View original document

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

October 15, 1995

Federal Reserve Bank of Cleveland

The Consumer Price Index
and National Saving
by Michael F. Bryan and Jagadeesh Gokhale


Llthough a majority of U.S. lawmakers now favor the goal of balancing the
federal budget within the next decade,
there is little consensus on how to
achieve it. One current proposal is that
the Consumer Price Index (CPI) should
be adjusted to better reflect the cost-ofliving increases that result from inflation. ' If adopted, this measure will not
only improve the long-run deficit outlook but, more importantly, will boost
the nation's flagging saving rate.
There are two possible ways to implement such a policy. First, it may be introduced only for a particular program. For
example, Social Security benefits—the
largest federal transfers—may be indexed for inflation at a lower-than-CPI
rate each year. This course was recommended in the early 1980s by Martin
Feldstein, then chairman of President
Reagan's Council of Economic Advisers.2 In this scenario, the reduction in
inflation indexing could be viewed as a
"deductible" against Social Security's
provision of retirement benefits and
other insurance.
Alternatively, such a policy may be
introduced comprehensively by adjusting the CPI itself.3 This approach is
motivated by the belief that, for several
technical reasons, the CPI overstates
actual increases in the cost of living.
Using a downward-adjusted price index
to measure cost-of-living changes would
decrease future outlays for all federal
transfers that are indexed by law. In
ISSN 0428-1276

addition, it would increase federal tax
revenues because of a slower upward
revision in income tax brackets—which
are also indexed for inflation.
Although the desire to reduce annual federal deficits is behind the current interest
in this idea, the real benefit of a new costof-living index will show up in higher
national saving.4 This is true because a
new index will not, by itself, lower the
government's absorption of resources.
Rather, increased national saving will
result from the lower private consumption induced by the policy. This Economic Commentary summarizes the case
for redefining the CPI and provides estimates of the additional national saving
that would be generated if such a policy
were implemented beginning in 1996.

• Causes of Upward
Bias in the CPI
To one seemingly direct question, "How
much have prices risen?" a statistician
quickly counters with another: "Prices of
what?" If the composition of goods and
services changes between two periods,
we can record price changes for the
things we consumed in an earlier period,
for the things we consume in the current
period, or for some combination of the
two. From a statistical perspective, there
is little basis for preferring any of these
approaches. Indeed, it is possible to
imagine many "correct" answers to the
first of the questions posed.

Adjusting the CPI to reflect the true
increase in the cost of living and lowering cost-of-living adjustments for
Social Security benefits will alleviate
two long-range problems—escalating
federal budget deficits and exceedingly low national saving.

Economists usually recast that question.
"When faced with changing prices,"
they ask, "how much would we have to
compensate someone to keep him just as
well off as before?" Here, of course,
there is only one correct answer, which
economists have dubbed the change in
the cost of living. To date, however, no
one has proposed a practical way to
quantify this change, and the task has
largely been thrown back to the statisticians. In constructing the CPI, the
Bureau of Labor Statistics (BLS) has
chosen to measure the cost of goods and
services consumed in a previous period
in terms of current prices.
In all likelihood, this approach overstates
actual changes in the cost of living, since
it does not account for the changes people make in the composition of goods
that they purchase—often in response to
price changes. If we were to compare
current spending patterns with those of
1982-84 (the period on which the CPI is
now based), we would expect to see a
variety of differences.

Presumably, household spending would
shift toward goods whose prices rise
more slowly than average and away
from those whose prices rise more
rapidly than average. Because the BLS
does not adjust the market basket of
goods and services to account for these
cost savings, the change recorded by the
CPI tends to overstate the true change in
the average person's cost of living. In
economists' language, if we were to
compensate the average person on the
basis of CPI changes, he or she would
most likely be better off than before.
Thus, the dilemma caused by using the
CPI to adjust government expenditures
like Social Security payments is that the
adjustment will tend to exceed the actual
rise in recipients' cost of living and so
will make them better off, when the purpose of the adjustment was just to prevent their standard of living from falling.
The compositional changes already
mentioned, which economists call "substitution effects," are only one type in a
long list of potential compositional
changes that might produce an upward
bias in the CPI. Others include alterations in retailing patterns (recently shown
to be a potentially important source of
CPI bias caused by the arrival of discount stores), and improvements in the
quality of goods that might be recorded
as cost increases, but in fact reflect better, more satisfying products. Another
compositional change occurs when new
goods are introduced, since they previously had no recorded price. Yet, these
many possible kinds of CPI bias have a
common origin: Tying the market basket
of goods to an earlier time causes the
index to overstate actual cost-of-living
changes because the composition of
goods that people purchase is altered in a
way that either minimizes cost changes
or maximizes benefit changes.
From a statistical perspective, there is no
obvious way to "fix" the CPI. The BLS
could try to update the market basket
more frequently (adjustments are now
made about every 10 years), but that
would require additional, expensive surveys of household expenditure patterns.
Moreover, there is no guarantee that the
resulting index would conform more
closely to the "true" cost of living. In
fact, it is easy to show that such adjust-







SOURCE: Authors' calculations based on National Income and Product Accounts data.


Government consumption rates
Private consumption rates






SOURCE: Authors' calculations from Survey of Current Business, various issues.

Percent of resources consumed






Age of consumer



SOURCE: Jagadeesh Gokhale, Laurence J. Kotlikoff, and John Sabelhaus, "Understanding the Decline in U.S.
Saving: A Cohort Analysis" (footnote 7).

ment can result in a price index that consistently understates the cost of living.
Because the CPI is not likely to be fixed
soon, and because it very probably contains an upward bias, the most practical
course may be merely to adjust the costof-living estimate by some amount. But
how much of an adjustment is appropriate? No one is certain, and the range of
answers has been staggeringly wide.
Federal Reserve Chairman Greenspan
recently suggested that the bias was in
the neighborhood of 0.5 to 1.5 percentage points per year, a figure that is consistent with a recent Federal Reserve
Board staff estimate.5 Other assessments of CPI bias, given in testimony
before Congress, have ranged from a

mere 0.2 percent per year to 1.7 percent
or more. Research results from the Federal Reserve Bank of Cleveland fall at
the low end of the range (about 0.7 percentage point per year, on average).6 Furthermore, the Cleveland Fed's research
shows that CPI bias may vary substantially over time (from about 1 percentage
point per year between 1967 and 1981, to
virtually zero between 1982 and 1992).
After reviewing the wide range of opinions, a panel of economists chosen by the
Senate recently concluded that annual
CPI bias falls between 1.0 and 1.5 percentage points. Although unlikely to be
the final word on the subject, this suggests that the appropriate adjustment may
be about 1 percentage point per year.

(Billions of 1993 dollars)
Percentage-point change




Reduction in Social Security COLA




Revised CPI




NOTE: Net national saving was $152.1 billion in 1993.
SOURCE: Authors' calculations.


The CPI and National Saving

To Consume—or to Save?
The net national saving rate (NNSR) is
the percentage of net national output that
is not consumed during the year. It is
defined as one minus the sum of the private consumption rate and the government consumption rate. The NNSR has
fallen substantially in the United States
since the late 1970s (figure 1). Of the
two consumption rates (private and government), it is the private rate's increase
that is entirely responsible for the decline in NNSR (figure 2). Understanding
why the rate of private consumption has
risen over the last 15 years is the key to
understanding how adjusting the CPI to
correct for its upward bias will help
restore the NNSR to its pre-1970 levels.
According to a popular approach to consumption behavior, individuals consume
a fixed fraction of their resources at each
age. This fraction is known as the propensity to consume out of resources. As
figure 3 shows, consumption propensities rise with age: Older people tend to
consume at a faster rate than do younger
ones because they have a shorter remaining lifespan over which to finance consumption spending.7 Knowledge of consumption propensities and calculations of
the change in resources that would follow
a redefinition of the cost-of-living index
can be used to estimate the changes in
NNSR that the policy may produce.8

The Impact of
COLA/CPI Revisions
Table 1 shows the impact on national
saving of revising the CPI and lowering
the cost-of-living adjustment (COLA)
for Social Security retirement, survivor,
and disability benefits, beginning in
1996. The numbers reported are for

1993, the base year for the calculations,
when net national saving was $152.1 billion. In general, the impact of reducing
the Social Security COLA is positively
related to the amount of the reduction.
Cutting the COLA by 1 percentage point
would increase net national saving by an
estimated $43.4 billion, which is 28.5
percent of actual 1993 saving.
Revising the CPI downward would
boost national saving by much more,
since it would reduce not only Social
Security benefits, but also outlays for
federal railroad, civilian, and military
retirement, Supplemental Security Income (SSI), food stamps, and child nutrition.9 It would also increase incometax revenues, all else being equal,
because tax brackets would rise more
slowly in the future.10 The table shows
that a 1 percent reduction in the CPI
would add nearly $76 billion to national
saving—almost half as much as actual
1993 saving. Even a conservative reduction in the CPI of 0.5 percentage point
would produce close to $39 billion more
in saving.
We should note that many private pension
benefits are also partially or fully indexed
for inflation. Hence, a downward revision
in the CPI would induce a reduction in
private pension wealth and would add to
the impact on national saving already
described. Because this source of increase is not included in our calculations,
the numbers we report are likely to understate the saving impact of this policy.


ing. Such changes are justifiable because
the current practice of indexing benefits
to the CPI very probably overcompensates for cost-of-living increases that
result from inflation.'' However, the
most important benefit of this policy will
be an increase in national saving, which is
now exceedingly low. If allowed to continue, such low saving may compromise
future economic growth. But even minor
adjustments to the CPI could improve the
net national saving rate significantly.
Adopting the policy described in this
Economic Commentary is likely to
simultaneously alleviate two long-range
problems—escalating federal budget
deficits and low national saving.



1. The CPI, a measure of the level of prices in
the economy, is based on a standard market
basket of goods and services purchased by a
typical worker's family.
2. See Martin S. Feldstein, ed., American
Economic Policy in the 1980s. Chicago: University of Chicago Press, 1994, pp. 1-79. This
type of policy has also been proposed recently
by Senators Robert Kerrey and Alan K. Simpson as part of their plan to overhaul federal
entitlement spending.
3. This is another of the measures included in
the Kerrey-Simpson reform proposals.
4. Some may equate the deficit with negative
public saving and so translate a deficit reduction directly into an increase in public and,
other things being equal, national saving.
However, because the deficit is an arbitrary
accounting construct, from the viewpoint of
economic theory the distinction between public and private saving is also arbitrary. Thus,
inferences about changes in saving based
on changes in deficits are not valid. For a full
discussion, see Alan J. Auerbach and Laurence
J. Kotlikoff, "Demographics, Fiscal Policy,
and U.S. Saving in the 1980s and Beyond," in
Lawrence H. Summers, ed., Tax Policy and
the Economy, vol. 4. Cambridge, Mass.: MIT
Press, 1990, pp. 73-101.
5. See David E. Lebow, John M. Roberts,
and David J. Stockton, "Economic Performance under Price Stability," Board of Governors of the Federal Reserve System, Working
Paper No. 125, April 1992.


The current national debate about how to
reduce the federal deficit has stimulated
interest in lowering Social Security
COLAs and adjusting the CPI to better
reflect the true increase in the cost of liv-

6. See Michael F. Bryan and Stephen G.
Cecchetti, "The Consumer Price Index as a
Measure of Inflation," Federal Reserve Bank
of Cleveland, Economic Review, vol. 29, no. 4
(Quarter 4 1993), pp. 15-24.

7. The consumption propensities shown here
are reported in Jagadeesh Gokhale, Laurence
J. Kotlikoff, and John Sabelhaus, "Understanding the Recent Decline in United States
Saving: A Cohort Analysis," Federal Reserve
Bank of Cleveland, Working Paper, 1995
8. One concern about using consumption
propensities out of wealth is their stability
over time. The available evidence suggests
that these profiles have been fairly stable
since the mid-1980s. The additional national
saving after a policy change is estimated by
computing each generation's implied change
in total resources times its age-specific average propensity to consume. Summing over
all members of the generation and then over
all generations alive provides the required
estimate. Of course, the implicit assumption
here is that average and marginal propensities to consume out of resources are equal.
9. Program-specific decay factors were used
to calculate the reductions in Social Security
and railroad, civilian, and military retirement
benefits resulting from reduced inflation
indexing. This is necessary because benefits
for each year's new retirees will be determined independent of earlier years' COLA
reductions. For example, new retirees' Social

Federal Reserve Bank of Cleveland
Research Department
P.O. Box 6387
Cleveland, OH 44101
Address Correction Requested:
Please send corrected mailing label to
the above address.
Material may be reprinted provided that
the source is credited. Please send copies
of reprinted materials to the editor.

Security benefits in the year 2000 will be
determined by their past covered earnings
and will not be affected by COLA reductions
applied between 1996 and 1999 to the benefits of earlier retirees. Benefit projections for
SSI, food stamps, and child nutrition are
based on a fixed formula applicable to everyone in a given year and so do not require the
use of decay factors. The decay factors for
each benefit type were provided by the
Office of Management and Budget (OMB).
10. The method of computing additional
future income-tax revenues from this policy is
somewhat complicated. The OMB-provided
baseline revenue (based on the 1994 MidSession Review of the United States Budget)
is used to recover a projection of "revenue
without indexation" by applying a loss factor
of 0.5 times the cumulative CPI. The new
revenue series is then converted into the postpolicy revenue projection by applying the
adjusted cumulative CPI. The difference
between the baseline and post-policy projection is the additional revenue resulting from
downward adjustment of the CPI.

if the CPI did not overstate the true inflationinduced change in the cost of living, implementing such a policy would still generate
additional national saving. However, this
Economic Commentary does not intend to
suggest that we should therefore do away
with indexing for inflation altogether.

Michael F. Bryan is an economist and consultant and Jagadeesh Gokhale is an economic advisor at the Federal Reserve Bank
of Cleveland.
The views stated herein are those of the
authors and not necessarily those of the Federal Reserve Bank of Cleveland or of the
Board of Governors of the Federal Reserve

11. The indexing of federal transfers and
income-tax brackets is designed to protect
individuals and households against losses
due to general price inflation. Note that even

U.S. Postage Paid
Cleveland, OH
Permit No. 385