For a complete listing of all On the Margin articles, please visit: www.aei.org/onthemargin/.
Several tax parameters, including the standard deduction, the personal exemption, and the income levels at which brackets begin and end, are adjusted each year based on changes in the Consumer Price Index. Those adjustments ensure that inflation does not push taxpayers into higher tax brackets. Social Security benefits and other government benefit payments are similarly adjusted annually for inflation. In recent months, bipartisan support has grown for proposals to use a different measure of prices, the chained CPI, to make those adjustments. Because the chained CPI is expected to rise about 0.25 percent per year more slowly than the price indexes that are currently used, that switch would increase revenue and lower government benefit payments, thereby reducing the deficit and alleviating the long-term fiscal imbalance.
Many supporters of switching to the chained CPI have justified the switch on the grounds that the CPI is a more accurate measure of inflation. They describe the switch as a technical adjustment that should be made even if it does not help address the U.S. fiscal imbalance. As I explain in this article, that technocratic argument is incomplete. If there was no fiscal imbalance, we would not want to merely adjust tax brackets for inflation. Without a fiscal imbalance, we would want to take the more sweeping step of adjusting tax brackets based on nominal income, so that neither real economic growth nor inflation would push taxpayers into higher tax brackets. In that case, we would have no need for a more accurate measure of inflation. The only sound argument for indexing to inflation rather than to nominal income is the need for additional revenue to address the long-term fiscal imbalance. Proposals to refine inflation indexation therefore cannot be discussed apart from broader budgetary issues.
The real case for switching to the chained CPI is grounded in fundamental budget realities. The fiscal imbalance will ultimately have to be addressed by bipartisan agreements that restrain entitlement spending and increase revenue. It will not be possible to address the imbalance on the spending side of the budget alone or on the revenue side alone, nor will it be possible for either major political party to unilaterally tackle the problem in a durable manner. A switch to the chained CPI is attractive because it combines revenue increases and entitlement cuts in a way that has attracted bipartisan support.
Inflation Indexation Under Current Law
Inflation indexation was introduced to the code by legislation adopted in 1981, effective in 1985.1 Although the scope of indexation has been extended over the last three decades, its mechanics have remained essentially unchanged. Section 1(f)(3) bases the indexation on the CPI, and section 1(f)(4) clarifies that this acronym refers to the Consumer Price Index, which section 1(f)(5) defines to be the "Consumer Price Index for all-urban consumers published by the Department of Labor" (CPI-U). Section 1(f)(3) provides that the dollar amounts for each year shall be based on the "CPI for the preceding calendar year," which is defined by section 1(f)(4) to be the "average of the Consumer Price Index as of the close of the 12-month period ending on August 31 of such calendar year." So, the dollar amounts applicable in calendar year 2011 are based on the average CPI-U values for September 2009 to August 2010. The dollar amounts are effectively indexed with a 16-month lag, which allows the adjusted values to be determined well before the applicable year begins.
Late each year, the IRS issues a revenue procedure that lists the code provisions for which Congress has provided inflation indexation and sets forth the adjusted values for the upcoming year.2 Inflation indexation applies to several key tax parameters. The standard deduction is indexed under section 63(c)(4), the personal exemption is indexed under section 151(d)(4), the income levels at which each tax bracket begins are indexed under section 1(f)(2), and various features of the earned income tax credit are indexed under section 32(j). Indexation also applies to the income ranges at which various provisions begin to phase out and to limits on the values of some tax breaks. For simplicity, I refer to the indexation of "tax brackets" throughout the remainder of this article, but that term is intended to encompass the full range of indexation described above.
Some important provisions are not indexed, including the $1,000 value of the child tax credit prescribed by section 24(a), the income levels prescribed by section 86(c) at which a portion of Social Security benefits becomes taxable, and the income levels prescribed by section 1411(c) at which the 3.8 percent unearned income Medicare contribution tax on interest, dividends, and capital gains will start to apply, when the tax takes effect in 2013. The permanent values of the alternative minimum tax exemption, prescribed by section 55(d)(1), are also not indexed, but those values have been overridden by higher temporary values in each year from 2001 through 2011.
CPI indexation is not confined to the tax code. Under 42 U.S.C. section 415(i), Social Security old-age, survivor, and disability benefits are indexed to inflation for each year that a person continues to receive benefits. The indexation is not based on the CPI-U, but instead on a different index: the Consumer Price Index for Urban Wage Earners and Clerical Workers (the CPI-W). The benefits paid each year are proportional to the average value of the CPI-W in July through September of the preceding year.
Supplemental Security Income (SSI) benefits paid to low-income elderly and disabled households are also indexed to the CPI-W under 42 U.S.C. section 1382f(a), as are veterans' benefits under 38 U.S.C. section 5312. Federal civilian and military pensions are also indexed to the CPI-W, although not all those pensions receive complete protection from inflation. The federal poverty guidelines, which help determine benefits under a variety of federal programs, are indexed to the CPI-U under 42 U.S.C. section 9902(2).3
The Bureau of Labor Statistics (BLS) in the Department of Labor publishes both the CPI-U and the CPI-W. The CPI-U is intended to measure the cost of goods and services purchased by the 87 percent of the American population who live in urban or metropolitan areas, while the CPI-W is intended to measure the cost of goods and services purchased by a subset of that group, consisting of specified workers and comprising about 32 percent of the American population.4 The two price indexes usually move together closely, a pattern that is expected to continue. From December 1999 to August 2011, according to seasonally unadjusted data, the CPI-U rose by 34.61 percent, or 2.58 percent per year compounded annually, while the CPI-W rose by 35.27 percent, or 2.62 percent per year.
Although it is not entirely clear why the CPI-U is used to index the tax code and the poverty guidelines while the CPI-W is used to index Social Security and other programs, the issue is relatively unimportant, given the similar behavior of the two indexes. Recent proposals pose the more important question whether to replace both of these indexes with the chained CPI, an index that behaves quite differently from the CPI-U and the CPI-W.
Proposed Switch to the Chained CPI
Price indexes ultimately aspire to measure the increase in consumers' cost of living. That task is difficult because the prices of various goods and services rise at different rates. A price index should combine the price changes for the various goods and services in a way that properly measures their overall impact on consumer well-being. The chained CPI is an alternative version of the CPI-U that better accounts for how consumers respond to price changes.
When prices rise at different rates, consumers are likely to consume fewer of the items with larger price increases and more of the items with smaller price increases. For example, if the price of oranges rises by more than the price of apples, consumers are likely to buy fewer oranges and more apples. Consumers' ability to shift toward apples mitigates, although it does not eliminate, the harmful effect of the price increase for oranges. The CPI-U and the CPI-W do not, however, fully account for consumers' ability to shift between products.5
To be sure, the two price indexes do not completely ignore consumers' response to relative price changes. When computing the indexes, the BLS assumes that consumers switch between items within each of the 211 major categories of goods and services tracked by the price indexes. The computations do not, however, account for the ability of consumers to switch between those categories. For example, consumers are assumed to switch from one variety of apple to another variety if their prices rise at different rates, but they are assumed not to switch between apples and oranges. The BLS makes a simple assumption about the degree of within-category substitution, which allows it to account for that substitution in the initial version of each month's CPI when it is published in the following month.
In 2002 the BLS began publishing the Chained Consumer Price Index for All Urban Consumers, which fully accounts for substitution across the various categories, providing data back to December 1999. That chained CPI measures consumer substitution by using data from the Consumer Expenditure Survey on the actual mix of consumer purchases. Because of lags in the availability of those survey data, the initial release of each month's chained CPI is only an estimate. The initial estimate is replaced with an interim estimate in February of the following year and with the final value in February of the next year. In contrast, the final values of the CPI-U and CPI-W are published at the outset and are not later revised.6
From December 1999 to August 2011, according to the currently available data, the chained CPI rose by 30.26 percent, or 2.29 percent per year compounded annually. Over that interval, therefore, the chained CPI rose 0.29 percent per year more slowly than the CPI-U and 0.33 percent per year more slowly than the CPI-W. The slower growth of the chained CPI reflects the economic impact of consumers' ability to substitute between categories of goods and services in response to changes in relative prices.
The tables below provide illustrative information about the potential effect on the inflation adjustments for taxes and Social Security. Table 1 shows the actual inflation adjustments made to tax brackets for 2003 through 2012 and the adjustments that would have occurred if they had been computed using the currently available values of the chained CPI. It should be noted that chained CPI adjustments computed in real time would have differed from those shown in the table, because they would have been based on the estimated values of the chained CPI that were then available rather than the revised estimates that are now available. The mean difference for those 10 inflation adjustments is 0.25 percentage points.
Table 2 provides similar information for the annual Social Security cost of living adjustments (COLAs) for 2002 through 2011, again using the currently available values rather than those that would have been available in real time. The mean difference for those 10 inflation adjustments is 0.27 percentage points. The difference for the 2012 COLA is likely to be much larger, possibly around 0.7 percentage points.
By indexing tax brackets more slowly, a switch to the chained CPI would increase revenue. By reducing the COLA for federal benefit programs, the switch would also reduce benefit outlays. Both the revenue increase and benefit reduction would lower the deficit. In its most recent survey of deficit reduction options, the Congressional Budget Office presented budgetary estimates for a switch to the chained CPI and summarized the policy arguments for and against that change. Based on its expectation that the chained CPI will increase 0.25 percent per year more slowly than the CPI-U and the CPI-W in upcoming years, the CBO estimated that a switch would increase income tax revenue by $72 billion, reduce Social Security benefits by $112 billion, and reduce federal pensions and veterans' benefits by $24 billion from fiscal 2012 through 2021.7
The Committee for a Responsible Federal Budget has documented the extensive bipartisan support for proposals to switch to the chained CPI.8 Those proposals have appeared in bipartisan plans to address the deficit and have also been endorsed by individuals and organizations of varying ideological perspectives. Except when noted, the proposals have called for the switch to apply to both taxes and benefits.
In its November 2010 deficit reduction plan, the Bipartisan Policy Center proposed to switch to the chained CPI starting in 2012.9 The deficit reduction plan approved by a majority of the Bowles-Simpson commission in December 2010 called for a switch to the chained CPI for taxes and benefit programs, starting in 2012.10 A switch to the chained CPI was also proposed by the "Gang of Six," consisting of Sens. Kent Conrad, D-N.D., Richard J. Durbin, D-Ill., Mark R. Warner, D-Va., Saxby Chambliss, R-Ga., Tom Coburn, R-Okla., and Mike Crapo, R-Idaho. Their July 2011 deficit reduction plan called for the switch to start in 2012, except that SSI benefits would be fully exempt for the first five years and partially exempt for the next five years.11
The income tax reform bill introduced by Sen. Ron Wyden, D-Ore., and then-Sen. Judd Gregg on February 23, 2010, called for a switch to the chained CPI for tax bracket indexation, starting in 2013.12 The successor bill introduced by Wyden and Sen. Daniel Coats, R-Ind., on April 5, 2011, calls for a similar switch, starting in 2014.13 Because those bills are concerned with the income tax system, they do not include any change to the indexation of Social Security or other benefit programs.
A switch to the chained CPI has been endorsed by The Washington Post,14 Reihan Salam at National Review,15 Shannon Leon of the Progressive Policy Institute,16 Hoover Institution research fellow Charles Blahous,17 Jared Bernstein of the Center on Budget and Policy Priorities,18 and Donald Marron of the Urban-Brookings Tax Policy Center.19 The Wall Street Journal praised the chained CPI provision in the Gang of Six plan.20 The deficit reduction plan presented by the Center for American Progress in May 2011 proposed a switch to "a more accurate measure of inflation" for tax and benefit programs.21
That switch was actively considered when President Obama and congressional leaders discussed a possible grand bargain to reduce the fiscal imbalance earlier this year. Tax Notes reported at the time that the proposal was "a rare area of common ground for Republicans and Democrats."22 Unfortunately, no grand bargain resulted from those discussions.
The Fiscal Imbalance and the Chained CPI
There is little dispute among economists that the chained CPI provides a more accurate measurement than the CPI-U or CPI-W of the inflation affecting the general American public. Many supporters of switching to the chained CPI rest their case on its accuracy, arguing that the switch can be justified without reference to its role in reducing the fiscal imbalance.
For example, in a prominent paper on this topic, Adam Rosenberg and Marc Goldwein of the Moment of Truth Project say:
An overwhelming majority of economists from both parties agree that the chained CPI is a far more accurate measure of inflation than the CPI measurements currently in use. In addition to improving technical accuracy, switching to the chained CPI would have the secondary benefit of reducing the deficit.23 [Emphasis added.]
That argument clearly puts technical accuracy first and budgetary impact second. The Wall Street Journal similarly states in an editorial that a switch to the chained CPI is "justified as more accurate financial measurement and should not count as a violation of any antitax pledge."24
Although that technocratic case may seem appealing, it skips a crucial step in the argument. Of course, if tax brackets should be adjusted for inflation, the adjustment should be made with an accurate measure of inflation. But the first question is whether tax brackets should be indexed to inflation. Or, should they be adjusted based on some other economic variable or not adjusted at all? Consideration of those questions quickly reveals that the case for the chained CPI must be grounded in budgetary realities, not the intricacies of price index construction.
Inflation adjustment ensures that rising prices do not push taxpayers into higher brackets. That is sensible because the rate at which Americans are taxed should not depend on the inflation rate. Were it not for the fiscal imbalance, however, the inflation adjustment would be only our first step. We would actually want to adjust tax brackets for nominal income growth, which is the sum of inflation and real income growth.
Nominal income indexation would ensure that neither inflation nor real income growth pushed taxpayers into higher brackets. Taxpayers whose incomes kept pace with general economic growth would pay the same percentage of their incomes in tax from year to year. Of course, taxpayers whose incomes rose more rapidly than the general economy would move into higher brackets and those whose incomes rose more slowly than the general economy would move into lower brackets, based on the progressivity that Congress has set in the rate schedule. Taxpayers would pay higher tax rates if their relative incomes rose, but they would not pay higher tax rates merely because incomes rose throughout the country as a whole.
In the absence of the fiscal imbalance, nominal income indexation would yield the right outcome. After all, there is no reason real income growth per se should result in higher tax rates any more than inflation should result in higher tax rates. Our graduated income tax rate schedule represents a judgment that in any given year, a taxpayer with double the income of another taxpayer should pay more than double the tax. But that judgment does not imply that the country's aggregate real tax payments should more than double when its aggregate real income doubles because of general economic growth. In the absence of the fiscal imbalance, it would be appropriate for the country's real tax payments to double as real income doubled, as would occur under nominal income indexation.
From a logical perspective, the question of how the country's tax payments should change as the country becomes richer is separate from the question of how the tax burden should be divided between the rich and the poor in any given year. Nominal income indexation would allow those questions to be kept separate.
Without the fiscal imbalance, therefore, a switch to the chained CPI would be a move in the wrong direction. If the CPI-U overstates true inflation by 0.25 percent per year, using it to index the tax brackets causes the correctly measured real values of the bracket endpoints to rise by 0.25 percent per year rather than remaining constant. Without the fiscal imbalance, however, that would be all to the good. Under those circumstances, it would be appropriate for the real values to rise by 2 percent per year or so, in line with real economic growth. An increase of 0.25 percent per year would be too small, not too large, and we would not want to eliminate that increase by switching to the chained CPI. Chris Edwards of the Cato Institute has recently relied on that basic argument to oppose a switch.25
The argument does not apply to the United States, however, because we face a severe fiscal imbalance. Although the dire fiscal outlook is well known, it warrants a brief review. The description below is drawn from the CBO's alternative fiscal scenario in its June 2011 long-term budget outlook.26 Under that scenario, Social Security spending grows from 4.8 percent of GDP in 2011 to 6.1 percent in 2035 because of population aging. Medicare spending, gross of premiums, grows from 3.7 to 6.7 percent of GDP over that period in response to rising medical costs. Spending on Medicaid and related healthcare programs grows from 1.9 to 3.7 percent of GDP, reflecting rising medical costs as well as provisions of the March 2010 healthcare reform law that expand Medicaid and create new subsidies to help some households buy private health insurance. In total, then, those entitlement programs grow from 10.4 percent of GDP in 2011 to 16.5 percent in 2035. That shift of more than 6 percent of GDP into those programs over that 25-year period is a staggering reallocation of economic resources.
The expansion of those programs helps push total non-interest spending up to 21.5 percent of GDP in 2021 and 25.0 percent in 2035. Those values are far higher than the average 18.3 percent level that prevailed in 1971 through 2007. To be sure, the future spending levels seem somewhat less dramatic if they are compared with the 22.7 percent share reported for 2011, but the latter value is temporarily high because of the Great Recession and the policy measures that were adopted to combat it.
The alternative fiscal scenario assumes that revenue is stable at 18.4 percent of GDP from 2021 onward. That level is only slightly higher than the 18.2 percent average share observed for 1971 through 2007, although it is much higher than the temporarily low 14.8 percent share in place for 2011. At the risk of stating the obvious, that revenue level is incompatible with the spending trajectory outlined above. Under the alternative fiscal scenario, federal debt held by the public therefore soars to 101 percent of annual GDP in 2021 and to 187 percent in 2035. Those debt levels are vastly higher than the 36 percent average value observed in 1971 through 2007 or even the temporarily high 69 percent value estimated for 2011.
Simple mathematics requires that revenue rise as a share of GDP or that entitlement spending grow more slowly than its current-policy trajectory or both. Of course, mathematics cannot determine the choice between those policies. That choice will be made by the American people's elected representatives through the democratic political process.
In various writings over the last three years, I have emphasized four long-run fiscal realities that I believe will arise from the political process.27 One reality is that entitlement spending will be curtailed to some extent, relative to the current-policy path. That entitlement restraint must be adopted because remaining on the current-policy path would reduce national saving and require astronomically high marginal tax rates in future decades, impeding long-run economic growth. Another reality, however, is that revenue will rise above its historic share of GDP because the extent of entitlement restraint that the American people are prepared to support will not be sufficient to avert an increase in the revenue share.28
The limited willingness of the public to support entitlement restraint has been documented in several polls. In my earlier articles, I cited a poll by Alan Blinder and Alan Krueger that asked Americans how they would prefer to reduce the long-run Social Security deficit. Only 5 percent of Americans favored relying primarily on benefit cuts, while 30 percent favored relying primarily on payroll tax increases and 34 percent favored a mix of the two measures.29 In a survey in Tax Notes last year, my colleague Karlyn Bowman described results from several polls that asked Americans how they would prefer to reduce the budget deficit. In a March 2010 Quinnipiac poll, for example, only 19 percent of the public said that cutting the growth of Social Security benefits should be part of a government approach to the deficit, and only 21 percent gave that response for cutting the growth of Medicare benefits. Many other poll results discussed in Bowman's article reveal a similar lack of public enthusiasm for entitlement restraint.30
Conservatives who advocate limited government should work to ensure that entitlement restraint is a major part of the solution to the fiscal imbalance. They should seek to shape public opinion by pointing out the advantages that entitlement reductions have over tax increases in terms of economic growth and liberty. But it will not be politically possible to close the massive fiscal gap solely on the spending side of the ledger. As I have recently emphasized, the corollary is that progress on the fiscal gap will arise from bipartisan agreement rather than from unilateral action by either party. Of course, any agreement will require compromises by all involved.31
The reality that revenue must rise as a share of GDP undermines the case for nominal income indexation. Indexing tax brackets only to inflation and thereby allowing real economic growth to push taxpayers into higher brackets is a simple and effective way to help raise the revenue required to finance entitlement growth. To be sure, it would be more elegant and conceptually pure to first increase the tax bracket endpoints in line with nominal income growth and then make a separate adjustment lowering them based on some function of medical costs and population aging. After all, it is the latter factors, not real income growth, that produce the need for a rising revenue share. Nevertheless, indexing tax brackets to inflation rather than nominal income growth offers a rough and ready way to respond to the fiscal imbalance.
Because indexation of tax brackets and benefit payments is described as being based on inflation, using a more accurate measure of inflation to perform the indexation promotes transparency by bringing the actual adjustment more in line with its description. The real case for switching to the chained CPI, however, is grounded in fundamental budget realities. A switch to the chained CPI combines revenue increases and entitlement cuts (with the latter roughly twice as large as the former over the next decade) in a way that has attracted bipartisan support. Those properties make it a promising candidate for an agreement to reduce the deficit.
A potential vehicle for an agreement now exists. The Budget Control Act of 2011 created a Joint Select Committee on Deficit Reduction, informally dubbed the supercommittee, which is instructed to recommend $1.5 trillion of deficit reduction by November 23. Any plan recommended by a majority of that 12-member bipartisan bicameral committee will receive an up-or-down vote on the House and Senate floors, with limited debate and no amendments. If the supercommittee does not make a proposal or if its proposal is not enacted into law, automatic spending cuts affecting defense, non-defense discretionary spending, Medicare provider reimbursements, and some small entitlements will take effect in fiscal 2013 through 2021.32 The prospect of those cuts is intended to provide an incentive to reach an agreement. Given the partisan gridlock that now prevails in Washington, the supercommittee faces long odds. If it does reach agreement within the allotted time, however, it will need to adopt measures that have already been proposed and scrutinized and that have secured support in both parties. A switch to the chained CPI satisfies those criteria. Of course, that switch will provide only a modest portion of the deficit reduction that is needed, so it should be accompanied by other spending reductions and tax increases.
The technocratic case for a switch to the chained CPI is incomplete because the fact that this index offers a more accurate measure of inflation does not alone justify a switch. The real case for switching to the chained CPI is that it offers an opportunity for bipartisan agreement to reduce entitlement spending and raise revenue, reducing the long-run fiscal imbalance and enhancing America's future prosperity.
Alan D. Viard is a resident scholar at AEI