Designing Nonmarket Accounts for the United States: Interim Report (2003)

Researchers and policy makers have long recognized the importance of the national economic accounts as a useful data system. Professional economists, among others, have also been well aware of the accounts’ inadequacies in providing meaningful measures of economic and social performance. Additionally, it has been recognized that improved accounts are essential for increasing the accuracy of productivity, price and real output statistics—particularly in difficult-to-measure areas such as health and education.

These observations notwithstanding, large and well-funded research efforts to address inadequacies in the national accounts have been relatively few. One such effort was the Measurement of Economic and Social Performance project at the National Bureau of Economic Research, from 1972 to 1977, funded by the National Science Foundation. First under the direction of F. Thomas Juster and, later, Richard Ruggles, the project carried out research on a number of specific limitations in the conventional economic accounting system.¹ Topics studied included the system’s lack of usefulness for describing income distribution among households and between geographical regions; its failure to account for many production and investment activities taking place in households; its treatment of many “intermediate” and investment activities as if they were “final” consumption activities; and its overly simplistic accounting for the complex fabric of federal, state, and local governmental activities.

Another area inadequately handled in the national accounts involves the services generated by the natural environment—both the negative contribution of pollution and environmental deterioration, and the positive contribution of these services to well-being when conditions are improved. Air and water quality, for example, are linked to the productive capacity of a society, broadly defined, but the national accounts do not adequately measure the value of investment or disinvestment in these assets. Expenditures to improve air or water quality are counted as contributing to national output (provided such outlays are counted as final demand expenditures), but they are valued at the cost of the inputs used rather than as the value of the output produced. And, if a business increases its production, the value of that production is measured, but there is no offset in the accounts to reflect any adverse effect that the production process may have on nearby streams or on ambient air quality.

Environmental accounts have been developed at the Bureau of Economic Analysis (BEA) and elsewhere in order to better reflect, relative to the National Income and Product Accounts (NIPAs), interactions between the market economy and the natural environment. And such efforts as the System of Integrated Environment and Economic Accounting (SEEA) and the Environmental and Natural Resources Accounting Project (ENRAP) have in fact been used—by the Environmental Protection Agency, the U.S. Department of Agriculture, the state of California, and others—to develop data for various estimation and policy development purposes. Even developing countries as Indonesia and the Philippines have used environmental accounts to help set policy priorities. In these cases, the interest was less with the accounts per se, and more with obtaining comprehensive data. The accounting system permitted the generation of consistent data sets at relatively low costs.

More generally, the omission of many nonmarket activities from the national accounts may significantly distort policymakers’ sense of economic trends. A fuller accounting of national production might lead, for example, to different conclusions regarding the level of output today relative to some earlier period, or in the United States compared with another nation. This deficiency—that the NIPAs fail to consider the full complement of inputs and outputs, specifically those that are non-marketed—would be less important if marketed inputs and outputs were independent of non-marketed inputs and outputs, but they are not.

To take one frequently cited example, failing to account for the output produced within households may lead to misleading comparisons of economy-wide production, as conventionally measured. The female labor force participation rate in the United States has grown enormously since the early part of the 20th century. To the extent that the entry of women into paid employment has reduced the effort women devote to household production, the long-term trend in output, as measured by gross domestic product (GDP), may exaggerate the true growth in national output.² Similarly, the lesser relative importance of home production in the United States as compared to many developing countries may exaggerate its national output relative to theirs. Perhaps less well

recognized, there may be similar problems with the measurement of national output over the business cycle. If people who lose their jobs during cyclical downturns take advantage of their absence from paid employment to increase the effort they devote to home production, the short-term decline in national output may be somewhat dampened relative to that measured by GDP.

Knowing more about the level and distribution of nonmarket activity may also change perceptions of the extent of economic inequality among U.S. households and how that has changed over time. This, in turn, may affect where welfare and poverty lines are drawn (Michael, 1996). Policy interventions that have positive effects on marketed outputs could have offsetting negative effects on nonmarketed outputs which, if they had been properly anticipated on the basis of more complete historical data, could have affected policy choices.

The conventional accounts also neglect significant amounts of nonmarket investment, such as the increases in human capital that result from education and job experience. There are a number of reasons why producing satellite accounts for human capital and formal education would be of use for the research and policy communities. First, because human capital, and particularly K-12 education, is such a large component of the capital stock, separate human capital accounts would add information that would be helpful for interpreting investment, capital, and ultimately economic growth as measured by the traditional accounts. Second, the education sector is large and important in its own right. Spending on education accounts for a substantial share of both state and local government budgets, and private educational services are a $30 billion dollar a year industry. Understanding trends in output and productivity growth in the education sector therefore is of interest. Third, the opportunity cost of students’ time is an important aspect of investment in human capital, and is missed in the traditional national income accounts. The value of time students spend in school is a nonmarket input, the value of which likely dwarfs the expenditures on marketed inputs associated with the educational process. Further, the accounts do not measure the value of additions to the stock of human capital produced by educational investments, even though that value is probably larger than the combined value of all measured investments in physical capital.

It is also worth keeping in mind that the national economic accounts serve multiple purposes. For example, health accounts could be designed to measure health improvements, in which case they would need to keep track of a wide range of factors, including things like diet and the environment. Optimally, expenditures and outcomes would be tracked so that changes in well-being associated with different actions could be monitored; in turn, expenditures (both private and public) could be effectively managed to achieve desired outcomes. Additionally, some version of health sector accounts needs to be designed in the best way to deflate expenditures for productivity measurement and real GDP. Currently, it is difficult to even determine if medical prices are rising or falling, since output quality is poorly monitored for the sector. How various uses of the accounts are weighted, in terms of importance, will affect decisions about what types of account to construct.

An overarching question for nonmarket account design is scope—where in the range of economic-related activities to draw the border of inclusion. This question raises

others: Should the expanded accounts be constructed as satellites or integrated with the core accounts? Should the objective be primarily a welfare measure or an output measure (and to what extent are the two really detachable)? We return to these questions below, but first address the question of which goods—near market, public, intangible assets, etc.—are to be included in an extended (initially experimental) set of accounts.

The scope and coverage of national accounting systems has been changing since the initial efforts by William Petty to estimate England’s national income in 1665.³ By modern standards, Petty’s accounts, albeit based on fragmentary data, were fairly wide in scope, covering, besides purchases in the market, imputed values for household production (Kendrick: 285). Far more narrow were the concepts of the French physiocrats, who believed that only agriculture produced a true net product, or the concepts of Adam Smith and Karl Marx, who believed that the only measure of a country’s productive capacity was in its ability to produce material goods, that is, excluding services.

However, beginning with the writings of Alfred Marshall (1920) and A. C. Pigou (1920), the modern trend, in terms of a conceptual objective, has been to widen the coverage of national accounts to include all activities that generate “utility” or welfare, including activities that are not reflected in market transactions. Yet it is a matter of some dispute as to how far this trend should go. Issues of scope—coverage as well as classification—have been at the center of much of the modern debate concerning the adequacy of national accounting systems. In the remainder of this section, we offer some preliminary thoughts on where the boundary should be, beginning with a discussion of what is already included in the core accounts.

Pigou wrote that national accounts should include elements that reflect economic welfare and that can “be brought directly or indirectly into relation with the measuring rod of money” (Pigou, 1920, p. 11). He emphasized that the word “can” might mean anything from “can easily” to “can with mild straining” to “can with violent straining.” National accounting practice in most countries leans far more toward those elements that “can easily” be measured in money terms than those that can be measured only with “violent straining.”

For a variety of historical reasons, partly philosophical but more fundamentally practical, the national accounts produced by BEA generally exclude activities that do not involve a market transaction or produce a marketed output. There are exceptions—the most quantitatively notable is the imputation for the rental value of owner-occupied housing. That this imputation is based on assumptions that are approximately as crude as those for, say, valuing the time spent cleaning a house at the price a cleaning service would charge suggests that the delineation is not purely the byproduct of practical considerations.

Imputations are made for other nonpriced, nonmarketed entities. In the U.S. system of accounts, these include wages and salaries paid in kind, food and fuel consumed on farms, and the services provided by banks, insurance companies, and other

financial intermediaries that are not reflected in explicit service charges.⁴ In addition to these items, which are (except for owner-occupied housing) relatively small in magnitude in the accounts of industrialized nations but may be large in the accounts of less developed economies, imputations are made for the services of capital and of the government. For example, in the U.S. accounts, an explicit adjustment is made to reported depreciation so that capital consumption allowances more closely reflect true economic depreciation. And national accounting systems have historically valued the services of government in terms of the costs of providing those services, though this is changing.⁵

These estimates clearly involve a process of valuing services that are not traded in well-defined markets. While rental markets do exist for many capital items, the value of the services of owned capital is not directly observable. Similarly, the value of freely provided governmental services to those consuming them is also not directly observable, although the cost of providing the services is.

One key characteristic of the nonmarketed items that are covered in conventional accounting systems is that their consumption is very closely related to the sales and purchases of marketed goods and services, making the estimation reasonably straightforward. For some nonmarket items, the imputation process is far more difficult, although these distinctions are a matter of degree. If “imputations procedures” always refers to data that are not directly observable, then it is clearly the case that the development of nearly all national accounting data, whether market, near market, or nonmarket, involves some degree of imputation.

It is impossible to lay out a case for the appropriate scope of nonmarket accounts without specifying the measurement objective. Perhaps the most fundamental issue is whether the primary intent of the accounts is to provide a measure of national output or national welfare. A defense of either goal requires dealing with the thorny nature of how these two concepts—output and welfare—interrelate. The two cannot be disentangled completely. For example, defining “the output of interest” surely implies, in some sense, “the output that contributes to well-being.” However, some goods that are welfare producing are not considered “output” in any conventional (accounting) way.

Some of the nonmarket activities currently unaccounted for can be categorized as output in a fairly narrowly defined way; for others, it is conceptually fuzzier. It seems clear that the value of household production should be included in an expanded account, even if the goal is limited to fuller accounting of national output. In contrast, tracking changes in the amount of time available for leisure or sleep may not be appropriate in such an account, though these factors certainly affect well-being (and, at extremes, productivity). These complications notwithstanding, a set of experimental accounts could

be expanded under either guiding principle, although the expansion would surely be much broader and ambitious if the goal is to provide an accurate indicator of welfare trends.⁶

The idea of an augmented account working in an integrated way to track changes in welfare implies tinkering with the core accounts (at least as they would be used in the augmented version). Furthermore, it is not just a matter of where to draw the line; in some cases, the phrasing of the accounting objective may lead to opposite practices. For example, in measuring “output,” one would likely want to include the value of parents’ time transporting kids to school and other activities. The market value of hiring a driver might be used to place a value on this time (though there are alternatives, discussed below). As the amount of time driving goes up, so to does the value of this component of household production. Yet, as driving time increases, parents’ welfare may actually decrease, as time is taken away from leisure and other utility-generating activities.

If the goal is to measure change in welfare as accurately as possible, the “intermediate” and “final” categorizations of many goods and services would need to be reassessed—as is done, for example, in the system of Nordhaus and Tobin (1972), in which certain “defensive expenditures” are subtracted from output totals. It might make sense to subtract such expenditures as crime prevention (e.g., home security or police protection) or even pollution cleanup because there is an underlying utility-based rationale for doing so. Often these expenditures are not intended to increase welfare; instead, they are intended to minimize welfare-damaging activities (activities that are not adequately measured in economic accounts).

Additionally, some goods that are welfare producing but not considered “output” in an accounting sense may interact with other nonmarket activities that are. For instance, leisure interacts with education—education leads to higher productivity and income; however, it may also be pursued because it yields more future leisure or to a more fulfilling kind of leisure. All this means that the distinction between an output account and a welfare account is not clear-cut and may, at best, provide only a guiding principle.

Alternatively, the accounting objective might be to measure income and output as fully as possible, with each account entry reflecting some aspect of economic well-being (see Nordhaus, 2002, p. 2). This assertion does not imply that the accounts or some portion thereof should necessarily “measure” human welfare. The fact that a nonmarket, welfare-affecting entity such as clean air may lack a market price and thus lack a natural money measure does not mean that it is conceptually dissimilar from more marketed elements that reflect economic welfare.

Restricting the coverage of the accounts in this manner does mean the exclusion of many factors that influence general welfare. For example, an account so defined will ignore changes in societal norms and values or trends in the connectedness of communities, except to the extent that such things affect elements that are or can be measured in money terms. In addition, while the destruction of environmental capital, such as natural forests or a particular plant species can be accounted for, the accounting will reflect only those environmental services perceived lost to present generations. In

principle, those in present generations can be queried or their behavior observed in ways that would permit placing a monetary value on such losses. Because the future cannot be observed, the value of lost opportunities to future generations would not be accounted for unless such future losses are also perceived to be losses to present generations.

One operational approach might be to make an item eligible for inclusion if it is possible to conceive of some procedure that would reveal the monetary value current generations place on it. Under this definition, a good like leisure would probably be considered in scope. Even such a seemingly noneconomic entity as “family disintegration” could enter the accounting structure if one could conceive of a method that revealed what society would be willing to pay to prevent such disintegration. Since our abilities to conceive of such procedures have increased in the years since Becker’s seminal work on marriage and the family (e.g., Becker 1973, 1974), the boundary between those entries that are proper candidates for inclusion or exclusion as determined by that rule will undoubtedly change over time.

In contrast, one could preserve more of the “output-based” methodology. Such a decision would probably eliminate the leisure example above from the scope of coverage (either on its own or as part of a household account). The position would be justified by stating that the measurement goal should be a full accounting of national output, market and nonmarket, and not of happiness or well-being, the latter being a much more futuristic goal.

The output-emphasized position can be defended along practical measurement lines as well. Such activities as household production of education or cleaning services have market analogues and are (relative to leisure) more closely aligned with what is currently viewed as output in the accounts. Because improving output (and corresponding input) measures is a prerequisite to any of these visions for an expanded set of accounts, this is where the panel is directing most of its energies. However, the panel’s final report will also assess prospects for progress for accounts characterized by a range of measurement objectives.

In sum, even if the goals for a set of augmented accounts are limited primarily to measurement of output—with home production probably being the clearest example— there is still plenty of work to do. Also, the range of activities that seem most amenable to convincing measurement involve input and output valuation methods that resemble those used in the current accounts. Moving toward a welfare measure implies harder work, especially in such areas as health and the environment, and also further deconstruction of current components of the National Income and Product Accounts (as used for the expanded account).

The panel’s final report will discuss a number of areas that should be considered for inclusion in a set of augmented accounts that would reflect more fully and more directly a range of nonmarket activities, including measurement of:

• household production and effect on the stock of human capital;

• investments in education and the resulting stock of human capital;

• investments in health and the resulting stock of health capital;

• selected activities of the nonprofit and government sectors;

• investments in the flow of new knowledge and the productive capacity thereby created; and

• social and environmental assets.

This identification of areas in which supplemental accounts might usefully be developed reflects several considerations. Each is associated with output that is both quantitatively significant and likely to have changed in relative importance over time.⁷ The panel is also studying areas in which work is ongoing and data are developing and that involve conceptual issues relevant to nonmarket accounting more generally. The areas covered are also broadly illustrative of the range of issues one might expect to encounter in the creation of a complete set of nonmarket accounts. The panel is also working on cross-cutting issues that do not fall neatly into these sector-oriented chapters: Time use, for example, is a key concept tied to data issues on which the panel is focusing; specific activities, such as volunteerism, also apply to more than one of the identified sectors.

This list of areas is fairly comprehensive, but at least one clear candidate for inclusion has been left off. Some efforts to measure the value of nonmarket activities have assigned a value to leisure time (e.g. Nordhaus and Tobin). Though it is discussed in this report, the panel has tentatively decided not make leisure a priority for a number of reasons. The new American Time Use Survey (ATUS) will provide useful estimates of the amount of time individuals devote to leisure, and changes in those amounts would admittedly be very important in a measure of well-being (and it is probably possible to assign a value to it based on what individuals are willing to pay for it). Leisure, however, does not contribute to the value of goods and services produced; unlike time withdrawn from nonmarket production, time withdrawn from leisure is unlikely to entail foregone output, or to require the purchase of a substitute. The panel thinks it pragmatic to first consider expansion for experimental accounts in areas more closely resembling those in the core accounts. The panel would like to prioritize work that addresses deficiencies in “output” measurement, more narrowly defined, which implies setting a boundary of areas where prices in principle can be derived from market comparisons.⁸

As noted on the list, the panel considers the nonprofit and government sectors to be areas of concern, given their size in the economy. An essential characteristic of these sectors is the extent to which they produce public-good outputs. Most of these outputs can be cross-classified in other sectors, such as health and education; as such, it is not clear what, if any, parts of these activities should be partitioned off in a separately designated nonprofit account.⁹ Additionally, while the panel would categorize accounting for natural resources and the environment as a top priority, it will likely spend less time on it as many of the key issues were addressed in Nature’s Numbers (National Research Council, 1999).

The connectedness of many of the nonmarket “sectors” makes clear-cut classification of activities impossible. For instance, there is blurriness between household production and other sectors. Additions to the stock of human capital may flow not only from investment that occurs in the formal educational sector, but also from investments that occur in the home (e.g., studying or reading to a child) and thus might be considered a form of home production. Similarly, improved health may result from better medical care, better education that contributes to better individual decisions about diet and exercise, or improved air and water quality, among other possible contributing factors. The full set of inputs to improved health outcomes cannot be precisely identified, and some of these inputs also may contribute to other desirable outputs.

The panel sees no realistic alternative to considering the different areas of nonmarket activity separately, but nonetheless recognizes the need to delineate the interactions and complementarities among these different areas as the development of supplemental nonmarket accounts proceeds. It may make sense to organize these accounts based, in part, on the methodological complexity. For example, the panel has discussed the idea of a limited scope household account, which would only include such activities as consumer services (e.g., laundry or meal preparation, basic child care) that result in output that can be relatively easily compared to market-based alternatives. This approach would leave more difficult-to-value activities (e.g., studying or exercising) that might be classified as investment to be dealt with elsewhere, perhaps in the human capital or certain health accounts. In these areas, constructing an account requires decomposing a much more complex production function, and valuation of both inputs and outputs is harder. For example, exercise may be an input in human capital formation (and health), but it may also have a consumption component. Degree-of-difficulty considerations might also suggest a more narrowly defined education account, limited to time spent on formal education tasks, with many inputs taken as given.

Measurement in many of these areas, even the conceptually simpler ones, will be difficult. An important part of the panel’s work involves evaluating promising measurement approaches and determining if they could be implemented, wholly or in

part, using existing or planned data sources. Developing recommendations about new data that might be helpful and how they might be collected is also an important part of the panel’s work.