Skip to main content

Chained CPI Can Be Part of a Balanced Deficit-Reduction Package, Under Certain Conditions

A proposal included in several deficit-reduction packages — those from fiscal commission co-chairs Erskine Bowles and Alan Simpson, the Domenici-Rivlin panel, and the Senate "Gang of Six" — would shift from the regular or official Consumer Price Index (CPI) to the "superlative" or "chained" CPI when indexing various federal benefit and tax-code provisions. This proposal would gradually trim the growth of benefit programs and boost future tax revenues.

Because many economists believe the official CPI is upwardly biased and regard the chained CPI as a more accurate measure of inflation, the Center on Budget and Policy Priorities has long supported switching to the chained CPI for adjusting federal benefits and taxes — if it is accompanied by several necessary adjustments to prevent significant hardship, as described below. The policy clearly amounts to a reduction in future Social Security benefits, which many find objectionable. However, we believe that the chained CPI is a reasonable component of a comprehensive package to put the budget on a sustainable course, under the following conditions:

  • The change needs to apply to both benefit programs and the tax code, and the proceeds from applying the chained CPI to the tax code should go entirely for deficit reduction. Using some or all of those proceeds to finance a reduction in tax rates should not be acceptable.
  • The change should be accompanied by a modest benefit increase (or "bump") for long-time Social Security beneficiaries. Switching to the chained CPI would lower Social Security benefits (relative to currently scheduled benefits) by amounts that grow with each additional year that an individual receives benefits. This could lead to hardship among very old people. As a result, a modest benefit increase for people who have received benefits for many years should accompany the switch to the chained CPI. Both the Bowles-Simpson and Rivlin-Domenici plans include this adjustment along with their chained CPI proposal.
  • Policymakers should exempt the Supplemental Security Income (SSI) program from the switch or make other changes to SSI to mitigate the chained CPI's impact. Applying the chained CPI to SSI, which serves the nation's poorest elderly and disabled people and still leaves them well below the poverty line, raises particular problems. Policymakers should exempt SSI from the switch or soften the impact on SSI beneficiaries as described below.
  • The proposal should specifically identify the programs to be moved to the chained CPI to avoid confusion or unintended effects. The CPI appears in hundreds of places in the U.S. Code and other laws, affecting not just annual cost-of-living adjustments (COLAs) but also caps, eligibility thresholds, reporting requirements, fines or penalties, and payment rates for various programs. Only about a dozen such provisions dominate the budgetary effects of switching to the chained CPI. To avoid confusion and future litigation, lawmakers should amend specific statutes as appropriate, rather than enacting blanket language, and focus on major provisions with significant budgetary effects — essentially annual cost-of-living adjustments in retirement and related benefit programs and annual inflation adjustments in the tax code.

Using Chained CPI Would Affect A Number of Programs and Save Significant Amounts

Many of the federal government's retirement, disability, and income-support programs — including Social Security, federal civilian and military retirement, railroad retirement, SSI, and veterans' compensation and pensions — pay annual COLAs that are linked to the CPI. Many elements of the tax code — notably tax brackets, personal exemptions, standard deductions, limits on contributions to 401(k) plans and similar accounts, and key parameters of the earned income and child tax credits — are also adjusted annually for the CPI.[1] In the case of benefit programs, the goal is to preserve recipients' purchasing power. In the case of taxes, the goal is to protect filers from owing higher taxes in real (inflation-adjusted) terms when their incomes are unchanged in real terms (that is, when their incomes rise only with inflation). [2]

Since 1996 — partly in response to the recommendations of the Advisory Commission to Study the Consumer Price Index [3] — the Bureau of Labor Statistics (BLS) has made numerous improvements to the official CPI that have caused it to rise more slowly than it otherwise would have.[4] In addition, BLS began publishing an alternative, "chained" CPI to address the commission's concerns about what is called "upper-level substitution bias." [5] That change in methodology, however, did not apply to the official CPI, which continues to suffer from upper-level substitution bias.

The chained CPI, first available for December 1999, has grown on average by about 0.3 percentage points per year more slowly than the official CPI. The Social Security actuaries assume the gap between the two CPIs will continue to average 0.3 percentage points per year in the future; CBO assumes the gap will average 0.25 percentage points per year. CBO has estimated that moving to the chained CPI on January 1, 2012 would reduce the deficit by over $220 billion in the first ten years (not counting the interest savings that would result; see Table 1), though this figure exceeds the likely savings in the coming decade for two reasons.

First, January 1, 2012 is already behind us. Second, the $220 billion figure does not include the costs of the needed adjustments in Social Security and other programs described later in this report, without which this is not a reasonable policy. The savings in the first ten years thus would be smaller, probably in the $100 billion to $150 billion range if the change takes effect several years from now (after the economy has recovered more). The savings would then grow substantially in subsequent decades.

Table 1:
Budgetary Effects of Adopting the Chained Consumer Price Index Effective December 201
(By fiscal year, in billions of dollars)
  2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 Total, 2012-21
Federal retirement and veterans' benefits -0.3 -0.8 -1.3 -1.7 -2.2 -2.6 -3.0 -3.6 -4.0 -4.5 -24.0
Social Security -1.4 -3.3 -5.3 -7.5 -9.8 -12.1 -14.5 -16.9 -19.3 -21.9 -112.0
Revenues 0.7 1.6 3.1 5.2 6.8 7.8 9.8 11.0 12.3 13.7 71.8
Deficit -2.4 -5.7 -9.7 -14.4 -18.8 -22.5 -27.3 -31.5 -35.6 -40.1 -207.8
Source: Congressional Budget Office, Reducing the Deficit: Spending and Revenue Options, March 2011, mandatory spending options 26 and 27 and revenue option 3. Omits effects on SSI, railroad retirement, and certain smaller programs. Revenues include revision made by JCT in June 2011.

Over the long run, the Social Security actuaries estimate that moving to the chained CPI — together with a benefit "bump" for longtime beneficiaries, which we view as an essential adjunct — would close almost one-fifth of the program's actuarial gap over 75 years, and over one-tenth of its gap in the 75th year.[6]

Current Experimental CPI-E Not a Good Alternative

Some opponents of using the chained CPI note that an experimental CPI for the elderly, or CPI-E, has often tended to grow faster than the official CPI.[7] The CPI-E gives significantly greater weight to medical care and housing, and lower weight to other items such as food, transportation, and education and communication) because a larger share of the expenditures of elderly people, as compared to the population as a whole, goes for medical care and housing while a smaller share goes for other items. [8] Over the last 20 years, using the CPI-E would have led to higher COLAs than using the official CPI in 14 years, lower COLAs in four years (including the January 2012 COLA), and a zero COLA in 2010 and 2011 just as the official CPI did; it would have made little cumulative difference (see Table 2).

Table 2:
Average Cost-of-Living Adjustment Using . . .
  CPI-W (actual) Chained CPI CPI-E
Last 20 years 2.6% n.a. 2.7%
Last 10 years 2.5% 2.2% 2.5%
Last 5 years 2.3% 2.0% 2.1%
January 2012 COLA 3.6% 3.4% 2.9%
CPI-W = Consumer Price Index for Urban Wage and Clerical Workers (current law); CPI-E = Experimental CPI for the Elderly. Chained CPI is first available for December 1999; because the chained CPI is subject to revision for two years, CBPP assumes that COLAs would be based on original (not revised) values, an approach outlined by CBO in "Using a Different Measure of Inflation for Indexing Federal Programs and the Tax Code," Congressional Budget Office Economic and Budget Issue Brief, February 24, 2010. All averages (20, 10, and 5 years) include the January 2012 COLA.
Source: Center on Budget and Policy Priorities based on data from Bureau of Labor Statistics.

Chiefly because medical care has a higher weight in the CPI-E and rises faster than other prices, the Social Security actuaries assume that the CPI-E will grow faster than the standard CPI by an average of 0.2 percentage points a year. Using it to determine Social Security COLAs would worsen the program's deficit by about one-fifth over 75 years, and by about one-tenth in the 75th year. [9]

The fact that the chained CPI may understate inflation, on average, for the elderly because they face higher out-of-pocket health costs is a drawback to switching to the chained CPI. Nevertheless, it is not a reason to forgo the deficit reduction that moving to the chained CPI would bring. To the contrary, we believe the best policy is to switch to the chained CPI but to accompany this move with other measures and adjustments.

We reach this conclusion for several reasons:

  • Increases in health care costs affect elderly people unevenly — seniors' high average out-of-pocket spending on health care is skewed upward by a minority of elderly people who face very high health care expenses. In other words, seniors' median out-of-pocket expenses (the expenses of the typical elderly person) are well below the mean (or average) expenditure amount. [10] Policymakers can address the very high expenses these seniors pay, in part, by restructuring Medicare cost-sharing in a deficit-neutral fashion so that Medicare beneficiaries who can afford to do so pay modestly more upfront for health care and much of the resulting savings are used to strengthen Medicare's protection against catastrophic expenses.
  • Even with such a change, the chained CPI will still understate inflation for many elderly people. While the effect would be small in any single year, it would compound over time, hitting beneficiaries hard in their 80s and 90s. This is a particular concern because as people grow older, their other income typically shrinks and they may exhaust much or all of their assets, even as their medical costs continue to rise.
  • Policymakers can mitigate this problem through other modest adjustments in Social Security and SSI. Both the Bowles-Simpson and Domenici-Rivlin plans proposed a bump in Social Security benefits — essentially an average increase of 5 percent phased in over several years — for beneficiaries after they have been receiving benefits (and hence, reduced COLAs every year) for 15 or 20 years. Both Bowles-Simpson and Domenici-Rivlin proposed that the bonus be an equal dollar amount for all eligible beneficiaries (rather than an equal percentage increase) in order to target the benefit adjustment toward those who need it most, an approach that the Administration backed in the Obama-Boehner negotiations last summer.

Switching to the chained CPI offers a rare opportunity to achieve deficit reduction on both sides of the budget in a manner that economic analysis supports. Policymakers should separately address high out-of-pocket health costs among the elderly and disabled, a problem that the CPI-E handles only clumsily. (If policymakers decided instead to use the CPI-E, they should devote the necessary resources to enable BLS to increase the sample size and make other necessary improvements. With roughly $1 trillion in benefit programs potentially affected by a switch to the CPI-E for indexing, it would be essential to develop a more rigorous index. Such an effort should also include developing a chained version of the CPI-E to address the same biases that afflict the standard unchained CPI.)

SSI Needs Special Attention

SSI merits special treatment under a switch to the chained CPI, for two reasons. First, as noted above, SSI beneficiaries are among the poorest elderly and disabled people in the country; SSI benefits lift single elderly and disabled beneficiaries only to about three-quarters of the poverty line. Second, in the absence of protective measures, the switch to the chained CPI would hit SSI beneficiaries much harder than Social Security beneficiaries — and would cancel out the effect of any Social Security benefit "bump" on those Social Security beneficiaries who are poor enough to qualify for SSI.

  • In Social Security and other retirement programs (such as federal pensions), the chained CPI would not affect recipients' initial benefit, which is based on their wages; it would simply cause their subsequent COLAs to be lower than they would otherwise be. If the move to the chained CPI occurred in 2015, a person who began to draw Social Security at age 62 in 2030 would initially get the same benefit as he or she would receive if the official CPI were still used to adjust benefits. After 20 years on the program, however, the individual's benefit would be nearly 6 percent lower than it otherwise would have been, the result of an average reduction of about 0.3 percent per year in the annual COLAs. [11]
  • By contrast, in SSI, the initial benefit level itself would be reduced by 0.3 percentage points every year. The person who began to receive SSI in 2030 (after the policy had been in effect for 15 years) would get an initial benefit over 4 percent lower than it would otherwise be — and after 20 years on the program, he or she would get a benefit 10 percent lower than it would otherwise be, due to the cumulative reduction in COLAs since 2015.
  • Furthermore, if no offsetting adjustments were made to SSI, then poor elderly or disabled people who received both Social Security and SSI would lose the 5 percent benefit "bump" in Social Security that various commissions have recommended in conjunction with a move to the chained CPI, because their SSI benefit would be reduced one dollar for every dollar of the Social Security benefit bump. [12] As a result, the only Social Security beneficiaries who would not get any relief from the benefit bump would be the poorest beneficiaries who need it most .
  • For some poor beneficiaries who receive both Social Security and SSI, the effects could be disastrous. If no protection is provided in SSI, then the Social Security benefit "bump" would lift them modestly over the SSI income limit, and because they were no longer eligible for SSI, they would lose their Medicaid coverage as well.

For these reasons, policymakers should not move to the chained CPI unless they exempt SSI or make appropriate adjustments in SSI, as outlined in the box on the next page, as well as in Social Security.

Policymakers Should Protect SSI Recipients From Impact of Chained CPI

Exempting SSI from a switch to the chained CPI would be a simple and effective way to protect its beneficiaries. If policymakers choose not to exempt SSI, they should soften the blow by: 1) addressing several serious deficiencies in SSI that relate to inflation-indexing; and 2) providing the same type of benefit "bump" for people on SSI as bipartisan commissions have recommended for people who have been on Social Security for many years. Specifically, if the chained CPI is applied to SSI, policymakers should:

  • Index SSI's $20 "disregard" of Social Security and similar income, so that the $20 figure is adjusted each year in accordance with the chained CPI. SSI reduces a beneficiary's SSI benefits when the individual receives income from Social Security or another "unearned income" source such as a retirement plan. The first $20 a month of such income is disregarded, but all the rest of the Social Security or other unearned income reduces the individual's SSI benefit on a dollar-for-dollar basis. This affects large numbers of SSI recipients who worked for low wages and receive a small Social Security benefit that leaves them well below the poverty line, which is why they qualify for SSI.

    The $20 disregard amount has never been adjusted for inflation since 1974, when SSI began. Had it been indexed, it would now be about $90. Policymakers should, at a minimum, index the $20 threshold to the chained CPI going forward. Better yet, they could increase its dollar level to make up a portion of the ground lost to inflation since 1974 and thereby enable these recipients to keep a more reasonable share of their Social Security income as a reward for a lifetime of work.

    This issue is important. The $20 disregard was established as a core part of the SSI benefit structure to ensure that poor people who have a substantial work history, and thus receive Social Security or a work-related pension, receive more in total retirement income than people who lack a comparable work history and receive only SSI. But the $20 amount has lost so much ground to inflation that it is worth less than a fourth what it was when SSI started, and the rewards for work have diminished substantially. Indexing the $20 disregard would offset a portion of the benefit loss from the chained CPI among SSI recipients who also receive Social Security. (It would not help SSI beneficiaries who lack outside income.)
  • Index the SSI asset limits. SSI has austere asset limits: $2,000 for individuals and $3,000 for couples. These limits have been frozen since 1989, and they, too, are far below where they stood in inflation-adjusted terms when SSI was created under President Nixon. These penurious limits prevent SSI recipients from budgeting for major one-time expenses like fixing a leaking roof or faulty heating system, and they are particularly outdated and problematic in light of the shift toward defined-contribution pension programs, which count in full against these asset limits. Poor elderly and disabled people with small 401(k) balances can be disqualified from SSI as a result. Moreover, the failure to index these limits for inflation causes them to erode further in real terms every year. The limits should be indexed to the chained CPI going forward.
  • Provide the same benefit bump as in Social Security to people who have been receiving SSI for 15 or 20 years. Even if we increase the unearned-income disregard as outlined above, SSI benefits would still be reduced by one dollar for every dollar that a poor Social Security beneficiary received over the disregard amount, including any proceeds from a Social Security benefit bump after 15 to 20 years on that program. Thus, most poor people who received both Social Security and SSI would get no relief at all from the Social Security bump. Furthermore, poor beneficiaries whose Social Security bump pushed them over the SSI income limit would lose Medicaid eligibility as well. It thus is essential that policymakers also provide a modest bump in the SSI benefit level after someone has received SSI for a number of years. This would be analogous to the Social Security bump and should be designed to be similar to, and compatible with, what policymakers provide in that program.

End Notes

[1] For historical reasons, benefit programs are indexed to the CPI for urban wage and clerical workers (CPI-W), while the tax code uses the CPI for all urban consumers (CPI-U). The CPI-W and CPI-U move in tandem, diverging only negligibly.

[2] There is no comprehensive list of programs and provisions that are linked to the CPI. We suggest that policymakers focus on COLAs and tax-code provisions.

[3] Informally known as the Boskin Commission. See

[4] 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.

[5] When relative prices change, consumers alter their purchases in response. Lower-level substitution refers to shifting their purchases between closely related items (such as different cuts of steak). Upper-level substitution refers to shifting between more loosely related items (such as chicken versus steak). The official CPI adjusts for lower-level substitution bias but not for upper-level substitution bias and hence slightly overstates inflation.

[6] Based on the menu compiled by the Office of the Chief Actuary, Social Security Administration, at Provisions included in this calculation are savings from the chained CPI (option A3) and costs from a benefit increase for long-term beneficiaries (options B6.3 or B6.4 for the Domenici-Rivlin or Bowles-Simpson variants, respectively).

[7] The CPI-E suffers from several limitations, notably a small sample size and sketchy coverage of factors like senior discounts. See Kenneth J. Stewart, "The experimental consumer price index for elderly Americans (CPI-E): 1982–2007," Monthly Labor Review, April 2008, pp. 19-24; Clark Burdick and Lynn Fisher, "Social Security Cost-of-Living Adjustments and the Consumer Price Index,"Social Security Bulletin, Vol. 67, No. 3 (2007), pp. 73-88; and National Academy of Sciences, At What Price? Conceptualizing and Measuring Cost-of-Living and Price Indexes (2002). Nevertheless, even if BLS received sufficient funding to address these shortcomings, we expect that the resulting index would tend to grow slightly faster than the standard CPI.

[8] See Bureau of Labor Statistics, "Experimental Consumer Price Index for Americans 62 Years of Age and Older, 1998-2009" ( for the current weights. Most notably, out-of-pocket medical expenditures make up about 5 percent of the CPI-W market basket (and 6.5 percent of the CPI-U), but 11 percent of the CPI-E. It is important to note that the CPI imputes housing costs for owner-occupants on a rental-equivalent basis; thus, the fact that elderly people have high home-ownership rates and are more likely than younger homeowners to be mortgage-free does not diminish the share of housing in their market basket in the various CPI measures.

[9] Based on the actuaries' menu, op. cit., option A6.

[10] Michael D. Hurd and Susann Rohwedder, "The Level and Risk of Out-of-Pocket Health Care Spending," Michigan Retirement Research Center Working Paper 2009-218, September 2009; Lynn Nonnemaker and Shelly-Ann Sinclair, "Medicare Beneficiaries' Out-of-Pocket Spending for Health Care," AARP Public Policy Institute, January 2011. Note that Hurd and Rohwedder believe that the Health and Retirement Study — used by Gopi Shah Goda, John B. Shoven, and Sita Nataraj Slavov in "How Well Are Social Security Recipients Protected From Inflation?" (National Bureau of Economic Research, Working Paper 16212, July 2010) — exaggerates the amounts of out-of-pocket spending for medical care.

[11] For simplicity, this assumes that the beneficiary files for early retirement at 62 (so that 20 years later he or she is 82). The Social Security actuaries depict effects on hypothetical workers who retire at age 65 (and depict the subsequent effects at ages 75, 85, and 95). In the actuaries' illustration, benefits are already lower at age 65 by 0.9 percent as the result of the move to a chained CPI (and 6.5 percent lower at age 85). That is because the switch affects COLAs starting at age 62, even if the beneficiary did not actually file for benefits until 65. See, Table 1B2.

[12] Nearly one-third of adult SSI recipients under age 65, and almost three-fifths of recipients over 65, also get Social Security based on their work history. See Center on Budget and Policy Priorities, Introduction to the Supplemental Security Income (SSI) Program, January 10, 2011 (