Sections

Commentary

The Politics of Immaculate Conception: Congress should set the budget, not the CPI

March 1, 1997

Last December the Advisory Commission to Study the Consumer Price Index issued a report arguing that the CPI has been overstating inflation for the past few decades—by 1.1 percent in recent years. The commission recommended that the overstatement be corrected forthwith.

Years of neglect and underfunding have seriously eroded the quality of the U.S. statistical system, and the commission’s attention to the adequacy of the CPI as a measure of inflation is welcome. But the accuracy of national statistics was not the only issue Congress had in mind when it appointed the commission. Adjusting the CPI in accord with the commission’s recommendation would have major effects on taxes, federal benefits, and the goal of future budget balance.

The commission’s conclusion that the CPI overstates inflation was not surprising. It is consistent with earlier research on the price indexes, and few analysts would disagree. But there is less empirical basis for the commission’s estimate of the magnitude of the bias. While the estimate is consistent with a number of recent survey articles, it is important to understand that all the reports are extrapolating from a common small set of empirical studies, none of which was intended to provide an unbiased assessment. The empirical studies, most arising from research conducted at the Bureau of Labor Statistics, focused on specific problems with the CPI—and, naturally, on those areas of the CPI where the effect would be largest. At this time, we do not know the extent to which the results of studies of those narrow areas can be generalized to other parts of the index.

The issue of bias in the CPI involves both technical aspects of constructing an index and measures of quality change. The technical issues are straightforward, and the commission estimates their impact on the rate of change in the CPI at 0.5 percent per year. The commission has highlighted some basic problems and provided specific suggestions that would help resolve them. Congress should provide the funding to update the index, and the BLS should move in the direction proposed by the commission. The result would be a better index.

The problems with the commission’s recommendations arise in connection with the issue of quality change, the remaining 0.6 percentage point of overstatement. In the first place, the commission’s report fails to reflect fully the extent to which the current CPI already adjusts for quality change. Of course the quality of some goods and services has improved. But that is not the point. The real question is whether the quality improvements have been more than are embodied in the current procedures. For, though few people outside the BLS seem to know it, the index already reflects a large amount of quality gain. In 1995 the total price increase in a subsample of the CPI, covering about 70 percent of the total, amounted to 4.7 percent. But the BLS determined that 2.6 percentage points, or slightly over half the increase, represented improvements in quality. That is, quality improvements offset about half of the gross price increase, yielding an estimated 2.2 percent inflation rate.

The CPI commission is arguing that the quality adjustment should have been 0.6 percentage points larger, 3.2 percent rather than 2.6 percent. They could be correct, but on the basis of existing evidence, we don t know that. I don’t see how anyone can be that precise. Indeed, one disappointing feature of the commission’s report is that it offers no new procedures to improve the estimate of quality change; it simply asserts that it is larger than estimated by the BLS.

Much of the dispute arises out of different views of how producers go about introducing price and quality changes. If most quality changes are small and incremental in nature, current BLS techniques may overlook them and end up misstating the price change. But if price and quality changes are tied closely together in the introduction of new models, the BLS methodology may overestimate quality improvements and understate inflation, a possibility ignored by the commission. In most cases, major changes in a product’s characteristics will result in its being dropped from the index, implicitly attributing most of any price change to quality. Thus, price increases that are imbedded in new models are ignored and assumed equal to those of other products in that month. On average, a product’s characteristics will undergo major change every two years.

In addition, the commission argues that the CPI should be an index of the cost of living and not just the average cost of a fixed market basket of goods and services. That seems right in concept. A cost of living index would reflect the fact that consumers can avoid some price increases by switching to substitute products. But pushing that idea to the extreme can open a can of worms that exceeds the capacity of the current methodology. Where do we draw the line between economic and noneconomic aspects of the cost of living, and are they separable? How should we value the time required to shop for the lowest price and increases in the range of consumer choices?

Congress is drawn to the issue not out of a fascination with the arcane issues of index number construction, but because the CPI is used to adjust large elements of the federal budget—Social Security benefits and income tax brackets—for annual increases in the cost of living. A revised CPI offers an “immaculate conception” version of deficit reduction in which spending is cut without Congress taking the blame. While the first-year savings from a lower CPI would be small, they build up over time. A 1.1 percentage point reduction in the annual increment to the CPI would save $150 billion in the tenth year, or about 1 percent of the GDP. It is becoming evident, however, that procedural changes by the BLS will generate only a fraction of the commission’s suggested overestimate of 1.1 percent. Thus, it is suggested that a group of “experts” who are knowledgeable of the actual increase in the nation’s average cost of living could provide Congress with a more accurate measure, again providing Congress with some cover for benefit reductions.

Were it not for the politics of immaculate conception, I don’t think that Congress would be considering this particular form of benefit reduction. Consider, for example, its impact on the nation’s oldest citizens. The typical retiree can expect to receive Social Security benefits for about 20 years. The importance of these benefits rises with age: private pensions have no inflation adjustment and retirees tend to use up their own assets. The proposed adjustment to the CPI would have no effect on the benefit of a new retiree, but benefits in each succeeding year would rise by a smaller percentage. After 10 years, benefits would be reduced by about 12 percent; after 20 years, by 25 percent. Currently, the average income of those over age 80 is only two-thirds that of people aged 65-69, and the poverty rate rises from 10 percent for families with a head aged 65-69 to 18 percent for those over age 80. The recommendation of the commission would exacerbate that trend. If the recommendation had been in effect for the past two decades, the proportion of those aged 65-69 with income below the current poverty standard would rise only marginally to 11 percent, but for those over age 80 it would soar above 30 percent.

Congress will have to scale back Social Security benefits in future years, but it should consider carefully the implications of how it makes the cuts. It makes far more sense to concentrate the cuts at the beginning of the retirement period when individuals can mitigate their effect by postponing retirement or working part-time, not when retirees are in their 80s and have no employment options. Nor should Congress cede its responsibility to make critical public policy decisions to a group of putative experts. Unless they are expert shoppers, it is not at all clear what they are supposed to know about annual changes in the cost of living that cannot be incorporated in the procedures of the BLS.