The Consumer Price Index (CPI) is one of the most widely used statistics in the United States. As a measure of inflation it is a key economic indicator. It serves as a guide for the Federal Reserve Board’s monetary policy and is an essential tool in calculating changes in the nation’s output and living standards. It is used to determine annual cost-of-living allowances for social security retirees and other recipients of federal payments, to index the federal income tax system for inflation, and as the yardstick for U.S. Treasury inflation-indexed bonds.
There has long been both research and policy debate about the appropriate conceptual framework for the CPI and whether it might be overstating changes in consumers’ costs of living. Forty years ago the Stigler committee outlined the difference between the CPI and a “true” cost-of-living index and recommended that the Bureau of Labor Statistics (BLS) undertake research to move the CPI closer toward a cost-of-living index. The subject was given new public prominence in the 1990s by increasing congressional concerns over the budget and the role of the CPI in determining social security cost-of-living allowances and tax indexation. A 1996 report by a congressionally appointed committee—known as the Boskin commission after its chair—estimated that the CPI was overstating the rise in the cost of living by about 1.1 percentage points a year and recommended changes in the way the CPI is designed and estimated.
Underlying some of the arguments and questions about the CPI is a fundamental issue of the nature of the index. Traditionally, a consumer price index measures the change in expenditures required by a household to purchase a fixed-weight basket of goods and services when prices change between some initial
reference period and a subsequent comparison period. The panel labels this a cost-of-goods index (COGI) (for convenience, we use the term “goods” throughout this report to denote goods and services, unless otherwise specified). In contrast, a cost-of-living index (COLI) measures the change in expenditures a household would have to make in order to maintain a given standard of living.
In 1997 BLS told Congress that it had been using the cost-of-living concept for many years as a framework for making decisions about the CPI and that it accepts the COLI as the measurement objective for the index. Recognizing the many theoretical and measurement issues involved in embodying a cost-of-living concept in an index, BLS asked the Committee on National Statistics of the National Academies to convene a panel of experts “to investigate conceptual, measurement, and other statistical issues in the development of cost-of-living indexes.”
A COGI VERSUS A COLI
For dealing with many of the issues considered in this report, there are close parallels between the COGI and COLI approaches. Nevertheless, having a clear conceptual basis for the index is important. It serves as an authority that can be appealed to when making difficult choices among alternative procedures or for accommodating the new developments constantly being generated by a technologically innovative economy.
The cost-of-living approach provides a rationale for taking account of the fact that, when prices change, consumers do not continue to purchase the same fixed basket, but shift their purchases toward goods whose relative prices have fallen. The concept of the COLI explicitly takes into account the effect of this substitution behavior in reducing the expenditure required by a consumer to maintain a given standard of living when prices change.
Probably the single most difficult and important task in index construction is dealing with the ongoing flow of quality changes among consumer goods and services. Many economists consider the economic theory underlying the COLI a helpful way of initially approaching the problem because it prompts the question: “What are the particular attributes of goods that consumers value?” This may provide a way to start, but the panel found that, beyond this point, current techniques for addressing problems associated with changing item quality can be analyzed with minimal use of the theory underlying the COLI and that the techniques could be applied within either a COGI or a COLI framework.
While a COLI framework offers some conceptual advantages, giving up the relative simplicity of the COGI comes at a cost. Conditions that complicate the estimation and cloud the interpretation of COLIs—such as changes in consumer tastes or changes in buying patterns caused by changes in income—may be present in practice. A CPI constructed on cost-of-living principles can, therefore, only be an approximation to the COLI that it seeks to measure. Moreover, re-
stricting the COLI to cover only the universe of private goods and services, as the BLS does and the panel recommends, requires that it be a “conditional” COLI; that is, it should measure changes in consumers’ costs of living on the assumption of stability in conditions—such as the weather or the quality of publicly provided goods—that are outside the universe of private goods. But the choice of exactly which outside conditions should be held constant in the conditional COLI is sometimes controversial and cannot itself be derived from the theory underlying the COLI. And the fact that consumers’ demands for private goods and services often do change in response to outside conditions provides another reason why there is a range of circumstances under which a CPI constructed on cost-of-living principles can only approximate a COLI.
If asked to assess the relative merits of the two conceptual approaches as a guide for construction of the CPI, various members of the panel would strike the balance differently. All panel members find it difficult to think about the definition of goods and about quality change without considering what it is that consumers value, and agree that it is impossible to think about substitution behavior without the concept of a constant standard of living which allows price changes to be converted into a monetary equivalent. For all these issues, especially the last, the cost-of-living framework is central. However, some panel members are skeptical about our ability to define a constant standard of living in an economy in which the nature of goods and services is constantly changing. They point out that the conceptual framework underlying the COLI is not always well defined in the presence of quality change and, therefore, they conclude it provides, at best, a limited advantage over the COGI approach in handling this most difficult of issues. They are also concerned about the BLS adopting an approach that differs from that of many statistical offices around the world.
Despite these differences, all panel members agree that the COGI and the conditional COLI that the panel recommends share many common aspects. We also concur that neither conceptual approach, viewed in its pure form, can provide the single guide to index construction but that each can make a contribution toward dealing with the various problems that arise in designing the CPI. Taking a pragmatic approach, the panel found that it could come, sometimes by different routes, to unanimous agreement on all of the specific recommendations in this report. But in its inability to achieve unanimity behind a recommendation that the cost-of-living framework be the sole appropriate basis for construction of the CPI, our panel differs from the Stigler committee and the Boskin commission.
THE SCOPE OR DOMAIN OF THE INDEX
For the reasons set forth in Chapters 2 and 3 of this report, we arrived at two general conclusions, largely about the conceptual basis for price and cost-of-living indexes, which serve to guide our more detailed conclusions and recommendations that appear later in the report.
An unconditional cost-of-living index is an unsuitable conceptual basis for the CPI. While research aimed at better understanding the economic effects related to changes in such matters as life expectancy, crime rates, or the environment would be useful for evaluating various aspects of public policy, the CPI should not change in response to changes in such factors. (Conclusion 2-1)
Within the general conceptual framework of cost-of-living indexes, the appropriate theoretical concept for the CPI is a conditional cost-of-living index that is restricted to private goods and services and in which environmental background factors are held constant. (Conclusion 2-2)
The BLS should not conduct research on its own aimed at producing a CPI with a substantially broader domain. That said, the panel encourages the BLS— jointly with other federal statistical agencies, particularly the Bureau of Economic Analysis (BEA)—to undertake or sponsor research aimed at producing, on an experimental basis or in satellite accounts, more comprehensive measures of national output, income, and prices. These accounts would seek to include the effects on output, income, and prices from changes in some of what we have labeled “outside conditions” in those cases where there may be at least some chance of measuring those effects—perhaps, for example, changes in the status of the natural environment.
Households differ from one another in their consumption patterns and shopping behavior and often pay different prices for the same goods. Part of this heterogeneity is associated with differences in households’ economic and demographic characteristics and in their geographic location. This fact gives rise to two kinds of issues: First, for such purposes as adjusting social security payments and the tax system, and for measuring changes in real income, when can one aggregate the data for the whole population into a single official price index; when are different price indexes needed for specific population subgroups; and how can the data needed to produce such subgroup indexes be collected? Second, when a single overall index is produced, how should the costs of living of individual households be combined into a single national index? Should equal weight be given to each household’s cost of living (a “democratic” index) or, as is now the case, should costs of living be weighted by the overall consumption spending of each household (a “plutocratic” index)?
The Consumer Expenditure Survey indicates the extent to which various economic and demographic groups allocate their budgets differently among categories of goods and services. The panel believes, however, that substantial
variation may also exist among different groups of households with respect to the particular types and qualities of goods they purchase and the prices they pay within each category. But because the price data used to produce the CPI are collected from retail stores and not directly from households, it is impossible to associate the economic and demographic characteristics of buyers with the items they buy and the prices they pay. As a consequence, it is impossible to investigate satisfactorily the two major aggregation issues: To what extent does inflation or changes in living costs differ among the various economic and demographic groups? And to what extent would a democratic index behave differently from a plutocratic one?
With current survey techniques and methods, collecting price as well as expenditure data from households on a scale sufficient to produce the CPI and an array of group indexes would be extremely expensive and possibly even infeasible; we therefore propose a more modest plan:
BLS should pursue an exploratory research program that would, initially only on a small scale, investigate and assess several alternative approaches—including, but not limited to, the use by survey respondents of handheld scanners and computers—for collecting prices in a way that allows them to be associated with household characteristics. A first objective might be the production of indexes for a few commodity categories and several demographic groups. (Recommendation 8-1)
ACCOUNTING FOR SUBSTITUTION BEHAVIOR
When prices change, consumers tend to shift their purchases toward those goods and services whose relative prices have decreased, thereby reducing any adverse consequences of the price changes on their costs of living. A fixed-basket index does not reflect this substitution effect. The BLS has recently made some changes in the method of constructing price indexes for many categories, or strata, of goods (utilizing geometric means of individual price relatives) in an effort to capture within-strata substitution effects. It will shortly begin producing a superlative index to approximate substitution effects among strata. But because some of the data necessary to construct a superlative index will not be available to meet the CPI’s publication schedule, the superlative index will be available only after a 2-year lag.
The panel agrees that the BLS should continue to produce, as its main index, a real-time CPI. employing a selective use of geometric means for producing individual strata indexes and Laspeyres weights to combine the strata indexes into the overall CPI. Further research should be conducted on consumer shopping and substitution behavior with an eye to improving knowledge of the appropriate application of geometric means at the lower level of index construction.
The BLS should also proceed as planned to begin publishing a superlative index with a 2-year lag. For purposes of producing a timely index for determining cost-of-living allowances for social security benefits and other indexed programs, we recommend an additional series:
The BLS should publish, contemporaneous with the real-time CPI, an advance estimate of the superlative index, utilizing either a constant-elasticity-of-substitution method or some other technique. (Recommendation 7-1)
Dealing with the ever-changing mix and quality of available goods and services poses the most numerous and difficult problems in constructing the CPI. Items constantly disappear from store shelves and are replaced in the index with similar but somewhat different items carrying different prices. The BLS must continually make judgments about how much of a price difference represents “pure” price change and how much represents a quality difference. Increasingly, BLS has been turning to explicit quality adjustment techniques, principally hedonics, in which statistical regressions are used to assign monetary values to differences in the particular characteristics of a type of product and to adjust its reported prices accordingly when the characteristics of the good change.
Hedonic techniques currently offer the most promising approach for explicitly adjusting observed prices to account for changing product quality. But our analysis suggests that there are substantial unresolved econometric, data, and other measurement issues that need further attention. The panel makes a number of recommendations to deal with this set of opportunities and problems:
BLS should continue to expand its experimental development and testing of hedonic methods and its support of relevant outside research. This research should not be confined to that relating to price adjustment but should also examine the role of hedonics in statistical audits of the other BLS quality adjustment methods and in the review of replacement item selection procedures and comparability decisions. (Recommendation 4-2)
The above recommendation does not suggest that BLS should immediately expand the use of hedonics in constructing component indexes for its flagship CPI. In fact, the panel takes the opposite position:
Relative to our view on BLS research, we recommend a more cautious integration of hedonically adjusted price change estimates into the CPI. (Recommendation 4-3)
This recommendation is based on theoretical considerations, not on empirical grounds. As documented in the report, the recent BLS expansion of hedonic price
adjustments to appliances and electronics has not had a large impact on those item subindexes. Our conservative view on integrating hedonics techniques has more to do with concern for the perceived credibility of the current models. While there is an established academic literature on estimating hedonic functions, researchers are much less experienced using them across a wide variety of goods in price index construction. Thus, while members of the panel agree that BLS and others should vigorously continue to research the viability of hedonics, the methods may, in their current state of development, only be justifiably applied to a narrow class of goods.
So long as hedonic techniques are restricted to replacements for items that have disappeared from store shelves, as is now predominantly the case, their use will not have a significant impact on index growth. Only if extended on a broader basis (e.g., to items coming into the index through the rotation of the retail store sample) will the use of those techniques make much difference. Such an extension would be unwarranted until the recommended research, development, and testing program makes progress on the measurement issues we have identified. To assist in this task, we recommend the following:
An independent advisory panel, consisting of econometricians, statisticians, index experts, marketing specialists, and possibly product engineers should be formed to provide guidance on both conceptual and application issues pertaining to hedonic methods. (Recommendation 4-8)
BLS, working with the recommended advisory panel, should assess the impact of modeling imperfections on the validity of their hedonic adjustments prior to their introduction into the index. This would provide an analytic basis for proceeding sensibly in the face of external pressures to proceed quickly in this area. The advisory panel should also provide outside review to help guide decisions about potential new applications and about which BLS pilot studies are adequately developed to be incorporated into the index. Together, our recommendations emphasize the high priority that the hedonics research program should receive.
Another class of product changes involves the appearance of goods with genuinely new characteristics (such as mobility for phones). These goods are sufficiently unlike existing ones in that they do not enter the CPI as part of the item replacement procedure or even when the sample of retail outlets is rotated. Hedonic techniques do not hold much promise for measuring the effect on the index of the introduction of such goods.
If a new good displays new characteristics, it is likely to become eligible for inclusion in the market basket only when item strata are redefined and upper-
level weights reestimated. To the extent that new goods offer previously unavailable benefits to early purchasers and because they typically experience price reductions early on, some declines in the cost of living are missed during the period before the new goods are incorporated into the index. Prominent examples of this phenomenon occurred when mobile phones and VCRs were introduced into the index many years after their appearance on the market.
Some proponents of the COLI approach argue that econometric methods should be used to estimate the “virtual” price reduction that occurs when a new product appears. Those estimates, in turn, could be incorporated into the index. However, the panel had serious doubts about the effectiveness of econometric techniques in this regard, and some members dispute the conceptual validity of treating the benefits from introducing new products as a price decrease:
Virtual price reductions associated with the introduction of new goods should not be imputed for use in the CPI. (Conclusion 5-1)
Members of the panel recognize that, outside of price measurement, there is nowhere in the national accounts for the effect of new products to be included, and real growth in the economy may therefore be understated. Rather than modifying the CPI, the panel suggests that research in this area be directed toward developing a separate experimental COLI that is adjusted, to the extent possible, to account for changes as new products and technologies diffuse throughout the economy.
Additionally, because, once introduced, new goods frequently display very different price trends from established ones, the panel does endorse BLS’s recent efforts to update weights every 2 years, to streamline sample rotation, and to perform targeted product introductions, all of which should enhance the probability that new products will enter the CPI basket more quickly than has historically been the case.
Another potential bias of the CPI, when used as a COLI (or possibly even as a COGI), arises because different stores sell identical items at different prices. If price variation is not proportional to differences in the quality of the retail service offered (as the ongoing trend to lower-price, lower-service outlets might suggest), consumers can lower their living costs by altering their shopping behavior. These types of “price reductions” are not fully captured by the CPI. Currently the underlying conceptual apparatus of the CPI assumes that when lower-price outlets enter the sample, there is no net price reduction, because all of the price difference between the old and the new outlet reflects a difference in the quality of service.
Because current techniques cannot consistently and accurately separate quality changes from the price effects associated with the value of retail service, BLS
has little choice but to continue this practice, though the body of this report does discuss a couple of alternatives. However, in principle, when outlet rotation results in a change in the observed price of an identical product, an attempt should be made to decompose the difference into quality (or convenience) and pure price components instead of attributing it entirely to the former.
With longer-term modifications in mind, the panel recommends pursuing research into price variation across outlets with differing characteristics. (Recommendation 5-2)
PRICING MEDICAL CARE
Medical care, one of the eight major product groups in the CPI, currently accounts for just less than 6 percent of consumer expenditures included in the index. Total expenditures on health care amount to almost 18 percent of consumption outlays, but the domain of the Medical Care Price Index (MCPI) in the CPI is limited to consumers’ out-of-pocket expenditures, thus excluding costs paid by Medicare, Medicaid, and employer-financed health insurance (as well as other smaller items). In the case of health insurance premiums paid by households themselves, the BLS does not price the premium cost of the insurance directly but imputes to it the prices of the underlying medical care services that are purchased with the premium.
Because of the complicated institutional setting in which medical care services are provided and financed, together with the rapid pace of development of new medical technologies, their appropriate pricing probably constitutes the most difficult single task in producing the CPI. The panel makes a number of recommendations in this area:
BLS should select between about 15 to 40 diagnoses from the ICD (International Classification of Diseases), chosen randomly in proportion to their direct medical treatment expenditures and use information from retrospective claims databases to identify and quantify the inputs used in their treatment and to estimate their cost. On a monthly basis, the BLS could reprice the current set of specific items (e.g., anesthesia, surgery, medications), keeping quantity weights temporarily fixed. Then, at appropriate intervals, perhaps every year or two, the BLS should reconstruct the medical care index by pricing the treatment episodes of the 15 to 40 diagnoses— including the effects of changed inputs on the overall cost of those treatments. The frequency with which these diagnosis adjustments should be made will depend in part on the cost to BLS of doing so. The resulting MCPI price indexes should initially be published on an experimental basis. The panel also recommends that the BLS appoint a study group to consider, among other things, the possibil
ity that the index will “jump” at the linkage points and whether a prospective smoothing technique should be used. (Recommendation 6-1)
Additionally, the panel concluded that a price index including a more broadly based measure of the changing cost of medical care would be valuable for a wide range of policy purposes.
BLS should include the portion of health insurance paid for by employers in one version of the CPI, perhaps calling it an “expanded-scope medical CPI.” Because many commonly used income measures exclude employer-provided benefits, and because the Consumer Expenditure Survey is based only on out-of-pocket expenditures, the original conception of the MCPI domain should still be maintained in constructing the traditional (flagship) CPI. The panel also recommends examining the practicality of including other employer-paid employee benefits (e.g., dental and cafeteria plans) in the expanded-scope CPI. (Recommendation 6-2)
To inform public policy discussions and to evaluate the performance of the U.S. health care sector, a medical care price index that encompasses purchases from all payers is needed.
A task force should be convened by the BLS, in collaboration with the Centers for Medicare and Medicaid Services and other appropriate agencies, to implement construction and publication of a total medical care expenditure price index, encompassing purchases from all health care payers—governments, private third-party insurers, and consumers. (Recommendation 6-3)
The most difficult issue in the construction of the MCPI concerns adjustments for quality change. New treatments can yield improved outputs in the form of extended and better quality life. The panel believes that an outcomes-based measure is in principle superior to an input-based measure, but we recognize the formidable measurement challenges and do not know how best to proceed. This area is new and requires a good deal more research, much of it interdisciplinary. After BLS has implemented Recommendation 6-1, it can then consider whether, how, and why the outcomes of the treatments for those diagnoses are changing over time, and finally consider how outcomes changes should best be evaluated in computing a quality-adjusted medical care price index.
INDEX DESIGN AND INDEX PURPOSE
The CPI and its individual components are used for a wide range of sometimes dissimilar purposes. In some cases different uses may call for different index designs. But no statistical index can perfectly match what is desired for a
particular purpose, and practical considerations limit the number of indexes that can or should be produced. Chapter 7 evaluates the extent to which the CPI and existing or proposed supplemental indexes meet the needs of various users.
The panel concludes that a superlative index is appropriate for adjusting benefits to keep pace with the cost of living. In this context, the panel suggests the following:
It would be feasible and appropriate to calculate cost-of-living allowances provided for social security and other programs from an advance estimate of the BLS published superlative index. Any divergence between that estimate and the superlative that appears 2 years later could be incorporated as a correction to the cost-of-living allowance provided for that year. (Conclusion 7-1)
A related question is whether social security cost-of-living allowances (COLAs) should be based on a special index for the elderly. Using data from the last several decades, BLS has produced a special index for the elderly (CPI-E) by weighting the price indexes for various categories of goods according to the purchasing patterns of the elderly rather than the general population. This index did not rise at a significantly different rate than the overall CPI. Different groups not only have different overall consumption patterns but face different prices and buy different qualities of goods; the BLS has called attention to this limitation of its experimental index. In the absence of an index that can capture these differences, we see no rationale for basing social security COLAs on the type of indexes constructed in the BLS studies. But the CPI-E should be periodically updated to make sure that no significant differences with the CPI have developed.
Adjusting social security benefits for retirees with a wage index would be an alternative to CPI indexation. The panel was not charged to make recommendations on this issue, but we do spell out the implications of this and other indexing methods for public policy.
The data inputs used to calculate the CPI subindexes originate from several sample-based sources, most notably the Consumer Expenditure Survey (CEX), the Point of Purchase Survey (POPS), the Commodities and Services Survey, and the CPI Housing Survey. The panel considered two distinct approaches for upgrading this apparatus. One is to assume that the basic data collection structure will remain as is and then to seek ways of improving each of the survey components. Another is to redesign, from scratch, the entire data collection structure so that it reflects advances in data collection technology and so that the data collected are more consonant with the ultimate computation of the CPI.
The panel’s foremost concern with the CEX, which is the primary tool for establishing CPI weights at the basic item level, is the extent of biases in house-
hold-reported expenditures which, in turn, affects the accuracy of upper-level CPI item category weights.
Before additional resources are directed toward increasing its sample size (beyond the current plan), the accuracy of the CEX should be carefully evaluated. Assessing the net advantages of using the BEA’s per-capita personal consumption expenditures (PCE) data to produce the upper-level weights for the national CPI should be part of this evaluation. (Recommendation 9-1)
Comparison of the CEX and PCE estimates suggests that, even allowing for errors in the latter, the CEX generates biased weights for a number of items. Even if the current system is ultimately maintained, the effort will produce additional guidance about how the CEX might be improved.
If categories can be reasonably well matched between the CPI and PCE, so that comparable item strata indexes can be created, a program should be set up to produce an experimental CPI that uses PCE-generated weights at the upper (218 item) level but that is otherwise no different from the CPI. (Recommendation 9-2)
Even if it is confirmed that the CEX is the best choice for establishing upper-level expenditure weights, the panel is hesitant to recommend expensive increases in the sample size. The panel’s calculations suggest that, if the goal is only to reduce the standard error of the national-level expenditure weights, resources spent to increase the sample size of the CEX beyond that which is currently planned would be largely wasted.
In considering alternative data collection approaches, the panel suggests that BLS (1) investigate the possibility of combining the POPS and CEX into an integrated survey that obtains expenditure and outlet-use data at detailed product levels, along with household demographic information needed for subgroup indexes and (2) continue its work on increasing the utilization of both store- and household-based scanner data.