View all Articles
Commentary By Charles Blahous

Reforming CPI: Not a "Grand Bargain" but a Prudent Reform

Economics, Economics Tax & Budget, Finance

Recent reports indicate that the budget/debt negotiations will not produce a “grand bargain.” At best they will produce a smaller set of targeted reforms slightly improving but not correcting the unsustainable trajectory of federal finances. But whether the budget discussions produce a big deal or a small one, however, both sides would do well to implement a more accurate measure of economy-wide inflation, namely the “chained” C-CPI-U.

Basic background: Many aspects of federal law from income tax brackets to Social Security payments are indexed to grow each year with price inflation, more specifically with the Consumer Price Index (CPI). There are different versions of CPI now in use, including the CPI-U (measuring inflation facing all urban consumers) and the CPI-W (measuring inflation facing urban workers). Some programs use one of these and some the other, but generally the two are close in value anyway.

Over the years, many economists have noted that these measures tend to overstate actual price inflation as felt by consumers. Simplifying considerably, this is because the rising price of one item often causes consumers to buy a different item instead, one whose price hasn’t risen as much. The mix of items that consumers buy thus changes over time, with the increase in the total cost of living being less than if no purchase substitutions had occurred.

Over the years the Bureau of Labor Statistics (BLS), which calculates these various inflation measures, has implemented improvements to correct for these changes in buying patterns. The current CPI-U and CPI-W, however, do not adequately account for changes across purchasing categories. That is to say, consumers don’t limit their purchasing substitutions merely to other items within the same spending category; they also shift their purchasing preferences between categories according to inflation trends within each. To address this, the BLS developed another index known as the superlative or chained CPI (C-CPI-U), which accounts for cross-category substitutions.

This C-CPI-U has averaged something close to 0.3 percentage points per year less than CPI-U or CPI-W in the years since 2000. Advocates of its adoption across federal programs argue that not only would C-CPI-U more faithfully reflect inflation than measures now in use, but would substantially reduce federal deficits as well due to its effects on outcomes ranging from income tax bracket growth to Social Security COLAs.

Press articles recently reported that the Obama Administration suggested the adoption of chained CPI in the ongoing budget discussions. Unfortunately, the reform was described in the worst possible way as the Administration having proposed “Social Security cuts” rather than merely the next technical improvement in the implementation of current policies. This led to an immediate denunciation of the idea by Congressional Democrats, considerably lessening its chances of being adopted.

This is highly unfortunate, as CPI’s refinement is a reform whose time ought to have come. It would improve the accuracy of federal processes, improve the budget outlook and serve the interests of negotiators on both sides of the table. It is strongly to be hoped that the option can be kept alive. Among the reasons:

C-CPI-U is the most accurate available estimate of economy-wide inflation. Some federal policies (like the fixed income thresholds for the recently-enacted 0.9% Medicare surtax) aren’t indexed at all. Others (like Social Security’s benefit formula) are indexed to wage growth. But currently expressed policy in many other areas of the federal budget is to index for general price inflation, no more and no less. To use the best available measure of such inflation is therefore not a “benefit cut” or a “tax increase” as much as it is the most faithful available method of complying with the policy basis of various statutes.

CPI-U and CPI-W weren’t originally inserted into existing laws because their sponsors thought that they overstated inflation; they were inserted because the sponsors were attempting to capture inflation, and those metrics were the best available at the time. To now use the more recently-developed C-CPI-U is in effect to better conform these various aspects of federal law to Congressional policy intent.

The purpose of CPI-indexation is not to attain targeted benefit or tax levels. Many on the left oppose using C-CPI-U because the continued use of CPI-W would lead to higher Social Security benefits, especially among the oldest seniors. Many on the right are similarly concerned about chained C-CPI-U because continuing to use current CPI-U would constrain the growth of federal revenue collections, relatively speaking. I share the policy goals of keeping tax burdens manageable and of ensuring adequate benefits for the most vulnerable seniors. But continuing to overstate inflation is not the appropriate means of achieving these goals -- even with respect to these respective policy advocates’ interests.

The policy goal of increasing benefits for the oldest seniors is more efficiently pursued by changing Social Security’s benefit formula to do so, rather than by overstating inflation in the COLAs provided to all beneficiaries. Income taxes are also better contained by lowering marginal rates rather than by faulty indexing. Moreover, pressure for higher taxes is driven predominantly by growth in federal spending, which would grow faster under current CPI indexing. Both sides of the aisle will also find it easier to argue for their respective policy priorities in an improved fiscal environment.

A case could be made that income tax levels should rise with average income, rather than prices, to prevent bracket creep from steadily increasing individual tax burdens. But as long as the current policy is to index for inflation, the most accurate available measure should be used. No particular policy rationale is served by indexing for inflation inaccurately.

The federal balance sheet would improve, especially over the long term. Deficit reduction alone is not a dispositive reason to embrace C-CPI-U. The fact that its adoption would improve the fiscal outlook is, however, a substantial benefit. To more fully appreciate this, imagine the opposite scenario: imagine that federal laws were currently indexed to C-CPI-U. A proposal to switch to CPI-U or CPI-W would then rightly be criticized both for resulting in less accurate indexing, and for adding recklessly to projected long-term deficits. If C-CPI-U were the measure already on the books, there would be hardly any question that it should be the operative method going forward.

Proposals to adopt an alternative measure of inflation would produce absurd results. Some have argued that an experimental index of inflation developed specially for seniors (CPI-E) should be used to index Social Security COLAs, even though doing so would increase costs and worsen Social Security’s projected shortfall. Methodologically, however, the experimental CPI-E suffers from the same problems as CPI-U and CPI-W in that it fails to account for upper-level product substitutions.

Even if the CPI-E didn’t suffer from significant methodological shortcomings, however, it could not sensibly be applied to Social Security benefits. Social Security beneficiaries come in various forms, from retirees to the disabled to child survivors. It would make no methodological sense to use a purchasing index for the elderly to adjust benefits for child survivors; nor would it make sense for the young disabled. It would also create a nightmare of complexity to have different beneficiary populations using different measures of CPI, shifting between them as they move from one category to the other (e.g., from disabled to old-age benefits). The purpose of inflation indexation is not to model the purchasing patterns of every individual or subgroup, but to model general price inflation, which C-CPI-U does better (even for Social Security’s beneficiary population, on average) than CPI-E.

C-CPI-U has distributional advantages also. Although the method of indexation should not be chosen based on distributional considerations, it should be said in response to some concerns raised that C-CPI-U does carry distributional benefits. Lowering deficits, debt and long-term spending levels would all reduce tax burdens on younger generations. And on the Social Security side, the biggest existing distributional inequity is the net income loss faced by younger generations as a result of the excess of benefits over taxes contributed for earlier generations. Under current benefit formulas, people who have already entered the Social Security system will receive $18.8 trillion ($2011 PV) more than the amount of taxes contributed over their lifetime, creating a deficit that would subtract roughly 4% from the lifetime wage income of younger generations, even if those generations receive all benefits now being promised. We needn’t “cut” the benefits for people now on Social Security, but formulaically exaggerating inflation will grossly exacerbate the program’s intergenerational inequities.

CPI reform is not Social Security reform, and for both tactical and substantive reasons should never have been presented as such. By itself, it won’t fix the long-term budget outlook, which task requires serious further reforms of Social Security and the health care entitlements. It would, however, improve the long-term outlook for the federal budget as well as for Social Security.

Not every distributional consequence of CPI reform will be to everyone’s liking, but that is true of any technical refinement of the federal government’s indexing methods. Altogether, CPI reform is a long overdue correction that would serve the interests of negotiators on both sides of the aisle, of taxpayers – and of the nation as a whole.

Charles Blahous is a research fellow with the Hoover Institution and the author of Social Security: The Unfinished Work.