Whether the desired index is a COGI or a COLI, decisions must be made about its “domain,” that is, about the universe of the things it is to cover. As we explain in Chapter 2, the conceptual underpinning of a cost-of-goods index (COGI) is sufficiently unstructured that its domain can be straightforwardly defined as whatever is considered appropriate. It is natural to think of the index as the price of what people buy. Throughout its history, the Consumer Price Index (CPI) has been defined to include only the prices paid for private consumer goods and services.1
The Bureau of Labor Statistics (BLS) has, in recent decades, increasingly emphasized the concept of the cost-of-living index (COLI) as a framework for making decisions about the CPI and “accepts the COLI as the measurement objective for the index,” with the caveat that “while the CPI may be described formally in the context of a cost-of-living index, there is no single all-purpose definition of this target” (Bureau of Labor Statistics, 1997c:3). From the standpoint of an individual consumer, an all-encompassing COLI, defined without limits on its scope or domain, would have to take into account effects on the consumer’s standard of living arising from changes in the social and physical environment, such as air quality and the crime rate, and in the provision of public goods furnished by the government. The BLS, however, has continued to define
the COLI to include only the effects arising from changes in the prices of private goods and services purchased in the market, i.e., the same domain as that specified in the traditional fixed-weight index. More precisely, according to the BLS (1997c:3): “The cost-of-living index approximated by the CPI is a subindex of the all-encompassing cost-of-living concept, specifically a sub-index that is conditional on the excluded factors that affect consumer well-being such as health status and the quantity and quality of government provided goods and services.”
A conditional cost-of-living index as defined by the BLS is the minimal expenditure ratio needed to attain a given standard of living in the face of changes in the prices of private goods and services, on the assumption that “outside conditions”—the status of the social and physical environment and the provision of goods by the government—remain at some specified level. Changes in the various conditions outside the universe of the provision of private goods and services can affect a household’s cost of living in two ways. They can directly alter its standard of living—the crime rate worsens, additional households become crime victims, and more people begin feeling less secure. But the increase in the crime rate may also change the household’s taste for private goods; it may increase its spending on home security systems and cut spending on other goods, especially those, like patronage of downtown restaurants, that it believes will increase exposure to crime. Changes in the relative prices of private goods would then not have the same effect on a household’s cost of living as it would with the original crime rate. A COLI conditioned on the stability of environmental conditions would measure the expenditures on private goods that would be needed to maintain a given standard of living for the household in the face of changes in the prices of private goods on the assumption that no change in the crime rate had occurred. It would therefore exclude both the direct and the indirect effects of any changes in outside conditions that did occur.
For most if not all of the purposes for which it would be needed, a conditional cost-of-living index should measure the expenditure ratio needed to maintain the reference period’s standard of living, given constancy of the environmental conditions at their reference period status. If changes in environmental conditions do occur between the reference and comparison periods, the superlative index that is used to approximate the conditional COLI would, as desired, exclude their direct effect on the household’s standard of living. But if the environmental change alters some of the marginal rates of substitution among private goods—as in the crime rate example above—the superlative index will not provide a first-order approximation to a COLI conditioned on the status of the excluded variables in the reference period.
As we note in Chapter 2, Diewert (2000b) demonstrated that, when external conditions change, a superlative index will approximate a COLI that maintains the household’s standard of living at some intermediate level between the reference and comparison periods and under the external conditions that lie between
those two periods. Consequently, the conditional COLI that is needed cannot be reproduced by the superlative index. One interpretation of this result is that environmental changes that affect marginal rates of substitution among private goods will reduce the accuracy with which a superlative index measures the conditional cost of living.
The extent to which changes in outside conditions affect the accuracy of a superlative index in meeting its stated objective will depend on how significantly the changes in outside conditions alter the pattern of preferences for private goods between the reference and comparison periods and how those alterations interact with the changes in relative prices that occur over the same period. If the weights used in the index are frequently updated, as the BLS now plans, a slow and gradual drift in such outside conditions as the quality of the physical environment or the crime rate are not likely to have much effect. But an event such as an unusually severe winter might have more noticeable, even if temporary (and reversible), consequences.
AN UNCONDITIONAL COLI: CONCEPTUAL ISSUES
Incorporating into a cost-of-living index the effects of changes in environmental conditions and government-provided public goods would require analytical and measurement techniques that, in most cases, go well beyond the current state of the art. But even if measurement of such effects were feasible, conceptual question arise about whether a cost-of-living-index should take them into account. Put another way, what should be included and what excluded from the variables that are held constant in a conditional COLI? Should the index incorporate the net effects on consumer welfare (not already reflected in the costs of private production) from such public goods as those furnished by the military establishment, the preservation of wilderness areas, or the provision of law and order? Do the effects of changes in the state of national security belong in the index, implying, for example, that a large drop in the cost of living occurred at the end of the Cold War when the threat to national security almost surely declined? What about the motor vehicle accident rate (whose effects would probably be easier to measure than those associated with, say, the crime rate)? Should the index include increases in longevity arising from general improvements in medical knowledge and techniques (as distinguished from those associated with particular medical procedures which, conceptually, should be treated as a quality change in a private good)? How feasible is it to make such a separation? In its discussion of quality-of-life issues the Boskin et al. (1996) report cited as negative factors, presumably tending to drive up the cost of living, “such social issues as divorce, illegitimacy, and the reduced role of the nuclear family.” In concept, at least, should the cost of living be defined to include effects of intangible factors such as these?
The panel distinguishes sharply between what is appropriate for inclusion in
the official family of consumer price indexes2 on the one hand and the scope of broader satellite or supplemental measures on the other, which, developed with care and prudence, might provide valuable additional knowledge about the economy and society. Even if acceptably reliable measures were available that could quantify the effects on living standards of changes in some or all environmental conditions, including the provision of public goods, the panel believes those conditions belong on the list of conditioning variables. Their effects should not be incorporated into a measure that is used for such purposes as indexing public and private payments, the income tax code, private contracts, providing an overall measure of inflation for consumer goods, and as an indicator for monetary policy. Within the COLI framework, this requires that the index be defined as a conditional COLI, holding constant the reference period status of environmental and other goods not included in private goods and services.
Congress and the President decide how social security retirees and other transfer recipients should be compensated for the effects of rising prices, and it seems extremely unlikely that any support could be secured for a system that would reduce the purchasing power of social security recipients in response to, say, the end of the Cold War (as national security increased) or to the recent reductions in the crime rate. The concept of inflation would be altered beyond recognition and macroeconomic usefulness if it is defined to include the effect on living standards stemming from a rise in pollution and the crime rate or the effects of a severe winter. For example, in a comprehensive index that included the effects of changes in outside conditions, suppose that the subindex for private goods were stable while deteriorating outside conditions were forcing up the overall index. Surely we would not want the Federal Reserve to pursue a restrictive monetary policy to deflate the prices of private goods (whose stability, after all, is what is relevant for efficient business and consumer planning) in order to avoid an “inflation” in the official index as it responded to increased pollution, higher crime rates, or a colder winter.3 We discuss these points further in Chapter 7, which examines the relationship between the design of price indexes and the purposes to which they are put.
There are other question about index domain that pose issues that are less clear cut than most of those discussed above. Should the CPI, for example, ultimately aim to include the estimated effect on living standards arising from the introduction of new goods or from the increase in longevity and decrease in
morbidity from technological advances in medicine? Should it seek to incorporate the improvement in living standards that might accompany the introduction and gradual spread of mass merchandising and discount stores, including an explicit adjustment for associated differences in retail service quality and shoppers’ time? The conceptual background for considering many of these questions was explored in Chapter 2, and they are discussed in more detail, together with recommendations, in Chapters 5 and 6. In the last part of this chapter we briefly deal with issues related to the treatment of certain goods furnished by government and by private employers (fringe benefits).
A cost-of-living index that seeks to include the status of the social and physical environment and government-provided public goods must measure the effects of changes in those variables on consumer welfare. That is, it must measure the expenditures necessary to maintain a given standard of living, taking into account the value to consumers—positive or negative—flowing from changes in those variables.
Correspondingly, an expanded measure of changes in national output would include the value of changes in the provision of goods by government and, with the appropriate sign, the value of any changes in the flow of damages from various kinds of environmental “bads,” weighted by their implicit prices.4 Increases in pollution, crime, congestion or other environmental damages, arising chiefly from human activity, would reduce the expanded output measure, while activities (such as pollution controls) that reduced these damages and whose benefits were greater than their costs, would raise the augmented output measure.
Although there are major difficulties that have to be overcome in defining and then measuring the relevant quantities of many environmental and government-provided public goods, the problems encountered in assigning values to these goods are truly formidable. The fundamental difficulty stems from the fact that not only the goods provided by government but also most environmental goods are public goods. Public goods, such as clean air, have the characteristic that if they are available to one person they are available to all—globally, nationally, or regionally—whether or not particular individuals desire to purchase them. Only to a limited extent can individuals choose how much clean air or safety against crime they want to pay for and consume. They can purchase home security systems to reduce their risk from crime or install water filters to increase the
purity of their tap water. They can move from a city with a high crime rate to one with a low rate. But in the main they cannot buy a pro rata share of safe streets and ozone reduction or their own personal protection against global warming.5 This fact has a number of consequences for environmental goods (or any other class of public goods) that make it exceedingly difficult to evaluate their economic worth.
First, by their purchases and the prices they pay, people reveal information about their preferences among private goods. Some of their preferences for environmental and other public goods influence their private decisions—such as in what city to live and in what neighborhood to buy a home—and thereby affect the rents and land prices. But not all environmental and related preferences exert such influences, and when they do the information they convey has to be estimated indirectly, often through a chain of inferences, each step of which entails the likelihood of reducing accuracy. For example, the estimated wage premiums paid to people in high-risk jobs have been used to place a value on the life-threatening effects of air pollution and traffic accidents. The difficulties of valuing public goods have often driven researchers to the use of contingent valuation techniques. Individuals are asked in a survey to indicate what they would pay for a particular public good—e.g., to increase visibility in a specific national park. But these techniques are highly controversial. The hypothetical nature of the question posed, the lack of any budget constraint on the respondent, and the absence of any behavioral observations cloud the results from contingent valuation approaches (see Diamond and Hausman, 1997).
A second characteristic of public goods is that they cannot be bought and sold in the market. Changes in their supply must be decided through political processes, which typically do not lead to an effective balancing of costs and benefits at the margin. People can, by changing their place of residence, make choices among alternative comprehensive packages of public services provided by different local governments. But there are costs of moving, and people cannot make choices among the many individual components of the package of services offered by different local jurisdictions.6 The choices are still more restricted
among public goods furnished by state governments and, for most people, individual choices do not exist for the public goods provided at the national level, such as national security, the federal justice system, or the national air pollution standards. In such cases, households cannot adjust the quantities of public goods they consume in accordance with their individual preferences. The empirical evidence suggests that the marginal costs of providing an extra unit of an environmental good such as clean air through emission abatement measures are not likely to equal the sum of the marginal valuations of the good by the affected households.7 The demand price is not equal to the supply price and so, in most cases, it cannot be assumed that cost data will be a reasonable proxy for consumer valuations.
Of course, GDP does measure and include the gross amounts spent by federal, state, and local governments to purchase goods and services ($1.7 trillion in 2000). These goods and services include such diverse items as arms purchases and the wages of the armed forces; the wages and salaries of teachers, police forces, and other civil servants; the building of highways and dams; and the maintenance costs of the national parks. But these are not themselves public goods. They represent the costs to the government for the labor force, materials, and capital goods it uses to produce public goods and services for the population—national security, education, law and order, new medical technologies, and so on. And, as we noted above, the political process does not tend to produce public goods in such an amount that, at the margin, the sum of the values attached by the population to an additional unit of public goods typically equals the additional costs of producing it.
In sum, research efforts to develop measures of output and prices that include the effects of changes in some important environmental conditions might well yield useful knowledge. But measures of the effects of changes in such conditions on the welfare of consumers depend on a complex interaction of often controversial assumptions and inferences that are far too speculative to be professionally and publicly accepted as part of an official index used for the important public and private purposes for which the CPI is used.
SUPPLEMENTAL INDEXES AND SATELLITE ACCOUNTS
While we recommend that the CPI should continue to be confined to the domain of private goods and services, we also believe that a research program to develop supplemental measures of the economic value of changes in several important environmental conditions is a worthwhile objective. So long as its limitations are made clear, such research should be undertaken in an effort to generate rough-and-ready but still useful knowledge about the effects of some aspects of environmental changes on the material well-being of the population. If successful, it could also aid in the management of the nation’s environmental resources. But we do not believe that this is a task that can be efficiently undertaken by the BLS on its own and with the principal objective of producing a more comprehensively defined cost-of-living index. Rather, it can best be pursued in the process of developing an integrated set of expanded output measures and their associated price indexes. Several considerations point in that direction.
In the case of most private goods and services, real output is not measured by collecting data on physical quantities, but by deflating observed nominal expenditure data with an appropriate price index. The output measures are, in effect, derived from the price measures. The prices used are observed market prices, and even when those prices are adjusted for quality change, the adjustments are typically based on estimates themselves derived from market prices (hedonic adjustments) or from cost data. But principally because of their “public goods” characteristics, the measurement of changes in both the “output” and the “prices” of environmental goods have to be derived independently: nominal expenditure data do not exist. Complex measures of physical changes typically have to be estimated, and their implicit prices have to be estimated indirectly.
In the important case of air pollution, for example, ambient air concentrations of individual pollutants (such as carbon monoxide, sulfur dioxide, or particulates) must be measured and weighted in terms of human exposures (or, in some cases, exposures of various building materials). Physical dose-response measures, such as changes in the incidence of respiratory ailments per unit change in ambient concentrations, must then be estimated from clinical or epidemiological studies. In many cases, it is the flows of relatively short-lived pollutants into the environment that cause the damage; in other cases, it is the stocks of pollutants in the atmosphere that produce gradual and long-lasting flows of damages. Once dose-response output effects are identified, a value per unit of response must be constructed by one of a number of methods (noted above).
Some of the damage from air pollutants directly lowers human welfare by increasing morbidity and mortality, and much of this kind of damage is currently unaccounted for in the national economic accounts. Some of the damage, however, lowers the productivity and raises the costs of producing private goods— e.g., the effect of acid rain on building materials or the lost days of production from increased morbidity—and is already included in the current measures of the
output and prices of private goods. These latter costs need to be identified, but one must have an accounting system that makes sure that environmental effects are not double counted. While this is an illustration of air pollution effects, there exists, more generally, a complex interweaving of environmental changes, economic production, and consumption.
In summary, estimating the effect of environmental conditions on economic welfare, despite some overlap, will typically involve quite different types of analysis and estimation than that associated with the estimation of a COGI or COLI restricted to the domain of private goods. Most importantly, estimates of the effect of environmental changes on national output, income, and prices must be embedded in a consistent accounting framework that takes account of stocks and flows and outputs and prices and distinguishes the effects of the environment on various categories of human and economic activity.
In recent years the U.S. Department of Commerce’s Bureau of Economic Analysis has been engaged in a long-run effort to construct “satellite” accounts that would gradually extend, on an experimental basis, the domain of the government’s official measures of national output and income. A recent National Research Council (1999) report provides a highly useful discussion of the challenges, difficulties, and promises of this endeavor. Any research program of the BLS to develop experimental measures that substantially expand the domain of its various price indexes should be carried out in close concert and coordination with the BEA effort.
OTHER DOMAIN ISSUES
Taxes and Government Programs
Governments, particularly state and local governments, use sales taxes and similar levies to finance part of their expenditures. Such taxes tend to be passed along to consumers in the form of higher prices.8 As a consequence, an increase in the average rate of the sales tax throughout the country tends to raise, and a decrease to lower, the CPI. But as pointed out by Nordhaus (1997), the direct contribution of the education, roads, and other public goods to households’ standards of living is not captured by the CPI as an offset against the taxes. As a result, the CPI tends to overstate the rise in the cost of living when sales tax rates
are rising and vice versa. Nordhaus also argues that similar anomalies occur in a number of other instances: government environmental, health, and safety regulations require firms to incur costs that are passed on in higher prices for private goods, but the value of the associated benefits is not offset against them. Employers provide fringe benefits to workers, especially pensions and health insurance, whose costs are passed along in higher prices while, again, the benefits are ignored. Ideally, according to Nordhaus, one should estimate the value of the benefits provided and make the necessary adjustments. Since that is not feasible, he recalculates the CPI by subtracting the ratio of these items to aggregate consumption expenditures on the assumption that their cost to consumers is matched by the benefits they finance. He finds that, over the period 1960 to 1995, when both employer fringe benefits and pollution abatement costs were rising relative to aggregate consumption, a CPI stripped of all of those costs, which he calls an augmented COLI (ACOLI), would have risen by 0.4 percent a year slower than the published index (Nordhaus, 1997).
The panel believes that the issues raised by Nordhaus should be explored in the context of research on satellite accounts and the development of expanded and integrated national accounts that we discussed earlier. But we do not believe, nor does Nordhaus recommend, that adjustments of the kind outlined above be incorporated in the flagship CPI. Nor, we conclude, should they be incorporated in any measure used to index public benefits and the tax code.
In the case of sales taxes and mandated pollution control costs, we have discussed the limited or nonexistent mechanisms that exist for equating the marginal tax costs and benefits of public goods, especially at governmental levels higher than local communities. In addition, even if the assumption of cost-benefit equality were assumed, several difficulties remain. Many of the public goods furnished by government are intermediate goods used by business firms in the production of private goods whose benefits are reflected in lower production costs for private goods and therefore already included in the CPI. Also, if we subtract from the index an estimate of the price-raising effects attributable to mandated pollution control activities, must we not also undertake the formidable task of estimating, and adding to the index, the costs directly imposed on consumers from the “free” use of the environment by producing firms, which gave rise to the need for the environmental controls in the first place? It is difficult to justify the first step without the second.
The treatment in the CPI of goods provided in-kind to employees as fringe benefits raises a number of issues. There is a wide, even if not universal, consensus that, apart from some tax advantages and except during short-run transition periods, overall compensation packages are not affected by how they are parceled
out into money wages or fringe benefits.9 In the medium to long run, the greater part of changes in fringe benefits tend to be matched by offsetting changes in wages: workers trade money wages for in-kind benefits. The major categories of fringe benefits are employer contributions to pension plans and health insurance and their payments of social security and Medicare taxes on behalf of employees. Employer-paid pensions are a form of savings (with tax advantages). But the CPI is a “one-period” index. Viewed as a COGI, it measures the average change in the price of currently purchased consumption goods. Viewed as a COLI, it is similarly defined to include only the effect on living costs arising from changes in the prices of current goods.10 A change in pension saving (i.e., future consumption) ought not to be treated as a change in the price index or cost of living associated with current consumption.
There are several arguments for not adjusting the CPI to take account of the benefits furnished to workers through employers’ contributions to the federal social security program. While it has been accumulating surpluses in recent years, the combined social security and Medicare (Part A) system is still essentially a pay-as-you-go system. Because of coming demographic changes, benefit payments under current law will begin to rise faster than revenues under current tax rates. It is unlikely that future changes in the present value of taxes and benefits will be closely linked to each other. Payroll taxes are still to a significant degree a public levy to finance a collective and continuing set of transfers from one generation to another. And, in any event, even if they were fully funded, they would represent another form of “forced saving” whose future consumption benefits are not covered in a one-period CPI.
Employer-provided health insurance benefits raise more difficult questions. Unlike pension contributions and employer-paid social security taxes, they (principally) provide a form of current consumption, namely, medical care plus the provision of risk insurance.
Conceptually, it is not clear whether or not employer-paid health benefits should be considered as an item of private consumption and included in the medical care weights of the CPI. We argued above for excluding public goods from the domain of the index, including locally furnished public goods. And the existence and generosity of health insurance benefits in compensation payments are also to some extent the result of collective decisions and, as such, share some of the characteristics of local public goods. In both cases, decisions are made collectively, but people have some ability to choose individually by changing their residence or their job. By analogy, employer-paid health benefits should be treated like a local public good and excluded from the index.
However, it could be argued that health insurance fringe benefits are more nearly private goods, influenced by private choices, than are the goods provided by local government. In choosing among jobs, workers have the possibility of accepting a job with higher money wages but without health benefits, while making their own arrangements for purchasing medical care or self-insuring; or they can choose something in between. But there are many fewer options in the basket of local public goods: one cannot typically buy a bridge single-handedly. Moreover, alternative baskets of compensation items among which workers have to choose will typically contain many fewer goods than the alternative baskets of local public services.11
From a practical standpoint, most of the analyses of changes over time in real income are derived by deflating survey data on nominal money incomes by the CPI. In turn, that nominal income data typically exclude the value of employer-financed fringe benefits. As a consequence, in Chapter 6, which deals with the pricing of medical care, the panel pragmatically recommends that the official “flagship” CPI continue to exclude coverage of employer-paid health benefits. On the other hand, while the appropriate conceptual treatment of employer-paid fringe benefits is not an open-and-shut issue, the panel does recommend in Chapter 6 that the BLS also publish a supplemental medical care price index that, among other features, includes health care expenditures financed by employers.
CONCLUSION AND RECOMMENDATION
Drawing from the arguments presented in this chapter, the panel offers the following conclusion and recommendation:
Conclusion 3-1: Under either rubric, COGI or COLI, the domain of the “flagship” or official CPI and any subgroup indexes that are produced should be confined to the boundaries that the BLS has adopted—changes in the prices of private goods and services.
Recommendation 3-1: The BLS should not, on its own, conduct research aimed at producing a CPI with a substantially expanded domain. But we encourage BLS, jointly with other federal statistical agencies and particularly the Bureau of Economic Analysis (BEA), to undertake research aimed at producing on an experimental basis or in satellite accounts, more comprehensive measures of national output, income, and prices. These would take into account the effects of changes in outside conditions for which there may be at least some chance of reasonably measuring those effects—perhaps, for example, changes in the status of the natural environment.