Full Text: BLS Commissioner's Testimony At Finance Hearing On Whether To Modify Cpi.

Statement of

Katharine G. Abraham

before the

Senate Finance Committee

UNITED STATES CONGRESS

March 13, 1995

Mr. Chairman and Members of the Committee,

I appreciate this opportunity to appear before you to discuss the Consumer Price Index (CPI), for which, as you know, the Bureau of Labor Statistics (BLS) is responsible. I think it would be useful for me to begin by providing a brief summary description of the CPI for background purposes. Although the CPI is often referred to as a "cost-of-living" index, it is in fact a measure of the average change in the prices paid by urban consumers for a fixed market basket of goods and services. Historically, this market basket has been updated roughly every ten years. The current market basket is based on 1982- 84 expenditure patterns; we are in the process of updating the index to reflect 1993-95 expenditure patterns.

Although we talk about THE Consumer Price Index, I would note that the BLS actually produces Indexes for two different population groups each month, one including all urban consumers and a second consisting just of wage earners and clerical workers. The index typically reported in the media each month is the Consumer Price Index for All Urban Consumers, or CPI-U, which was introduced by BLS in 1978 and is representative of the buying habits of about eighty (80) percent of the population. BLS also produces the CPI-W, a continuation of the Index that was introduced about three quarters of a century ago, which covers only wage earners and clerical workers, who represent roughly 32 percent of the U.S. population.

The BLS calculates the CPI on a monthly basis and publishes it about two weeks after the end of the month to which it pertains. To calculate the CPI, BLS collects some 80,000 price quotations each month, covering all items that consumers purchase for daily living. These prices are collected in 85 areas throughout the country, in about 21,000 retail and service establishments. An additional 40,000 landlords or tenants and 20,000 owner-occupants are surveyed concerning rents and housing costs. The collection of such a large number of individual prices each month is dictated by the importance of producing a measure that meets high standards of statistical reliability. All of this information is collected and processed on an extremely tight schedule, and the Index is issued each month on a date announced in advance, to avoid any appearance that the timing of the release has been affected by non-statistical considerations. As the data for any one month are released to policy makers and the public, the process of collecting the next month's data is already well underway. In short, the production of the CPI is an extremely complex undertaking carried out each month according to a rigorous time schedule that leads to the release of each month's data in as timely a fashion as possible.

Having provided this background, I would like now to turn to the various issues that have been raised concerning the present method of determining the CPI. Perhaps the best starting point for a discussion of CPI measurement issues is to note that what many people mean when they talk about inflation is the change in the cost of living. Despite the best efforts of BLS and of economists in academic, business, and other public and private organizations, many people refer to the CPI, mistakenly, as a "cost-of-living" index. By design, as I have already noted, the CPI is a measure of the change in the cost of purchasing a fixed market basket of goods and services. As published BLS descriptions of the CPI make clear, the fact that the CPI is based on a fixed market basket means that it will tend to rise somewhat more rapidly than would a true cost-of-living measure. The reason for this is that the CPI does not reflect changes in buying or consumption patterns that consumers would be expected to make as they adjust to relative price changes, buying more of goods whose relative prices have fallen and less of goods whose relative prices have risen. By making these changes in how they spend their money, consumers may be able to maintain their level of well-being at a lower cost than that indicated by the CPI. BLS research suggests that not accounting for these substitutions in response to relative price change raises the annual change in the CPI by 0.1 to 0.2 percentage point.

A true cost-of-living measure also would take into account changes in the external environment that might impact consumers' out of pocket expenditures. These might include such things as, for example, a deterioration in air or water quality or an improvement in the quality of publicly provided services, such as education. I also would quickly add that, were we able, we would produce a true cost- of-living measure. Unfortunately, the "state of the art" in the area of price index construction has not advanced to the point where anyone knows how to construct true cost-of-living measures. In fact, although there are alternative price index formulations that provide a better approximation to cost-of-living measures than our current CPI, none of these alternatives is now feasible to produce on a "real time" basis.

Given that the CPI is not, and is not represented to be, a measure of changes in living costs, it is nevertheless appropriate to ask how well the CPI measures what it is designed to measure. In this connection, two broad areas of concern have been identified. The first, brought to light by researchers at BLS, has to do with the construction of the most disaggregated components of the index, and in particular with the way in which new items enter the index as part of routine sample replenishment and the way in which the treatment of new types of retailers -- such as discount stores -- impacts the index's measured rate of inflation. The second, and less well understood, issue is the question of how well the index accounts for changes in the quality of the goods and services that consumers purchase.

The sample rotation effect arises because procedures for systematically introducing new outlets and items into the CPI inadvertently tend to give higher weight than is justified to prices that are temporarily low in the month the new samples are introduced and lower weight than is justified to prices that are temporarily high. Thus, these procedures can cause an overstatement of price change in the period immediately following sample replacement. The BLS has taken steps to address the sample rotation problem effective with the data for January 1995. If further corrective measures can be identified, they will be incorporated as expeditiously as possible in the context of the ongoing CPI revision.

The outlet substitution effect can arise because consumers are free to substitute where they buy goods and services as well as what they buy. For example, if consumers don't consider the (possibly) lower level of customer service provided by a discount store to be of any consequence, they may shift to such stores and experience no loss of well being. Current CPI procedures would not capture any price decline associated with such a shift. Although it is unclear that there is in fact any bias associated with the CPI's treatment of discount outlets, further research on this issue would be valuable.

It is axiomatic that a measure that purports to estimate changes in prices must take account of the fact that the quality of the goods and services purchased in our economy can, and does, change, in some cases for the better and in some cases for the worse.

Today's cars, for example, are substantially more expensive than the cars sold in the 1970s. Today's cars, I think it would widely be agreed, also are substantially better than were the cars of the 1970s, in the important sense that they embody more of the features -- such as durability, safety, and lesser maintenance requirements -- that consumers value. In measuring the price change for cars over this period, the challenge is to isolate that part of the price increase associated with improvements in the quality of vehicles, as distinct from that part that is truly a pure price change. In the case of new automobiles, for example, adjustments made in the CPI to factor out quality change have had a very substantial impact. We estimate that the change in the new automobiles component of the CPI over the years from 1967 to 1994 would have been more than 80 percent greater than we actually reported had no adjustments been made for quality improvements. (The automobile index in the official CPI-U rose 172.1 percent from December 1967 to December 1994, while over the same period, without factoring out changes in quality, the new car component would have risen 313.4 percent.) From a different, and perhaps more important, perspective, the overall CPI today would be nearly 3 percent higher had we not made quality adjustments for the single index component of new cars. (The CPI-U All Items rose 341.6 percent from December 1967 to December 1994, while an All Items CPI-U without factoring out the new car quality adjustments would have increased 351.1 percent.)

The more general point is that efforts are routinely made in every index component to insure that changes in quality are not recorded as price changes. These efforts range from the prosaic case of adjusting for the fact that a 1 ounce candy bar is worth more than a 0.75 ounce candy bar to consideration of the more difficult question of what is the value to the consumer of a new, non-invasive diagnostic medical test that replaces an earlier, more taxing and riskier test for the same condition.

The emergence in the market of entirely new goods or services presents perhaps the most difficult quality adjustment problem. These "new goods" are so radically different from anything previously on the market that they have no obvious earlier counterparts with which their costs can be directly compared. Electronic calculators, video cassette recorders, and personal computers are often cited as examples of "new goods". Current CPI procedures lead to these new goods being included in the index in a comparatively timely fashion, and the new procedures to be introduced as a part of the ongoing CPI revision will allow them to be included even more quickly. What remain to be developed, however, are methods that enable direct comparison of a new good's price with that of its closest antecedent.

Some have estimated that the overstatement in the CPI is as large as 1.5 percentage points per year. Estimates of this size require that there be a large quality-adjustment bias, arising as a consequence of substantial improvements in the quality of the goods and services consumers purchase that are not reflected in the construction of the index. Although many believe that the CPI is biased upward because quality improvements are not fully accounted for, there is little direct evidence to support this view. Indeed, some have suggested that quality adjustment problems may lead to a downward, not an upward, bias in the CPI, at least during certain periods. Adjusting for changes in the quality of goods and services remains one of the most challenging tasks in constructing any price index. Solutions to the problem, it is widely agreed, are not obvious.

A promising strategy for improving the CPI's accounting for changes in the quality of goods and services would be to expand the collection of information on the characteristics of items, which would allow BLS to estimate the value of particular features and explicitly adjust items' prices for changes in those features. This so-called "hedonic" approach is currently used in several components of the index but its extension to other areas would require additional resources.

Assessments of the total bias in the CPI as a cost-of-living proxy vary considerably. Federal Reserve Board research staff have concluded that the CPI may overstate the change in the cost of living by 0.4 to 1.5 percentage points per year, though they also say that "these estimates are by necessity extremely rough." Another review done recently by researchers at the Congressional Budget Office concludes that the bias in the index is probably much smaller, in the range from 0.2 to 0.8 percentage point. Researchers at the Dallas Federal Reserve Bank conclude that "a figure of less than 1 percent . . . strikes us as a plausible estimate of the overall (upward) bias" in the CPI, but add "the true figure may be a lot larger or a lot smaller; at present we simply do not know."

In closing, Mr. Chairman, we are intensely aware of the sensitive nature of the data we produce, and of the critical need for these data to be as accurate as possible. As the revised methods we introduced in the most recent published CPI (released three weeks ago) indicate, we are, within the limits of our resources, pursuing these issues and introducing improvements as quickly as we can. In addition, now that the comprehensive updating of the CPI, normally conducted at ten-year intervals, has been funded, we are moving as expeditiously as possible to complete that major undertaking, while, of course, adhering to our normal standards of care and thoroughness to assure that mistakes are not made. In the interim, Mr. Chairman, we will continue to investigate the measurement issues that we and others have identified, and I can assure you that we will address them to the full extent that it is possible for us to do so.