noted earlier, the concepts of consumption and expenditure are related but not identical. Consumption relates to the actual using up of goods and receipt of human services and expenditures relate to purchases of goods and services. Each concept has important analytical uses. BLS conducts extensive analyses of expenditure data. However, there is no comparable sustained effort with respect to consumption data.

A continuing developmental program for the measurement of consumption should be added to the existing program expenditure statistics in the Bureau of Labor Statistics. Provision for input from the principal agencies engaged in the production or analysis of consumption and expenditure data should be made through the establishment of a committee to advise on priorities.

The Federal Government provides the largest share of the production of data on income, wealth and consumption. On the other hand, analysis of these data is probably shared with more users outside than inside the government. Input from users and data producers outside the government is received through a broad range of means including formal advisory groups such as the Advisory Committees for the Censuses of Population and Housing, exchanges at professional meetings such as those of the American Statistical Association (ASA) and the American Economic Association (AEA), seminars such as those held as part of the SIPP developmental program, the employment of consultants, and informal exchanges. Continuation of such exchanges of views with the larger research community is essential to further progress in the area.

Data users should be consulted at an early stage in the development of major household surveys focusing on the collection of data on income, consumption or wealth. Early consultation will reveal the preference of users for specific data items to be collected jointly in the same survey vehicle and identify those which might be approximated through matching files from other surveys and administrative records.

Data concerning income, wealth and consumption have a high degree of importance for policy decisions. This use places a premium on timely release of high quality data. Timeliness can only be achieved on an ongoing basis through the establishment of a tightly controlled production process. Such a process can allow little variation if the objectives of uniform quality and timeliness are to be met. Concurrently, "the Federal data system" is called on to be sufficiently flexible to produce data tailored to the analysis of emerging policy needs. The conflicts between the production of ongoing statistical series and the need to

be responsive to emerging policy needs produce a considerable strain on the major statistical programs.

The resolution of these unavoidable conflicts among objectives will require that every major ongoing program for the collection of data concerning the economic status of households include some provision for flexibility with respect to data content and collection techniques. Assurance of timely production of statistics meeting high standards of quality can be met through the employment of a core set of data items on major continuing survey vehicles. Modules supplementing the core data set can be used to obtain data to meet some of the additional requirements for data related to emerging policy needs. Such supplementary modules would be limited by the need to avoid damaging the provision of data from the core data set. The requirements for developmental work could be met through work separate from the main survey process and through the occasional use of survey panels for additional enumeration after the completion of the regular data collection program.

Survey vehicles for the collection of data on the economic status of households should be sufficiently flexible to be able to respond to changing data requirements for policy analysis. Use of a relatively stable set of core data items should permit timely release of continuing series while additional modules are processed separately.

The major continuing surveys designed to monitor changes in the economic status of households should also include an option for an additional quarter at the end of the regular data collection period. This will permit data collection which cannot be accommodated during the regular survey period and the testing of collection methods.

Assuming that strong organizational focal points are established for statistics of income, wealth and consumption, there will be a need to ensure the best use of the survey instruments. In the area of statistics of economic well being, there are frequent needs for joint analysis of these data. Consequently, the design. and operation of major statistical programs should facilitate joint analysis. There is also a need to ensure the most efficient data collection approach consistent with requirements. Achievement of the required integration of effort among the units responsible for the production of income, wealth and consumption data should be met through the exercise of OFSPS's responsibility for the oversight of statistical policy.

The Office of Federal Statistical Policy and Standards should ensure coordination of statistical programs relating to income, wealth, and consumption

through membership on the interagency policy advisory groups in these areas and through the Interagency Committee on Income Distribution.

The techniques employed to provide data concerning the economic status of households will involve the integration of separate surveys, and of survey data with data obtained from administrative records. The need for integration of different data bases derives from considerations of efficiency, respondent burden, accuracy and completeness of data. For example, the use of tax records in conjunction with household survey data can provide more accurate data at the upper end of the income scale. At the lower end records obtained in the course of eligibility determination for programs such as the Supplementary Security Income and Aid to Families with Dependent Children programs will improve the accuracy of income estimates. Under favorable circumstances, the merging of survey and administrative record sets provide the improved accuracy of administrative data with the flexibility in data content obtainable in household surveys.

At best, matching of administrative and survey data files is a difficult procedure which should be employed selectively. However, in order for file matching to be possible in those cases where it is the preferred approach, it is necessary to include provision for matching in the basic survey design. It would also be helpful to modify administrative data records in anticipation of matching requirements to the extent that this is feasible. The design requirements for statistical matching are particularly stringent in that there frequently is a need to have many identical data items for purposes of matching.

Use of existing administrative and survey files should be made to improve the accuracy of estimates and to provide a more complete assessment of economic well being than can be provided through a single data set. Survey and administrative files should be designed to facilitate matching which can be accommodated within limitations required to protect confidentiality.

Data users with increasing frequency require data relating income, wealth or consumption data on the level of the individual household. Yet to a large extent available data sources do not provide for the joint collection of these data, or if they do, the data which is an adjunct to the main purpose of the survey is either not directly comparable or it is too summary. For example, the Current Population Survey (CPS) which is currently the best source of household survey data provides no link with the Consumer Expenditure Survey through the inclusion of a set of expenditure questions. (However, it should be noted

that the Current Consumer Expenditure Survey will employ standard CPS questions on income.)

For many uses, the inclusion of a basic set of income, wealth or consumption questions on surveys devoted to detailed data collection on another aspect of economic well being would be sufficient. It would also permit more direct comparisons with results obtained from the more detailed surveys in other areas. The use of abbreviated data sets are closely related to file matching which would be greatly improved through the use of such sets.

Abbreviated data sets should be employed to improve the description of economic status and to improve the results of file matches. For example, a summary set of questions from the Current Consumer Expenditure Survey should be seriously considered for inclusion in the Survey of Income and Program Participation.

Income

The quality of current data on money income has gradually improved over the last decade. The improvement is particularly evident with respect to the annual money income series produced by the Bureau of the Census. Identification of source of income has become more detailed, improved methods for imputation of missing data have been developed and a more thorough analysis has been conducted concerning the extent to which money income falls short of aggregate totals developed by the Bureau of Economic Analysis.

However, serious problems remain. In the area of income estimation, income other than wage and salaries is seriously underreported with as much as 71% of interest income, 64% of workman's compensation and 57% of farm income not reported for 1975. Another shortcoming is the exclusion of almost all forms of income in kind. Such income is particularly important for the lower income families. There is also believed to be a significant amount of in-kind income received by higher income families in the form of employee paid health care, pensions, use of company property, and so forth. Families at all income levels received significant income in kind from occupancy of their own homes.

A third area of difficulty concerns the period for which household income is collected. The current series measures annual income, but income is not always received uniformly throughout the year. Eligibility for income maintenance programs is

2U.S. Department of Commerce, Bureau of the Census, "Consumer Income," Current Population Reports, Series P-60, No. 105, June 1977, p. 277.

frequently tied to income received during a fraction of a year. Moreover, the annual income data collected for a household in the Current Population Survey (CPS) does not always correspond to the household composition or income of the unit during the preceding year.

Improvement of income estimates is being approached through improved survey data collection including selective estimates of in-kind income, the use of administrative records and through improved estimating procedures for imputing income not reported. The development effort for the Survey of Income and Program Participation (SIPP) is the major current activity for the improvement of the measurement of income.

The four-year SIPP development effort sponsored by DHEW will involve all of the enumerated approaches. A series of tests will be conducted to attempt to determine the effect of improved methods of asking questions. Comparisons will be made with administrative records to check the survey results, fill in missing data, and correct survey answers. However, according to present plans the improved estimates anticipated from the SIPP will not be available before 1981. In the interim there will be a need for improved estimates of income for direct input into policy formation.

As the major survey from which estimates of the distribution of household income are currently derived, the CPS will be the principal vehicle available to fill the need for improved income data during the interim. Preliminary testing in a CPS test panel of selected improvements in income measurement developed for the SIPP could be employed to ensure that modifications in the annual income questions would not significantly affect the measurement of employment or unemployment for the month of March, which is the month annual income is collected. Early introduction of improved measures of annual income into the CPS program would facilitate comparisons with data from the SIPP program when it becomes operational.

Selected improvements in the measurement of income which are successfully tested for the Survey of Income and Program Participation should be incorporated into the Current Population Survey. The integrity of the employment and unemployment estimates derived from the CPS should be protected by use of a CPS test panel to determine the effect of additional questions.

Another way in which income estimates have improved is through the personal income estimates prepared by the Bureau of Economic Analysis

(BEA). Besides being of value in the preparation of improved income size estimates in the economic accounts, these BEA estimates provide control totals for the data collected in the CPS. The estimates of personal income include the distribution of food stamps, Medicare, renter versus owner occupied housing, food and lodging furnished employees, food and fuel produced and consumed on farms, selected categories of interest and imputed interest from life insurance policies. These estimates are based on administrative records of government programs, available survey data and other relevant information. They also represent a matching of IRS tax return and CPS household data as a method of cross-checking similar information reported in the two sources and for deriving estimating factors in allocating different sources of income by income size and demographic (age, sex, race, etc.) groups.

The Bureau of Economic Analysis measures of personal income by size and class should be expanded initially to include income after the payment of personal income and social insurance taxes: subsequently they should be broadened further to include disposable personal income, consumer expenditures, and personal savings.

The introduction of in-kind income into statistics of household income permits a more complete assessment of household well being. However, the identification of what should be counted as in-kind income is a difficult one. For example, choices need to be made as to whether income in kind should be defined to include the monetary value of property rights, the receipt of gifts, the monetary equivalent of leisure time, and child care expenses. The inclusion or exclusion of these and other forms of income could materially affect comparisons of income distributions.

The valuation of in-kind income is at least as difficult as deciding what should be counted and will itself affect the inclusion or exclusion of particular types of income. The most general problem of valuation is whether in-kind transfer income should be valued at the cost to the donor or at its worth to the recipient. The differences between donor and recipient valuations of in-kind income can be significant. One analyst described the valuation problem as follows:

.. It is usually assumed that a household receiving income in the form of cash will spend that income on a bundle of goods (including savings) in the manner that is most desirable from its own perspective. In technical terms, it is assumed to maximize utility given a budget or income constraint. In-kind income, however, is provided in

the form of a particular good or set of goods. After receipt of the in-kind income, then, the household's bundle of goods may include more (or in some cases less) of the particular good than it would choose if given cash. Because the household is not permitted to exercise complete freedom in its choice of a bundle of goods, that bundle with the in-kind component may be worth less to the household than if it had been given cash. For example, a family that is temporarily housed for a month in a hotel costing the government $50 a day can in no case be said to have $1500 of income for the month; clearly, if provided with $1500 in cash, the family would not have chosen to spend all of the $1500 on housing. . . .

3

Further discussion of the problem of defining and valuing in-kind income is given in Technical Paper VII of The Measure of Poverty and in a study entitled "The Cash Equivalent of In-kind Income Study" which was sponsored by the SIPP development program.'

Preparing distributions of in-kind income is both important and difficult to accomplish. The many conceptual and measurement problems make it unlikely that general agreement can be reached concerning the best approaches to take. Under these circumstances, Federal efforts should be directed to providing statistics on the major sources of in-kind income, including data which will enable analysts to choose alternative ways of valuing individual sources. The substantial exploratory work being conducted as part of the SIPP development program should be continued as an integral part of the ongoing operation of the SIPP. Advice on priorities and practical approaches to the problem of measurement should be provided by producers and users of income statistics.

An interagency task force should identify practical methodologies for measuring the distribution of income in kind. The study should evaluate the alternative methodologies of collecting direct survey data and using indirect estimating procedures for selected components.

Wealth

In recent years, the need for information on wealth has grown with respect to Federal policies in areas

'Janice Peskin, Technical Paper VII, "In-Kind Income and the Measurement of Poverty," of The Measure of Poverty, U.S. Department of Health, Education and Welfare, p. 9.

*Ibid.

'Unpublished, Office of the Assistant Secretary for Planning and Evaluation, Department of Health, Education, and Welfare, 1978.

such as the distributional impact of taxes, the evolution of the Social Security trust fund, pension reform, income maintenance programs, and personal savings. Each of these areas has important ramifications. For example, personal savings affect investment behavior and capital formation, two factors which are crucial to the general performance of the Nation's economy. Most of the advances that have been made in statistics concerning the economic status of households have been directed to the improvement of the measurement of income and expenditures. Some improvements in the provisions of current data on household wealth have been in support of individual program areas such as housing. For example, in the last few years the HUDsponsored Annual Housing Survey has provided current estimates of the value of owned homes.

In spite of data improvements in response to the needs of individual programs, current data are not available from which an assessment can be made of the distribution of household wealth and its contribution to well being. Much that is known about the distribution of personal wealth is due to the major effort embodied in the Survey of Financial Characteristics of Consumers sponsored by the Federal Reserve Board in 1962, a survey which was drawn from the Census sampling frame and IRS tax records. In the absence of continuing government sponsorship, this survey has been updated on a number of occasions by the University of Michigan. However, it has not been possible to conduct a survey on the scale of the 1962 efforts.

A brief summary of Federal statistical programs will reveal the paucity and unevenness of data concerning the distribution of wealth among households. With respect to physical wealth, ownership of real estate is the major component of wealth. Data giving the owner's estimate of current value of his own residence are available yearly from the Annual Housing Survey. Other information obtained from this survey permits cross-tabulations of value with (1) income and other household characteristics, (2) condition of the housing unit, (3) vacancy status, and (4) many other factors. These data are also available every fourth year for selected Standard Metropolitan Statistical Areas. On a periodic basis, the decennial census program provides estimates of housing value for all communities in the country and for blocks and census tracts within communities. Cross-tabulations of value data are available by many of the other characteristics collected in a decennial census.

The other principal source of estimates of the value of housing is the Consumer Expenditure Survey

conducted by the Bureau of Labor Statistics every ten years for the purpose of updating the Consumer Price Index. The most recent survey, which was conducted in 1972-73, provides extensive detail permitting the calculation of the owners' equity in his home. In addition, the survey provides data on the holdings of other property, including second homes and real estate used for business purposes. Since extensive data on many aspects of consumption, income, liabilities, and other wealth components are collected, the survey provides data for detailed analysis of household wealth. These data are available for the national level and to a lesser extent for regions.

Other existing data on the value of housing relate to aggregate estimates or provide insight into the current operation of the housing market. To summarize, detailed data on the national level of the value of housing is available annually, but coverage of all real estate owned by household members and data from which equity can be calculated are obtained only at ten-year intervals.

Motor vehicles are the second largest component of physical household wealth. There are no generally available current data on the distribution of household ownership of this asset. The Consumer Expenditure Survey provides an estimate of the current value of new and used vehicles owned for personal use. One major omission, which is important for some analyses, is the lack of data concerning personal use of vehicles registered by businesses. The 1980 decennial census will contain a question on the number of vehicles available for the personal use of members of the household, which will be of value for transportation analyses as well as for analyses of wealth.

With respect to major household appliances and other consumer durables, the Consumer Expenditure survey obtains extensive data which make possible estimates of the current market value of appliances purchased within specified recent periods. Data on clothing are limited to recent purchases and no attempt has been made to obtain estimates of personal wealth holdings such as jewelry, antiques, art collections and other miscellaneous wealth components.

Data on the distribution of financial wealth are limited to the Consumer Expenditure Survey and the Estate Tax Returns of the Internal Revenue Service. The Consumer Expenditure Survey contains data on the ownership of assets in bank accounts, United States Savings Bonds, and life insurance policies. It involves an overall estimate of ownership of corporate stocks and bonds, but no detail is obtained. The Consumer Expenditure Survey provides no data

on rights to private pension or retirement funds. The Estate Tax Returns have provided insight concerning the share holdings of top wealth holders in national wealth. However, as revisions to the estate tax become effective in 1978 and 1979, the number of persons required to file returns will be reduced to a small fraction of the present level.

Some improvements in the measurement of household wealth are being tried in the development program for the SIPP. To the extent feasible, those which show promise should be incorporated into continuing operational programs. Provision should also be made for the inclusion of measures of wealth in the Current Consumer Expenditure Survey under development at the Bureau of Labor Statistics. While a significant amount of work is planned, progress in the measurement of wealth would benefit greatly from the establishment of a lead agency which would provide a focal point as described in the beginning of the section of this chapter on adequacy.

The extent of need and the complexity of measurement for data on the distribution of household wealth would easily outrun resources available for statistics. Consequently, it is important that a study of agency requirements for wealth data be conducted to provide guidance in determining priorities in this area.

The Interagency Committee on Income Distribution should prepare a report on Federal requirements for data on the distribution of household wealth. The report should provide a basis for the development of wealth statistics pending the establishment of a continuing developmental program within the Bureau of Economic Analysis.

Consumption

Data on consumption fill an essential role in the assessment of the economic well being of households in that consumption measures the actual using up of goods and services to meet human needs. Data on current consumption thus explain the level of living supported by any given combination of household income and wealth. Data on consumption are of special significance in the area of nutrition in which detailed information on individual intake is needed to assess the role of human nutrition in maintaining good health.

Data on household consumption are difficult to measure. If consumption is measured on a current basis, the act of measurement is likely to have a significant effect on the measure itself. If the household is asked to recall consumption in some recent period, errors of memory will produce