Measuring Up: The Case for the Chained CPI
This report was originally published 05/11/2011 and has been updated on 12/12/2012.
With so much of government set to autopilot, the measurement we use for inflation plays a very important role in public policy. Currently, the federal government generally relies on the consumer price index (CPI) to index provisions of the budget and tax code to account for cost-of-living changes. However, this measure actually overstates inflation and, as a result, drives up the deficit unnecessarily.
Maintaining purchasing power in spending programs and indexing various parts of the tax code is an important policy goal. However, policymakers should ensure that the most accurate measure of inflation is being used.
To correct the problem of over-indexation, many have proposed switching to the chained CPI to provide a more accurate measure of inflation for indexed provisions in the federal budget. This switch was recommended by the National Commission on Fiscal Responsibility and Reform (“Fiscal Commission”) and the Bipartisan Policy Center’s Debt Reduction Task Force (“Domenici-Rivlin”). It was also a key component in negotiations on the debt ceiling last summer and the Super Committee last fall, and has a number of strong advocates from the left, right, and center (see http://crfb.org/blogs/bipartisan-support-chained-cpi). 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 chained CPI would have the secondary benefit of reducing the deficit – by over $235 billion over the next decade alone if implemented for 2014, and over $290 billion if implemented immediately.
Addressing our fiscal challenges will require many tough choices and policy changes – but switching to the chained CPI represents neither. Such a change offers policymakers the rare opportunity to achieve significant savings spread across the entire budget by making a technical improvement to existing policies. As such, across-the-board adoption of the chained CPI should be at the top of the list for any deficit reduction plan or down payment.
Currently, the federal government relies on two cost-of-living measures for most inflation-indexed spending programs and provisions of the tax code: the CPI-U (Consumer Price Index for All Urban Consumers) and the CPI-W (Consumer Price Index for Urban Wage Earners and Clerical Workers).
The CPI-W is used to calculate COLAs for Social Security and other federal retirement programs, while the CPI-U is used to index various provisions of the tax code, as well as poverty thresholds. The CPI-U is considered to be the more accurate of the two, since it covers 87 percent of the population, compared to the 32 percent covered by the CPI-W.
The Bureau of Labor Statistics (BLS) formulates these price indices by creating a representative “market basket” of goods based on the covered population’s consumer spending patterns; in the case of the traditional CPIs, this data is available with a two-year time lag. Then, using this market basket, BLS tracks price changes for the included items over time. BLS follows the prices of goods in the market basket, and measures inflation based on the overall change in the price of the basket.
In 1996, the Advisory Committee to Study the Consumer Price Index (the Boskin Commission) found that both CPI-U and CPI-W overstated inflation in a number of ways, which they estimated to total about 1.1 percentage points per year. The Bureau of Labor Statistics responded by making a number of changes to the way they measured CPI-U and CPI-W, which corrected much of the biases in the two indices.
However, a portion of the bias – upper level substitution bias for changes between categories – cannot be addressed through the existing CPI-U and CPI-W for technical reasons. Instead, BLS created a new measure of inflation – the chained CPI (also known as the superlative CPI or the C-CPI-U) – in 2002 to account for consumer substitution between categories. This measure has been refined and improved since it was initially published. Unlike the methodological changes in the calculation of CPI-U and CPI-W that are automatically reflected in the published measures used for indexing programs under current law, using the more accurate chained CPI for indexation instead of the CPI-U or CPI-W requires a statutory change in law.
The Technical Case for the Chained CPI
One of the reasons the Boskin Commission found that the CPIs overstated price changes was something called substitution bias.
Understanding substitution bias requires recognizing the way consumers respond to changes in prices. If one spends $100 per month on apples and the price of apples suddenly doubles, it is unlikely that their cost of living will go up by $100. Rather, an individual will substitute away from apples, buying less of them than they otherwise would and therefore buying more of something else instead.
Substitution bias occurs because the regular measure of CPI assumes consumers will buy the same basket of goods regardless of relative prices, not realizing that consumers can often soften the blow of increased relative prices by consuming more of a relatively cheaper good. Since the Boskin Commission, BLS has made a number of improvements in the calculation of CPI-U and CPI-W to account for what is known as “lower level substitution bias,” when individuals substitute within categories. In particular, BLS adopted a geometric mean in 1999 to account for consumer substitution within item categories as relative prices change. The geometric mean formula, though, does not account for consumer substitution taking place between categories – known as upper level substitution.
For example, if consumers respond to the price increase for Granny Smith apples by buying more Red Delicious apples instead (lower level substitution bias – changes within categories), this is accounted for in the current CPI.
But if consumers respond to the price of Granny Smith apples increasing by buying less apples altogether and purchasing more oranges instead (upper level substitution bias – changes between categories), this is not accounted for.
Upper level substitution bias is an artifact of BLS’s reliance on a fixed “market basket” of goods, which is based on old purchasing habits. This could be fixed by instead using a market basket based on the very newest purchasing habits, except that this would actually cause the opposite problem where substitution biases cause the CPI to understate inflation.
Moving to the chained CPI would address this bias by using a superlative index that updates expenditure weights and formulas in order to address consumer response to substitution between categories.
As the Congressional Budget Office (CBO) explains, the chained CPI “attempts to fully account for the effects of economic substitution on changes in the cost of living… [It] provides an unbiased estimate of changes in the cost ofliving from one month to the next by using market baskets from both months, thus ‘chaining’ the two months together.”
Since 2000, the chained CPI has, on average, been 0.25 to 0.3 percentage points lower per year than the standard CPI measures. Though this difference is small on average, it compounds over time; depending on which index you use, prices have either increased by 34 percent (CPI-U and CPI-W) or 29 percent (chained CPI) between 2011 and 2000 (see Figure 1). Over a longer time frame, this difference would become even more pronounced.
Budgetary Impact of the Chained CPI
Because a number of government programs and provisions in the tax code are linked to the CPI-U or CPI-W, the reliance on an overstated measure of inflation results in higher spending and lower revenue collections than intended or warranted. Relative to current law, which essentially projects continued overpayments and under collection, switching over to the chained CPI would result in substantial deficit reduction.
Instituting the chained CPI for Social Security cost-of-living adjustments (COLAs) alone would save $100 billion through 2022 if implemented in 2014. Using it for COLAs in other federal retirement programs would save another $27 billion over the same period, and there would be an additional $22 billion in deficit reduction from other areas of the budget. On the tax side, meanwhile, moving to chained CPI would cause tax bracket thresholds and other parameters to grow more slowly and raise an extra $62 billion over the ten-year period relative to current law.
Added together and combined with the net interest savings, we estimate that relying on the more accurate measure of inflation beginning in 2014 would reduce deficits by about $236 billion over the next decade and $54 billion in 2022 alone. Implementing the policy effective for 2013 would save closer to $290 billion over ten years. Actual savings could be larger, since these numbers use a more conservative differential (0.25%) than has been used previously (0.3%), and since our numbers assume the chained CPI is not applied to CPI “sub-indices”.
Though this would reduce the deficit relative to current law, it should not be considered a change in tax or spending policy. The provisions affected by the move to chained CPI are designed to be indexed to changes in overall cost of living. Rather than serving to raise taxes and cut benefits, switching to the chained CPI would simply be fulfilling the mission of properly adjusting for cost of living. As Robert Greenstein of the Center for Budget and Policy Priorities has said, “this change should not be regarded as a benefit cut or a tax increase. It should be regarded more as a technical change to achieve Congress’s stated goal of keeping pace with inflation in as accurate a way as possible. ”
In addition to this $236 billion in deficit reduction, there are numerous budgetary merits to using the chained CPI. First, it would contribute to reducing the long-term funding shortfall in Social Security. Switching to the chained CPI for COLAs would close more than one fifth of Social Security’s 75- year shortfall. This would be a significant down payment on bringing that program into long-term balance.
Another merit of the switch to the chained CPI is that the savings are credible but back-loaded. The budgetary savings from a 0.25 percent change in indexing is very tiny at first – in fact, switching to chained CPI only saves about $3 billion in the first year and $8 billion in the second. Over time, though, these savings compound, leading to increasing amounts of annual deficit reduction. Assuming implementation in 2014, the policy would save over $50 billion (over $60 billion including interest) in 2022 and more in subsequent years. The gradual ramping up of savings fits in well with the idea that deficit reduction should be enacted immediately but phased in gradually so as not to undermine the economic recovery and to give taxpayers and beneficiaries time to adjust to the small changes.
One final merit of using the chained CPI is its political appeal. Although switching to the chained CPI is a technical change that should not be viewed as a spending cut or tax increase, it is worth noting that the effects of switching to chained CPI would be quite balanced, with a portion of savings coming from Social Security, other spending programs, and the tax code.
Using a better measure of inflation would also give us a more accurate understanding of changes in real variables in the economy. As the Boskin Commission report stated, “Even if no federal program on either the outlay or revenue side of the budget were indexed, it would still be desirable to improve the quality of measures of the cost of living from the standpoint of providing citizens a better and more accurate estimate of what was actually going on in the economy.” But the magnitude of our fiscal challenges and the substantial fiscal impact of achieving this correction make this change particularly timely.
Distributional Impact of the Chained CPI
One commonly-voiced criticism of switching to the chained CPI is the claim that it is regressive, both on the tax and on the spending side. This claim is a false one, based on a misunderstanding of how the chained CPI works and a misreading of existing non-partisan analysis of the change. With regards to both Social Security and the tax code, switching to chained CPI would be roughly distributionally neutral – meaning that individuals of different income levels are affected similarly.
On Social Security, the effects of chained CPI would be more or less uniform across individuals. Because the cost-of-living-adjustment offers the same percent increase for all beneficiaries, the effect of switching to a more accurate but modest measure of inflation would be also roughly the same for everyone. Of course, in reality there would be some differences due to a number of factors, including how long individuals receive benefits from the program. As it turns out, however, these deviations skew slightly progressive rather than regressive. According to the Social Security Administration, switching to the chained CPI would reduce average benefits in 2050 by about three percent relative to the current CPI for those in the bottom three quintiles, and four percent for those in the top two quintiles.
On the revenue side, the average effects are similar – though there would likely be more deviation within each income group. According to the Tax Policy Center (TPC), each quintile would see a 0.2 percent reduction in their after-tax income over a decade. Because so many provisions in the tax code are indexed to inflation – including the standard deduction, personal exemptions, income levels where new tax brackets begin, limits on various tax expenditures, etc – few taxpayers would be unaffected by the change. Among those who would not see any change in their taxes, most are at the bottom of the income spectrum and currently face no (or a negative) tax burden. There is also a much smaller but not insignificant group at the very top of the income spectrum who pay the AMT and are therefore unaffected. Still, more than 40 percent of the additional revenues generated by the chained CPI would come from the top quintile alone.
While switching to the chained CPI would be distributional neutral, this distribution may not be desirable in the context of aiming to protect the most vulnerable from the adverse effects of deficit reduction. When thinking about distribution, however, it is important to keep in mind the total effect of the package rather than the effect of a single policy. As an example, if Social Security reform were to rely on chained CPI but also raise the taxable maximum and flatten the benefit formula, as the Fiscal Commission plan would do, the net outcome would be highly progressive. Moreover, to the extent there are specific undesirable effects of chained CPI on certain vulnerable populations, they should be addressed through small policy changes targeted to those populations. For example, the Fiscal Commission recommended instituting a flat dollar benefit bump-up for very old Social Security recipients (as well as a minimum benefit for lower income beneficiaries). Other policies such as enhancing SSI benefits or increasing the refundability of certain credits on the tax side could be included in a package to offset the impact of the change on lower income individuals.
Chained CPI is a more accurate measure of inflation and there is no reason to maintain a $450 tax windfall for those in the top quintile as a result of using an inaccurate measure of inflation in the tax code just to prevent a $25 tax increase for those in the bottom quintile. Likewise, there is little reason to provide higher than warranted increases in benefits for all Social Security beneficiaries just to protect lower-income beneficiaries when those concerns could be addressed by much more targeted policies and at lower overall costs.
The purpose of indexing government benefits and provisions in the tax code is to ensure that they are adjusted to reflect changes in cost of living. However, current law relies on measures on inflation that fail to fulfill this purpose accurately. The Bureau of Labor Statistics has said that the chained CPI provides a closer approximation to a cost of living measure than other CPI measures, a judgment shared by economists and statisticians across political and ideological spectra. Adopting the chained CPI for indexation therefore represents a more accurate and effective approach to achieving the desired policy goal of accurately preserving the real value of certain spending programs and tax provisions.
The magnitude of our fiscal challenges will require serious tax and spending changes that go well beyond technical issues. The Fiscal Commission offered $4 trillion of such savings, which should be enacted as soon as possible. But if policy makers cannot even agree to a technical improvement to rely on a more accurate measure of inflation, the prospects for making the hard choices seem grim.