Automatic Cost of Living Adjustments: Some Economic and Practical Considerations







Prepared for Members and Committees of Congress



Public policy discussions often involve dollar amounts. Because of rising prices, however, the
purchasing power of those dollar amounts changes over time. For that reason, policy makers have
seen fit in some cases to arrange for those amounts to be increased automatically, as prices rise, to
keep their purchasing power unchanged. Without such an arrangement, either the real value of
those amounts would fall or policy makers would have to take periodic steps to increase them to
offset the effects of inflation.
Most of the effect on the federal budget of automatic COLAs may be accounted for by three
programs: individual income taxes, Social Security, and income support programs tied to the
poverty thresholds.
The consumer price index (CPI) is based on prices and quantities of goods and services, which
can be directly observed. It measures the change in income required to purchase a fixed
marketbasket of goods and services, but that is not a “true” measure of change in the cost of
living. A true cost-of-living index would measure the change in income required to maintain a
constant level of satisfaction.
If existing measures of price change exaggerate what increase in income is needed to maintain
consumer satisfaction, then increases in benefit payments that are indexed are larger than they
need to be to maintain their value to beneficiaries, and income tax brackets shift upward more
than necessary to avoid bracket creep. Both effects tend to increase the budget deficit. There are
other price indexes that may be superior measures of change in the “true” cost of living, but they
have some practical disadvantages.
One standard for choosing a price index for automatic COLAs might be which one is most
appropriate. In some cases, it might be argued that what is appropriate is not a measure of price
change but rather a measure of change in income.
With respect to public policy in general, it has been argued that indexation may have the effect of
inhibiting policy makers from making changes to those programs that are indexed. It may be that
indexing a benefit implies a kind of guarantee of future benefits, making any future change likely
to be more contentious. It may be that there is a loss of policy discretion and an increase in policy
inertia with respect to those programs that are indexed. Indexed benefits rise over time with no
regard to changes in underlying economic conditions. However, budget priorities are constantly
shifting, and if economic growth is sluggish and revenues are less than expected, that may put
additional budgetary pressure on those programs that may be relatively more subject to discretion.
This report will not be updated.






Major Federal Programs Affected by Price Indexes.................................................................1
Price Indexes.......................................................................................................................1
Income Taxes......................................................................................................................1
Social Security....................................................................................................................2
Civil Service Retirement System........................................................................................2
Federal Employee Retirement System................................................................................2
Military Retirement.............................................................................................................3
Railroad Retirement............................................................................................................3
Poverty Thresholds.............................................................................................................3
The Cost of Living from a Theoretical Perspective..................................................................3
Price Index Complications..................................................................................................4
Indexing Alternatives................................................................................................................6
BLS’s Experimental Index for the Elderly..........................................................................6
The Chain-Weighted Consumer Price Index.......................................................................7
Other Price Indexes.............................................................................................................8
“Diet” COLAs.....................................................................................................................8
Additional Considerations.........................................................................................................8
Author Contact Information............................................................................................................9





ublic policy discussions often involve dollar amounts. Because of rising prices, however,
the purchasing power of those dollar amounts changes over time. For that reason, policy
makers have seen fit in some cases to arrange for those amounts to be increased P


automatically, as prices rise, to keep their purchasing power unchanged. Without such an
arrangement, either the real value of those amounts would fall or policy makers would have to
take periodic steps to increase them to offset the effects of inflation.
These automatic cost-of-living adjustments (COLAs) are also referred to as indexing. Values that
are adjusted automatically are said to be indexed because they are linked to an index of prices. As
might be expected, there are a number of consequences that follow from indexing.
One implication of indexing is that the index itself becomes an object of scrutiny to those who
might otherwise have no interest in it. The index also becomes a potentially powerful policy lever
in that it directly influences the federal budget, both outlays and revenues. Indexing may also be a
source of policy inertia in that it implies that the amount that is indexed is less susceptible to
discretionary changes.
This report looks at how indexing influences the budge and identifies major programs that have
indexing provisions. It also explains what price indexes attempt to measure and discusses some of
their weaknesses. Finally, it points out some practical things to keep in mind when establishing an
indexing provision.
Most of the effect on the federal budget of automatic COLAs may be accounted for by three
programs: individual income taxes, Social Security, and income support programs tied to the
poverty thresholds.
Indexation usually relies on one of two consumer price indexes (CPI). Both are published
monthly by the Bureau of Labor Statistics (BLS) of the Department of Labor. One is the
consumer price index for all urban consumers (CPI-U), and the other is the consumer price index
for urban wage earners and clerical workers (CPI-W). Both are intended to track changes in the
prices of the goods and services purchased by a typical household in the populations surveyed. Of 1
the two, the CPI-U reflects a larger share of the population. CPI data are usually released toward
the end of the month following the month in which the data are collected.
A number of elements in the individual income tax are automatically adjusted for inflation.
Indexation was introduced beginning with the Economic Recovery Tax Act of 1981 (P.L. 97-34),
and the number of provisions of the individual income tax code that are indexed has continued to
grow. In the absence of indexation, inflation in combination with progressive tax rates tends to
raise the average effective tax rate. As an individual’s income rises to keep pace with rising

1 For more detail on how the CPIs are calculated, see CRS Report RL30074, The Consumer Price Index: A Brief
Overview, by Brian W. Cashell.



prices, more of it is subject to higher marginal tax rates even though there has been no real gain in
purchasing power. This has been referred to as “bracket creep.” Linking the income levels that
define the tax brackets to a price index prevents that from happening. Only those increases in
income that represent real gains above and beyond inflation are subject to higher tax rates. Other
elements of the individual income tax code that are linked to a price index are the earned income 2
tax credit, the standard deduction, and the personal exemptions.
These automatic adjustments are based on changes in the CPI-U. Specifically, the automatic
adjustment in each year is based on the average level of the CPI-U for the 12-month period ended
in August of the previous year, relative to the average CPI-U for the 12-month period ended in 3
August of a particular base year. For example, the inflation adjustment factor for the standard
deduction for any given year is determined by the ratio of the average CPI-U for the 12-month
period ended in August of the previous year to the average CPI-U for the 12-month period ended
in August 1987. The data required to calculate the automatic adjustments for the coming year are
usually available in late September.
Social Security benefit payments are adjusted in January every year to reflect increases in the
CPI-W. The adjustment is based on the average value of the CPI-W for the three months of the
third calendar quarter in the previous year relative to the average of those same three months from
the year before. For example, the COLA for January 2008 will be based on the percentage change
in the CPI-W between the third calendar quarter of 2006 and the third calendar quarter of 2007.
The data required to calculate the COLA for the coming year are usually available in late October.
Retirement benefit payments under the Civil Service Retirement System (CSRS) are
automatically adjusted in January of each year to reflect increases in the CPI-W. Like the Social
Security COLA, the adjustment is based on the third quarter over third quarter change in the CPI-
W.
Retirement benefit payments under the Federal Employees Retirement System (FERS) are also
based on the third quarter over third quarter increase in the CPI-W. There are, however, some
additional adjustments to calculate the COLA. If the increase in the CPI-W is 2% or less, then the
COLA is equal to the CPI-W increase. If the increase in the CPI-W is more than 2% but less than
3%, the COLA is 2%. If the increase in the CPI-W is more than 3%, the COLA is 1% less than
the increase in the CPI-W. This COLA only applies to FERS retirees aged 62 and over.

2 See James C. Young, “Inflation Adjustments Affecting Individual Taxpayers in 2006: Analysis and Commentary,”
Tax Notes, October 3, 2005, pp. 105-114.
3 The base years are different for different components of the income tax.





Military retiree pension benefit payments are adjusted in the same way as Social Security
benefits. The January increase in benefit payments is increased by the third quarter to third
quarter increase in the CPI-W.
Benefits paid to retirees in the railroad retirement system are adjusted in the same way as Social
Security benefits. The January increase in benefit payments is increased by the third quarter to
third quarter increase in the CPI-W.
Poverty thresholds (or some percentage of them, usually more than 100%) are used to establish
eligibility for a number of federal government income support programs, including food stamps,
Medicaid, the subsidized portion of Medicare prescription drug coverage, and the low income
home energy assistance program (LIHEAP). Benefits paid by the programs are not indexed, but
the poverty thresholds themselves are adjusted each year to reflect increases in the CPI-U. Each
year, the poverty thresholds are increased by the increase in the CPI-U between the two prior
calendar years. For example, the poverty thresholds for 2007 were calculated by increasing the

2006 thresholds by the increase in the average CPI-U for 2006 over the average CPI-U for 2005.


The CPIs, as they are now calculated, are based on prices and quantities of goods and services,
which can be directly observed. They measure the change in income required to purchase a fixed
marketbasket of goods and services; however, that is not a “true” measure of change in the cost of
living. A true cost-of-living index would measure the change in income required to maintain a
constant level of satisfaction, or “utility,” in the jargon of economists.
The idea of utility is pervasive in economics. With a given income, which constrains choices,
consumers decide how to spend their money based on how much utility, or satisfaction, is derived
from the various goods and services available. There is, however, no unit of measure for utility, so
there is no way of quantifying the utility associated with a particular good or basket of goods and
services. The inevitable gap between any practicable cost-of-living measure and its theoretical
ideal makes it difficult to know exactly how close any actual measure may be to that ideal.
The issue of price index accuracy may seem arcane, but it is not trivial. Suppose that existing
measures of price change exaggerate what increase in income is needed to maintain consumer
satisfaction. If that is the case, then increases in benefit payments that are indexed are larger than
they need to be to maintain their value to beneficiaries. That means federal outlays rise more than
necessary to maintain the value of those benefits. Similarly, income tax brackets shift upward
more than necessary to avoid bracket creep, and tax revenues are less than they would be if the
price index were measured ideally. A price index that overstates the true increase in the cost of
living thus causes outlays to be higher than necessary and revenues to be less than was intended.
Both effects tend to increase the budget deficit.





There are a number of weaknesses in the CPIs as they are now calculated. Whether they cause
overstatement is subject to dispute. In 1996, a special commission established by the Senate
Finance Committee (the “Boskin” commission, named after its chairman) issued a report 4
estimating that the CPI overstated inflation by about 1.1% per year. Since that report was
released, BLS has introduced a number of changes that may have reduced some of any upward 5
bias in the CPI.
The CPIs are fixed-weight price indexes.6 From one period to the next, they measure the change
in total price of an unchanging mix of goods and services. Consumers, however, regularly vary
their spending habits.
Even as the average level of prices rises, relative prices also change. Some prices rise more
rapidly than others, and in some cases, prices may fall. Over time, as relative prices change,
consumers will tend to buy more of those goods and services whose prices are rising less rapidly
than average and fewer of those goods and services whose prices are rising more rapidly than
average. By changing spending patterns in this way, consumers can partially insulate themselves
from the general increase in prices, at least as measured by a fixed-weight price index like the
CPI.
To the extent that spending patterns change solely in response to changes in relative prices, it may
be that the CPI tends to overstate the effect of inflation on consumer well-being. However,
spending patterns change for other reasons. Over time, tastes and preferences change. Food
purchases may change, for example, for dietary reasons. As incomes rise, consumers may demand
more luxuries—for example, substituting restaurant meals for those prepared at home. If
changing spending patterns simply reflect changing consumer preferences, it is less likely that the
CPI overstates the effects of rising prices on consumer well-being.
BLS now publishes a price index that attempts to eliminate any bias resulting from changing
consumer spending patterns. The chained consumer price index for all urban consumers (C-CPI-
U) is not a fixed-weight index. To measure the change in prices between two periods, it uses both 7
prices and quantities from both periods, and so takes changing marketbaskets into account. Data
for the C-CPI-U are available beginning with December 1999. Although there is some variation
in the differences, on average, increases in the C-CPI-U have been between 0.3% and 0.4%
slower than the CPI-U.
Another difficulty with the CPIs is the introduction of new goods to the indexes. Typically, the
indexes register the change in price of existing goods and services, but when a new good is
introduced, there is no price for it in the previous period, so there is usually a delay before new
goods can be introduced to the index. Cell phones, for example, were added to the CPI

4 Advisory Commission to Study the Consumer Price Index, Michael Boskin, Chairman, “Toward a More Accurate
Measure of the Cost of Living: Final Report to the Senate Finance Committee, December 4, 1996, p. 93.
5 The Boskin Commission report was not without its critics. See Economic Policy Institute, Getting Prices Right: The
Debate Over the Consumer Price Index, edited by Dean Baker, 1997, p. 190.
6 BLS currently updates the CPI marketbasket every two years.
7 Unlike the CPIs the C-CPI-U is subject to revision. For a detailed explanation of the C-CPI-U, see CRS Report
RL32293, The Chained Consumer Price Index: How Is It Different?, by Brian W. Cashell.





marketbasket years after their introduction, and thus the CPI failed to reflect significant price
declines as they became ubiquitous.
The characteristics of goods also change over time, making it difficult to compare prices from one
period to the next. For example, cars are safer, color television replaces black and white, standard
sets are replaced by high definition ones, and computers make more calculations per second. A
true cost-of-living index would take into account that some of any change in price in a good
might be due to the change in characteristics.
Measuring the value of a change in quality is not always straightforward. In some cases, an
improved good may be sold alongside its outmoded version, and the difference in price can be
taken as a measure of the value of the improvement. Even that method may be flawed, however,
because the introduction of a new model may cause the price of the older one to fall because it is
perceived as obsolete. The introduction of a new generation of computers, for example, may
cause the price of the previous one to fall because consumers know that future software releases
may not run on that model.
One method that BLS uses to adjust some goods for quality change views goods as bundles of
characteristics or attributes. For example, personal computers are adjusted for quality change
based on the attributes of their components, such as processor speed, amount of memory, and
storage capacity. Changes in the characteristics of many goods and services, however, are more
difficult to quantify, and so quality adjustments are far from straightforward.
The characteristics of goods are not the only changes that must be addressed in a cost-of-living
index. How those goods are sold also changes over time. In recent years, the proliferation of
buying clubs and other “big box” retailers has provided consumers with a lower-price alternative.
The shift to self-service gasoline stations is another example. Just as the CPI may be slow to
reflect changes in the mix of goods and services that consumers buy, changes in the nature of
retail markets are not immediately taken into account.
BLS efforts to improve the CPI as a cost-of-living index began long before the release of the
Boskin Commission report, and they continue. Specific changes subsequent to the Boskin report
include the introduction of the new C-CPI-U, more frequent updating of the expenditure weights 8
used in the CPIs, and increased efforts at quality adjustments, especially for durable goods.
Robert Gordon, who was a member of the Boskin Commission, recently published an assessment 9
of the CPI 10 years after the commission’s report. He found that in some cases, the commission
may have overestimated the bias because of inadequate adjustment for quality change in the CPI,
but may have underestimated the upward bias attributable to inadequate accounting for the effects
of consumer substitution. Reviewing the post Boskin Commission changes in CPI methodology,
Gordon estimates the current upward bias in the CPI to be 0.8 percentage points.

8 John S. Greenlees, “The BLS Response to the Boskin Commission Report,” International Productivity Monitor,
Spring 2006, pp. 23-41.
9 Robert J. Gordon, “The Boskin Commission Report: A Retrospective One Decade Later,” National Bureau of
Economic Research Working Paper 12311, June 2006, p. 35.





A study published by BLS also reviewed some of the recent methodological changes in the CPI.10
It found that some of the bias reported by the Boskin Commission that was due to consumer
substitution had been reduced. The study also found that although BLS has greatly expanded
efforts to make adjustments for quality change, those adjustments have not had a significant
quantitative effect on the index itself. The BLS study also points out that some have argued it is
possible that some quality adjustments are overdone and that there could even be a downward
bias in the CPI. Most studies, however, have found reason to believe that there is some upward
bias. There also seems to be some suspicion that the value of life-extending medical advances, as
well as many technological advances, cannot help but be undervalued.
Once a benefit or an income threshold has been established, it is not surprising that provision is
made for inflation protection of those amounts. Most automatic adjustments are tied to one of two
CPIs, the CPI-U or the CPI-W. There are, however, other price indexes published by the
government that measure change in consumer prices. None is perfect, the choice may reduce to
what is most appropriate and what is practical.
Social Security benefits are indexed to the CPI-W. At the time the benefit COLA was first
established, there was only one CPI; the CPI-U was not published until 1978. The CPI-W tracks
the prices for goods and services purchased by households deriving more than 50% of their
income from wages and clerical workers’ earnings. That would seem to exclude many retirees,
and it might be that the CPI-U would be a more appropriate index to adjust Social Security for the
effects of inflation. From a practical standpoint, however, it would not have made that much
difference. Between 1978 and 2006 the CPI-W increased at an annual rate of 4.0%, and the CPI-
U increased at an annual rate of 4.1%.
In the case of Social Security beneficiaries, the claim is often made that the CPI-W does not
reflect the spending patterns of retirees. In particular, it is argued that the elderly spend more than
the average household on health care, and that health care costs rise more rapidly than prices for
other goods and services. For that reason, some have suggested that a separate price index for the
elderly population should be calculated.
Beginning with December 1983, BLS began calculating an experimental index for the elderly
using data already collected for the CPI program. If the purpose of indexing Social Security
benefits was to maintain the value of those benefits to retirees in an environment of rising prices,
it might seem appropriate to use a price index that tracked the goods and services purchased by 11
Social Security beneficiaries.
On average, the elderly consume a different basket of goods and services from the overall
population. They tend to spend more on health care and less on transportation, for example. As a
result, historically, the experimental index for the elderly has risen about 0.3 percentage points

10 David S. Johnson, Stephen B. Reed, and Kenneth J. Stewart, “Price Measurement in the United States: A Decade
After the Boskin Report,” Monthly Labor Review, May 2006, pp. 10-19.
11 Not all Social Security beneficiaries are elderly.





per year faster than the CPI-W.12 The difficulty with using a separate price index for the elderly to
adjust Social Security benefits is that, just as the inflation experience of the elderly varies from
that of the overall population, so too are there differences in spending patterns within the elderly
population itself. Switching to a price index would still leave some whose inflation experience
was underestimated by the index, but they would be fewer.
Although the experimental index differs from the CPI-W, as might have been expected, there are
other considerations to keep in mind. Because it is an experimental index, fewer resources are
used in its construction. It relies entirely on data collected for the estimation of the other CPIs.
The survey on which the marketbasket is based was considerably smaller than the one used for
the other CPIs and thus more subject to sampling error. Further, no additional survey was
conducted to determine whether the elderly shop in different outlets from the overall population.
If BLS were to initiate a more formal program to estimate change in the cost of living of the
elderly, it might yield different results.
One further consideration is that, if the experimental index for the elderly is found wanting, any
automatic cost-of-living adjustment affecting the elderly might at least rely on the CPI-U rather
than the CPI-W, since the elderly are under-represented in the CPI-W. Although the CPI-U might
be the more appropriate of the two, it would be unlikely to make much practical difference.
Referred to above, the chain-weighted consumer price index for all urban consumers (C-CPI-U)
differs from the CPI-U in that it takes into account changes in spending habits over time resulting
from changes in relative prices. For that reason, it is considered to be closer to a true measure of
changes in the cost of living.
Historically, the C-CPI-U has increased more slowly than the other CPIs. The annual difference 13
has typically been about 0.3 percentage point but has been as large as 0.8 percentage point.
While the C-CPI-U might be considered a better measure of changes in the cost of living, it might
be unsuitable for use in any automatic COLA provision. The reason is that unlike the CPI-U or
CPI-W, the C-CPI-U is subject to revision. Once the CPI-W and CPI-U are released, those
numbers are final and will not change. The C-CPI-U, however, is subject to two revisions after its
initial release. In the case of data for each January, the wait for the final statistic is the longest.
The first estimate is released in the following February. The first revision is released in February
of the following year, and the final revision is released in February the year after that. The wait
for the final number for any given January is more than two years.
If timeliness is an important aspect of COLAs, that may present a problem. In an era of stable
inflation rates, it might make little practical difference if COLAs were based on an increase in
prices that happened as much as two years earlier. In an era of accelerating inflation, however, 14
delayed adjustments would result in declines in the real value of the indexed amount.

12 See CRS Report RS20060, A Separate Consumer Price Index for the Elderly?, by Brian W. Cashell.
13 See CRS Report RL32293, The Chained Consumer Price Index: How Is It Different?, by Brian W. Cashell.
14 Although simplicity may be considered an important aspect of any COLA provision, it would be possible to make
adjustments to subsequent COLAs to reflect revisions in earlier data.





Using the first release of the C-CPI-U in an automatic COLA provision would be more timely but
might present other complications. For example, if a given benefit payment were indexed to the
first release of the C-CPI-U and subsequent revisions indicated a faster rate of increase,
beneficiaries would likely be displeased.
There is also a broad array of price indexes associated with the economic accounts used to
generate estimates of gross domestic product (GDP). Each of the components of GDP has a
corresponding price index. One of those is referred to as the chain-type price index for personal
consumption expenditures (PCE). The PCE price index tracks the prices of all household
consumption. Like the C-CPI-U, it also takes into account that household spending patterns
change over time, and so it avoids some of the overstatement of inflation that is typical of a fixed-
weight price index.
The PCE price index, as well as the other price indexes associated with the GDP accounts, is
useful as an economic indicator because when the methodology used to estimate it changes, the
entire time series is revised. At a given time, the PCE price index time series is based on a
uniform methodology. The PCE price index, however, is subject to more frequent revisions
which, like the C-CPI-U, may make it unattractive for use in any automatic COLA.
PCE price index data are quarterly. The first release is published in the month following the end
of the quarter to which the data refer (the reference quarter). That figure is revised again the
following month, and again the month after that. Further, it is subject to revision in each of the
next three Julys. Additional revisions are possible after that.
There is another possible automatic COLA provision, and as mentioned above, there is a version
of it applied to pensions in the FERS retirement system. This is sometimes referred to as a “diet”
COLA. In this indexing arrangement, the automatic adjustment is something less than the full
increase in the price index used. That might be done to limit the budgetary consequences of a
benefit program over the long run, but it might also be done because the CPIs are believed to
overstate the effect of rising prices on households. The historical gap between the C-CPI-U and
the CPI-U thus might be seen as a basis for reducing the automatic adjustment that would result
from a direct link to the CPI-U.
Aside from any theoretical issues related to indexing, there are a few practical matters that may
be important. First is that the timing of the COLA must take into account that the data will be
available. That may seem obvious, but it might be prudent to leave some time between when the
data are released and when the COLA takes effect. There are two reasons for that. One is for there
to be time to advertise the COLA to those affected. Another reason is the possibility that the
release of the data might be delayed. In 1995, for example, because of controversy over the
federal budget at the end of 1995, there was a temporary government shutdown. That shutdown
resulted in delays in the publication of the CPIs for December 1995 and January 1996.





One standard for choosing a price index for automatic COLAs might be which one is most
appropriate. However, in some cases, it might be argued that what is appropriate is not a measure
of price change but rather a measure of change in income. The poverty thresholds, as described
above, are adjusted annually by the change in the CPI-U. Using a price index to adjust them
implies that they represent an absolute standard of poverty, but some may consider them to be
relative standards that have meaning only in the context of the overall distribution of income. If
poverty is seen as a relative measure, it might be more appropriate to adjust it to keep pace with
rising incomes.
From time to time, indexing the statutory federal minimum wage has also been discussed.15 Like
the poverty thresholds, the minimum wage might be viewed either as an absolute standard, in
which case, if it were to be automatically adjusted, it might make sense to link it to a price index.
If, however, it is considered to be a relative standard for worker pay, it might make more sense to
link it to a measure that reflected changes in average wages.
In cases where the benefit to be indexed is intended for a specific purpose, such as health care as
opposed to general income support, it might be more appropriate to use a price index that
reflected changes in the cost of health care.
With respect to public policy in general, it has been argued that indexation may have the effect of 16
inhibiting policy makers from making changes to those programs that are indexed. It may be
that indexing a benefit implies a kind of guarantee of future benefits, making any future change
likely to be more contentious. It may be that there is a loss of policy discretion and an increase in
policy inertia with respect to those programs that are indexed. Indexed benefits rise over time
with no regard to changes in underlying economic conditions. However, budget priorities are
constantly shifting, and if economic growth is sluggish and revenues are less than expected, that
may put additional budgetary pressure on those programs that may be relatively more subject to
discretion.
Brian W. Cashell
Specialist in Macroeconomic Policy
bcashell@crs.loc.gov, 7-7816


15 See CRS Report RL33791, Possible Indexation of the Federal Minimum Wage: Evolution of Legislative Activity, by
William G. Whittaker.
16 R. Kent Weaver, Automatic Government: The Politics of Indexation, The Brookings Institution, 1988, pp. 276.