed, and with solid-
looking houses in the background” (Gordon,
2005:4).
21
To resolve the paradox Gordon suggested that
at some point in the past the CPI bias must have
been zero or even negative, and to address this
possibility he carried out research on two of the
three major necessities, clothing and shelter. For
apparel his major finding is based on applying
two methodologies to the same data for wom-
ens’ dresses from the Sears Roebuck catalog
over the period 1914-88. Hedonic regressions
are compared with a matched-model index that
duplicates the CPI pre-1988 methodology by
comparing dresses from one year to the next that
are absolutely identical in every quality
attribute. It had long been suggested that the
matched-model methodology would miss price
increases that occur with model changes, and
indeed this is what occurred in the Sears data.
The annual growth rate of the hedonic index was
fully 2.9 percentage points faster than the
matched model index based on the same data
and 1.3 percentage points faster than the CPI for
womens’ dresses.
22
Gordon’s results include a
close comparison of quality in the 1914 and 1988
dresses, something that is not possible with the
CPI, with the conclusion that quality in the
Sears sample deteriorated over the full period.
He suggests that the CPI for apparel is roughly
accurate for 1914-47 but is downward biased by
roughly 1.5-2.0 percentage points per year for
1947-88. He also suggests that methodological
improvements in the CPI after 1988 may have
largely or completely eliminated the downward
bias, and in any case he has no evidence after that
date based on hedonic regressions.
Rental housing is the most important single
component of the CPI, because the price of
rental housing is used as a proxy for owner-occu-
pied housing. Gordon and vanGoethem (2005)
examine a wide variety of evidence, including a
large biennial set of panel data on rental apart-
ments from the American Housing Survey, and
conclude that the CPI for rental housing is
downward biased for most of the period from
20 The formula bias was partly eliminated by the January, 1995 introduction of a procedure called “seasoning”
that is explained in Greenlees (2006).
21 Brueghel dropped the “h” from his name in the last ten years of his life, but his sons retained the “h”.
22 These results are taken from Gordon (2005:Table 13). The fact that the Sears matched-model index grew
1.62 percentage points slower than the CPI could reflect differing sample sizes that causes the Sears
index to miss more of the price changes, and also a kind of outlet substitution bias reflecting Sears’ rela-
tively low prices compared to other merchants in the first two-thirds of the sample period.
18
N
UMBER
12, S
PRING
2006
1914 to 2003 at an average rate of roughly one
percentage point per year. They find that the
period of most rapid downward bias was in the
first half of the postwar era, the same time inter-
val that yielded the most rapid downward bias in
the Gordon apparel results, and that method-
ological improvements in the CPI had reduced
the downward bias to roughly one-third point
per year from 1995 to 2003.
These results on apparel and rental shelter
require a retrospective downward revision in the
Boskin Commission estimates of quality change
bias. The Boskin bias estimate for apparel was
+1.0 percentage points per year, and this should
be reduced to zero.
23
The Boskin bias estimate
for rental housing was +0.25 percentage points
per year, and this should be reduced to the
Gordon-vanGoethem regression-based estimate
of -0.46 percentage points per year.
24
Ta k i n g
these together and using the weights in the
Boskin report reduces the Boskin estimate of
quality change bias from 0.612 to 0.429 points
per year for the 1995-96 period.
Post-Boskin Research on Quality
Change: Medical Care and
Pharmaceuticals
In looking back to the Commission's estimate
of a 3.0 percentage points per year upward bias
in the CPI for medical care services, fully 2.0
percentage points are accounted for by the
change from input costs to a treatment basis,
introduced into the PPI in 1992 and the CPI in
1997. This leaves only 1.0 percentage point for
all remaining improvements in medical care
technology, and if anything, recent studies sug-
gest that medical care technology improvements
may be reducing the true price of medical care
by more than 1.0 percentage point per year.
25
In a newer version of research that the Com-
mission had reviewed, Cutler, McClellan, and
Newhouse (1999) use information from hospital
records to price heart attack treatments using
the CPI's traditional input-cost approach and
find an upward bias of 2.0 percentage points.
Adding in the value of improved life expectancy
from better treatment procedures, the bias rises
to the range of 3.1-3.5 percentage points, and
further research raises the bias estimate to 5.0
percentage points per year. A study that implies
a much larger bias was carried out by Frank,
Berndt, and Busch (1999) on treatments for
mental depression. There is no component of
the CPI that is directly comparable, but this
research implies a possible bias of more than 10
percentage points per year.
26
Ellison and Heller-
stein (1999) analyze a large data set on the prices
of the cephalosporin class of antibiotics and find
a price increase for 1988-96 of 0.76 per cent per
year compared to the PPI cephalosporin compo-
nent which rose at 4.54 per cent a year, for a bias
estimate of 3.78 percentage points per year.
27
Newhouse (2001, Table 5) surveys available evi-
dence at that time and considers a 3 percentage
points upward CPI bias to be conservative.
The number of aspects of medical care subject
to research has been growing, and it seems rea-
sonable to extrapolate from areas where studies
have occurred to those which may have similar
characteristics. Shapiro, Shapiro, and Wilcox
23 The Boskin estimate had been based on an earlier version of the Gordon paper that compared the CPI with a
matched-model index for all apparel (not just womens dresses) based on Sears catalogue data. The hedonic
study of womens’ dresses had not yet been performed at the time of the Commission deliberations.
24 This is the average of their estimate for 1985-95 of -0.58 and for 1995-2003 of -0.33.
25 Some of this research appears in Triplett (1999), a book with a comprehensive and thoughtful introduc-
tion that provides the best available introduction into the issues, problems, techniques, and results in
this area.
26 A follow-up study by Berndt et al. (2000:15) supports the earlier paper and “does not materially change
findings from previous research on treatment of depression.”
27 Related research available to the Commission was Griliches-Cockburn (1994) and Berndt et al. (1996).
I
NTERNATIONAL
P
RODUCTIVITY
M
ONITOR
19
(1999) point to the numerous dimensions of cat-
aract surgery. An operation that once required a
week in the hospital is now a brief outpatient
procedure. Subsequent recovery is much faster;
complication rates have declined; intraocular
lenses have replaced cumbersome cataract spec-
tacles. As a result of improved technology and a
drastic decline in prices, the rate of cataract sur-
gery among individuals in the United States
aged 65 years has increased by almost a factor of
four. Because the operation now occurs earlier
in the disease, the period of steadily obscured
vision has been eliminated, with enormous ben-
efits in welfare. The authors point to similar
benefits in angioplasty, joint replacement, and
laparoscopic removal of gall bladders.
Post-Boskin Research on Quality
Change: Other Products
Scattered pieces of new research have
emerged for other products. Moulton et al.
(1998) concluded that there was an upward bias
in the CPI for television sets of three to five per-
centage points over the period 1993-97. Ohashi
(1999) developed hedonic price indexes for
VCRs during the first decade of their introduc-
tion (1978-87) prior to their introduction into
the CPI and found an average rate of price
decline of 12 per cent per year over that period.
Hausman (1999) criticizes the Boskin Commis-
sion for understating the CPI bias related to cel-
lular phones and estimates that the CPI for
telephone service (taking account of the gradu-
ally increasing weight on cell phones) is biased
upward for 1985-97 by between 0.8 and 1.9 per-
centage points per year.
Abel, Berndt, and White (2003) find that the
prices of Microsoft software declined over 1993-
2001 at an annual rate of 4.26 per cent, even
without taking into account any improvements
in quality. A much more rapid rate of decline of
15 to 18 per cent is recorded by White et al
.
(2004) for personal computer operating systems
and of 13 to 16 per cent for software productiv-
ity suites. An even faster annual rate of decline of
21 to 26 per cent for PDAs (Personal Digital
Assistants) is found by Chwelos et al
.
(2004).
Other than Hausman’s studies of Cheerios,
cellular phones, and Wal-Mart, there has been
relatively little new research on the impact of
new products and new outlets, and yet specula-
tive estimates of the value of these improve-
ments were at the core of the Boskin
Commission estimates of CPI bias for quality
change and new products. The traditional
supermarket industry is in a state of upheaval as
consumers shift market share both toward the
cheaper supercenters studied by Hausman and
Leibtag (2005) and at the same time toward the
greater variety offered by such upscale markets
as Whole Foods. Both low-priced supercenters
and high-priced markets that emphasize organic
and in-house prepared foods are viewed as
attractive new products, as shown by the
revealed preference of consumer behavior.
Conclusion
In evaluating the Boskin Commission report
ten years later, two issues must be sharply distin-
guished. First, did the Commission overstate the
CPI bias for the period to which it referred,
1995-96? Second, how much have improve-
ments in the CPI reduced that bias?
On the first question, recent research on
apparel and housing suggests that the Commis-
sions’s estimate of quality change and new prod-
uct bias may have been roughly 0.2 percentage
points per year too high. However, the striking
0.38 percentage point annual difference over
2000-2006 between the CPI-U and the chain-
weighted C-CPI-U suggests that the Commis-
sion’s estimate of 0.15 points greatly understated
the significance of upper-level substitution bias.
This is especially true, given the much more rapid
updating of upper-level weights in the CPI for the
2000-06 period from which the 0.38 percentage
20
N
UMBER
12, S
PRING
2006
points number is calculated. It is possible that,
with this new information, the Commission’s
estimate of upper-level substitution bias for the
1995-96 interval should have been 0.45 to 0.50
points, instead of 0.15 points, more than offset-
ting the Boskin overstatement of quality change
and new product bias. Thus my own retrospective
view is that the upward bias in the CPI in 1995-96
was if anything higher than the Boskin estimate of
1.1 percentage points and was perhaps 1.2 or 1.3
percentage points.
An important piece of evidence on the nature
of CPI bias comes from the longstanding excess
of the annual growth rate of the CPI over the
PCE deflator. Any such excess is notable,
because the PCE deflator uses the same underly-
ing micro price indexes as the CPI but weights
them differently. As discussed above, the CPI-U
grew 0.38 percentage points faster than the PCE
deflator during 1999:Q4 to 2005:Q4, a period
when most of the improvements in the CPI were
already in effect. Over the period more relevant
to the Boskin Commission bias estimates, 1992-
98, the CPI-U grew 0.63 percentage points
faster than the PCE deflator. This 1992-98 dif-
ference, which can be explained only by item
and category substitution effects (because the
PCE has the same exact treatment of outlet sub-
stitution, quality change, and new product
effects), provides
prima facie
evidence that the
Boskin bias estimate may have been understated.
What about the second question, reductions
in the bias due to CPI improvements since 1996?
The greatest amount of progress has been made
in reducing substitution bias, both at the upper
level and lower level, although the thorny issue
of outlet substitution bias remains untouched.
However, in light of the continuing large differ-
ence in growth rates of the C-CPI-U and CPI-
U, these improvements have reduced substitu-
tion bias from a higher base than the Commis-
sion recognized. Several specific improvements
address part of the Commission's estimated
upward bias involving quality change and new
products, including new hedonic indexes for
television sets and personal computers as well as
an improved treatment-based methodology for
measuring medical care prices.
What is my own estimate of the current CPI
bias? The new C-CPI-U evidence suggests that
category and item substitution bias appears to
remain at about 0.4 percentage points per year,
outlet substitution bias remains at about 0.1 per-
centage points per year, and bias attributable to
quality change and new products has been
reduced from the Boskin-era 0.4 percentage
points (revised downward from 0.6 percentage
points as explained above) to perhaps 0.3 percent-
age points primarily as a result of the switch from
input prices to treatment prices for medical care.
This sums to 0.8 percentage points per year.
Concluding as did the Boskin Commission on
issues of unmeasured improvements and deteri-
oration, I think that these point estimates sub-
stantially understate the value of inventions,
new products, and increased longevity. A cen-
tury ago, our forefathers had to shovel coal,
carry water into their dwellings, and heat it
manually before bathing or the tedious scrub-
bing of clothes could take place. The value of
running water, water heaters, forced air heating
fueled by natural gas and the cleaner air that has
resulted, is enormous, even if converted to an
annual growth rate over 100 years. Recent
research on the value of increased life expect-
ancy creates growth rates in welfare that swamp
the Boskin debates about 10 basis points here or
there. In particular, Nordhaus (2002) and
related research concludes that the value of
increased longevity over the past century is as
large as the value of measured growth in all non-
health goods and services.
Let me conclude with a small personal exam-
ple. Despite global warming, it still snows occa-
sionally in Chicago, but I don’t have to touch a
snow shovel. My trusty Toro snow blower is a
I
NTERNATIONAL
P
RODUCTIVITY
M
ONITOR
21
new product from the perspective of 50 years
ago. The value of its invention is not included in
the CPI. But, not only does it greatly ease the
job of removing the snow compared to the old-
fashioned snow shovel, but it indirectly has con-
tributed to my life expectancy.
References
Abel, Jaison R., Ernst R. Berndt and Alan G. White
(2003) “Price Indexes for Microsoft’s Personal
Computer Software Products,” NBER Working
Paper 9968, September.
Abraham, Katharine G. (1997) “The CPI Commis-
sion: Discussion,”
American Economic Review
,
87
(2), May, pp. 94-98.
Abraham, K., Greenlees, John S., and Moulton,
Brent R. (1998) “Working to Improve the Con-
sumer Price Index,”
Journal of Economic Perspec-
tives
,
12
(1), Winter, pp. 27-36.
Aizcorbe, Ana M. and Patrick C. Jackman (1993)
“The Commodity Substitution Effect in CPI
Data, 1982-1991,”
Monthly Labor Review
116
(12), December, pp. 25-33.
Berndt, Ernst R., Anupa Bir, Susan H. Busch, Rich-
ard G. Frank and Sharon-Lise T. Normand
(2000) “The Medical Treatment of Depression,
1991-1996: Productive Inefficiency, Expected
Outcome Variations, and Price Indexes,” NBER
Working Paper 7816, July.
Berndt, Ernst R., I. Cockburn, and Z. Griliches
(1996) “Pharmaceutical Innovations and Mar-
ket Dynamics: Tracking Effects on Price Indexes
for Anti-Depressant Drugs,”
Brookings Papers on
Economic Activity
, Microeconomics Issue, pp.
133-188.
Boskin, Michael J., E. Dulberger, R. Gordon, Z.
Griliches, and D. Jorgenson (1996) “Toward a
More Accurate Measure of the Cost of Living,”
Final Report to the Senate Finance Committee,
December 4.
Boskin, Michael J., E. Dulberger, R. Gordon, Z.
Griliches, and D. Jorgenson (1998) “Consumer
Prices, the Consumer Price Index, and the Cost
of Living,”
Journal of Economic Perspectives,
12
(1), Winter, pp. 3-26.
Cage, Robert (1996) “New Methodology for Select-
ing CPI Outlet Samples,”
Monthly Labor Review
119 (12), December, pp. 49-61.
Cage, Robert, Greenlees, John S., and Jackman,
Patrick (2003) “Introducing the Chained Con-
sumer Price Indices,” in Thierry Lacroix (ed.),
International Working Group on Price Indices
(Ottawa Group): Proceedings of the Seventh Meeting
(
Paris: INSEE) pp. 213-246.
Cardenas, Elaine M. (1996) “Revision of the CPI
Hospital Services Component,”
Monthly Labor
Review
119 (12), December, pp. 40-48.
Chwelos, Paul D., Berndt, Ernst R., and Cockburn,
Iain M. (2004) “Faster, Smaller Cheaper: An
Hedonic Price Analysis of PDAs,” NBER Work-
ing Paper 10746, September.
Cutler, David, McClellan, Mark, and Newhouse,
Joseph P. (1999) “The Costs and Benefits of
Intensive Treatment for Cardiovascular Dis-
ease,” in Triplett (ed.), 1999, pp. 34-71.
Dalton, Kenneth V., Greenlees, John S., and Stew-
art, Kenneth J. (1998) “Incorporating a Geomet-
ric Mean Formula into the CPI,”
Monthly Labor
Review
121 (10), October, pp. 3-7.
Diewert, W. Erwin (1998) “Index Number Issues in
the Consumer Price Index,”
Journal of Economic
Perspectives,
12 (1), Winter, pp. 47-58.
Ellison, Sara Fisher, and Hellerstein, Judith K.
(1999) “The Economics of Antibiotics: An
Exploratory Study,” in Triplett (ed.), 1999,
pp. 118-143.
Fixler, Dennis (1998) “Treatment of Mandated Pol-
lution Control Measures in the CPI,”
CPI
Detailed Report
, September, pp. 4-7.
Frank, Richard G., Berndt, Ernst R., and Busch,
Susan H. (1999) “Price Indexes for the Treat-
ment of Depression,” in Triplett (ed.),
pp. 72-117.
Gordon, Robert J. (1990)
The Measurement of Dura-
ble Goods Prices
(Chicago: University of Chicago
Press for NBER).
Gordon, Robert J. (2000) “The Boskin Commission
Report and its Aftermath,” in Mick Silver and
David Fenwick (eds.),
Proceedings of the Measure-
ment of Inflation Conference
(Cardiff UK: Cardiff
University), pp. 258-82. Also available as NBER
Working Paper 7759, June.
Gordon, Robert J. (2005) “Apparel Prices 1914-93
and the Hulten/Brueghel Paradox,” in C. Cor-
rado, E. Diewert, and C. Hulten (eds.),
Price
Index Concepts and Measurement,
Conference on
Research in Income and Wealth (Chicago: Uni-
versity of Chicago Press for NBER, forthcom-
ing) Also NBER Working Paper 11548, August.
Gordon, Robert J. and vanGoethem, Todd (2005) “A
Century of Downward Bias in the Biggest CPI
Component: The Case of Rental Shelter, 1914-
2003,” in E. Berndt and C. Hulten (eds.),
Hard-
to- Measure Goods and Services: Essays in Honor of
Zvi Griliches,
Conference on Research in Income
and Wealth (Chicago: University of Chicago
Press for NBER), forthcoming, also NBER
Working Paper 11776, November.
Greenlees, John S. (1997) “Expenditure Weight
Updates and Measured Inflation,” unpublished.
22
N
UMBER
12, S
PRING
2006
Greenlees, John S. (2006) “The BLS Response to the
Boskin Commission Report,”
International Pro-
ductivity Monitor
, Number 12, Spring, pp. 23-41.
Griliches, Z. and I. Cockburn (1994) “Generics and
New Goods in Pharmaceutical Price Indexes,”
American Economic Review
, December, pp.1213-
1232.
Hausman, Jerry (1997a) “Valuation of New Goods
Under Perfect and Imperfect Competition,” in
Bresnahan, T, and Gordon, Robert J. (eds.),
The
Economics of New Goods
(Chicago: University of
Chicago Press for NBER) pp. 209-37.
Hausman, Jerry (1997b) “The CPI Commission:
Discussion,”
American Economic Review
, 87
(2),
May, pp. 94-98.
Hausman, Jerry (1999) “Cellular Telephone, New
Products and the CPI,”
Journal of Business and
Economic Statistics.
Hausman, Jerry and Leibtag, Ephraim (2005) “Con-
sumer Benefits from Increased Competition in
Shopping Outlets: Measuring the Effect of Wal-
mart,” NBER Working Paper 11809, December.
Hicks, John R. (1940) “The Valuation of the Social
Income,”
Economica
,
7
(26), May, pp. 105-24.
Hulten, Charles R. (1997) “Discussion of Nordhaus,”
in Bresnahan T., and R.J. Gordon (eds.),
The
Economics of New Goods
(Chicago: University of
Chicago Press for NBER) pp. 66-70.
Moulton, Brent R. (1996) “Bias in the Consumer
Price Index: What is the Evidence?”
Journal of
Economic Perspectives
, 10 (4), Fall, pp. 159-177.
Moulton, Brent R. and Moses, Karin E. (1997)
“Addressing the Quality Change Issue in the
Consumer Price Index,”
Brookings Papers on Eco-
nomic Activity
, 28 (1), pp. 305-49.
Moulton, Brent R., LaFleur, Timothy J., and Moses,
Karin E. (1998) “Research on Improved Quality
Adjustment in the CPI: The Case of Televi-
sions,” presented to the Conference of the
Ottawa Group, April.
Newhouse, Joseph P. (2001) “Medical Care Price
Indices: Problems and Opportunities, the
Chung-Hua Lectures,” NBER Working Paper
8168, March.
Nordhaus, William D. (1997) “Do Real-Output and
Real-Wage Measures Capture Reality? The His-
tory of Light Suggests Not,” in Bresnahan T.,
and R.J. Gordon (eds.),
The Economics of New
Goods
(Chicago: University of Chicago Press for
NBER) pp. 29-66.
Nordhaus, William D. (1998) “Quality Change in
Price Indexes,”
Journal of Economic Perspectives
,
12 (1), Winter, pp. 59-68.
Nordhaus, William D. (2002) “The Health of
Nations: The Contribution of Improved Health
to Living Standards,” NBER Working Paper
8818, March.
Ohashi, Hiroshi (1999) “Quality-Adjusted Price
Indexes for Home Video Cassette Recorders in
the US, 1978-87,” presented at NBER Summer
Institute, July 21.
Shapiro, Matthew D., and David W. Wilcox (1997)
“Alternative Strategies for Aggregating Prices in
the CPI,” NBER working paper 5980, March.
Shapiro, Irving, Matthew D. Shapiro, and David W.
Wilcox (1999) “Quality Improvement in Health
Care: A Framework for Price and Output Measure-
ment,” NBER working paper 6971, February.
Slifman, L., and C. Corrado (1996) “Decomposition
of Productivity and Limit Costs,”
Occasional Stuff
Studies
, Federal Reserve Board, November 18.
Stewart, Kenneth J. and S. B. Reed (1999) “CPI
Research Series Using Current Methods, 1978-
98,”
Monthly Labor Review
122 (6), June, pp. 29-38.
Stigler, George, ed. (1961)
The Price Statistics of the
Federal Government. Report to the Office of Statisti-
cal Standards
, Bureau of the Budget. New York,
National Bureau of Economic Research, 1961.
Triplett, Jack (ed.) (1999)
Measuring the Prices of
Medical Treatments
, Washington, Brookings.
Triplett, Jack (2006) “The Boskin Commission
Report After a Decade,”
International Productivity
Monitor
, Number 12, Spring, pp. 42-60.
U.S. Bureau of Labor Statistics (1997) “Measure-
ment Issues in the Consumer Price Index,”
June
1997, response to the U.S. Congress
, Joint Eco-
nomic Committee.
U. S. Bureau of Labor Statistics (1999) “Scheduled
Updates for Expenditure Weights in the Con-
sumer Price Index,”
CPI Detailed Report
, Febru-
ary, pp. 5-6.
U. S. General Accounting Office (2000),
Consumer
Price Index: Update of Boskin Commission's Esti-
mate of Bias.
Washington, draft report, January.
White, Alan G., Abel, Jaison R., Berndt, Ernst R.,
and Monroe, Cory W. (2004) “Hedonic Price
Indexes for Personal Computer Operating Sys-
tems and Productivity Suites,” NBER Working
Paper 10427, April.
White, Joseph B. (2006) “Cars Last Longer, Driving
a Shift in All Aspects of the Auto Sector,”
Wall
Street Journal
, February 28, pp. D3.
