Consumer Expenditure Survey Data: Uses By Bls

Eva E. Jacobs, U.S. Bureau of Labor Statistics
ABSTRACT - Consumer Expenditure Surveys conducted by the Bureau of Labor Statistics provide essential data for BLS statistical series as well as for general economic analysis and making purposes. Use of these data by BLS in the CPI, expenditure tables, and Family Budgets as well as the differences in concept among the series are described.
[ to cite ]:
Eva E. Jacobs (1981) ,"Consumer Expenditure Survey Data: Uses By Bls", in NA - Advances in Consumer Research Volume 08, eds. Kent B. Monroe, Ann Abor, MI : Association for Consumer Research, Pages: 251-253.

Advances in Consumer Research Volume 8, 1981      Pages 251-253

CONSUMER EXPENDITURE SURVEY DATA: USES BY BLS

Eva E. Jacobs, U.S. Bureau of Labor Statistics

ABSTRACT -

Consumer Expenditure Surveys conducted by the Bureau of Labor Statistics provide essential data for BLS statistical series as well as for general economic analysis and making purposes. Use of these data by BLS in the CPI, expenditure tables, and Family Budgets as well as the differences in concept among the series are described.

The long history of consumer expenditure surveys in the United States reflects the variety of needs which they were designed to serve, which have generally been associated with the need for cost of living information. The inquiries over time have been directed to the welfare of immigrants in 1870, the influence of the costs of living on the cost of production in connection with setting tariffs in 1888, the incorporation of the data as weights in the Consumer Price Index and the use of the data in developing Bureau of Labor Statistics (BLS) Family Budget cost estimates. BLS has always published tables, and more recently tapes, of expenditure, income and asset data to permit users to track changes in consumption patterns and to examine the relationship among economic variables for different types of families. The words "different type of families" cannot be overemphasized because it is thc provision of demographic characteristics associated with the economic data that is the major contribution of this survey. This is in contrast to the other widely used measure of consumption, the personal consumption expenditure component of the national accounts, which is an aggregate estimate of consumption of all families in the U.S. and does not tell us who is buying what.

The survey data are essential to several major BLS programs and I will describe how these uses affect the design of the survey.

Fur the Consumer Price index (CPI), the data are used for the derivation of the expenditure weights of the components and for the selection of items to be priced. Since the 1977 revision, the CPI has been constructed for two populations. One is the total urban population, about 80 of the total non-institutional population. Derivation of weights for this group requires expenditure data and geographical identification. The other is for the traditional concept of urban wage earner and clerical workers, about 50 percent of the population. The determination of this group requires data on geography, occupation, employment status, and income because the group is defined as limited to certain occupations based on the proportion of income derived from these occupations.

The particular definitions of the CPI components determine to some extent the kinds and number of questions that are asked in the survey. It is not enough to ask the respondent for his monthly mortgage payment although the response may well be reliable and might be sufficient for certain types of analyses of outlays. Homeownership cost in the CPI includes the net cost of houses purchased in the base period and the interest cost contracted for over half the term of the mortgage. In order to estimate these weights it is necessary to know the price of houses purchased and sold, the interest rate contracted for, number of years, down payment, etc.

Some components require additional detail because of experience with price movements. Telephone charges provide an example. Since intra-state and interstate prices may move differently, expenditures for long distance and local service are treated as separate items. It is not sufficient to know what the total telephone bill was for CPI purposes.

The requirements of the CPI change over time and the change in the 1978 procedure for selecting the detailed item to be priced for the CPI has allowed us to reduce sums of the detail collected. Prior to the revision, BLS pricing agents were given detailed descriptions of items to be priced. These items were selected from the detail collected in the survey. In the new disaggregation process the agents have more general descriptions to choose from. For example, previously the pricing agent was given a specific type of toaster to price - all agents attempted to price the same specification. Now the agent is provided with the item "small kitchen electrical appliances." Through the disaggregation process, which involves the relative sales of items in the store, the pricing agent selects the specific item, a toaster, in one outlet, a can opener in another. Since this procedure was introduced after the 1972-73 survey, the question was changed for the continuing survey from asking about a long list of items, all of which are frequently responded to with "no", to a question that asks about small kitchen appliances.

This may be satisfactory for the CPI but may prove to be a great disappointment for marketers of small electrical appliances. This is part of the constant conflict between the desire for more and more data and the desire to reduce respondent burden. Other requirements for the CPI have added new kinds of questions to the continuing survey. With the 1977 revision of the CPI, a separate point of purchase survey was introduced from which the sample of outlets which agents visit to collect prices for most components is drawn. For certain components, such as telephone and gas and electric utilities, the name of the company supplying the service is now obtained in the expenditure survey. The responses will form the basis for selecting outlets for these components in the future. However, point of purchase information may not be made available for individual outlets because of CPI confidentiality requirements.

The second major activity in the Bureau concerns use of the data for general economic analysis of consumption.  The Bureau has published a series of tables which shows selected characteristics, expenditures, income and assets and liabilities for all families and families classified by income, age, family type, region, etc. From these for example, one can compare the consumption patterns of the elderly with the general population, or the prime working age group. Or one can compare one earner husband-wife families with two earner families. For this use even mere data are required than for the CPI since the CPI does not cover income taxes, social security taxes, life insurance, cash gifts and contributions. All of these are included in a family's outlay and clearly influence the amount that will be spent for consumption items. Assets and liabilities come into play since borrowings or asset depletion are used to finance expenditures and loan repayments and additions to assets constitute additional outlays.

The methodology and format of the 1972-73 survey were markedly different from past surveys and these new features had a considerable impact on the process of developing the published tables. In earlier surveys, the household's income, total expenditure and assets data were collected through an annual recall procedure. The 1972-73 survey provided for two surveys -- one, consisting of a quarterly interview of one sample covering major expenditures and another, consisting of a separate sample keeping a two week diary from which frequently purchased components such as food detail and household supplies could be obtained. The new form had some inherent problems. One complication was that the time periods covered in the two surveys were not the same, although there was an 18 month overlap. That was finally resolved through a deflation procedure for the diary portion to provide an estimated two year period to match the interview. There was an additional difference in the time period covered by the income question. Since income is one of the most important variables in the tabulations, it was necessary to make an adjustment to make the income variable in the diary comparable with the interview. These adjustments having been made, the two surveys were integrated by groups of consumer units. For example, the food component of diary families of four persons were combined with interview families of four persons to form the four person family column in the integrated table showing expenditures by family size.

What was learned about changes in consumption shares from 1960-61 to 1972-737 The major movement was away from food and towards transportation for all segments of the population. The exception was that the food share increased at the lowest income level, probably as a result of the food stamp program. These were the expected results, since food shares have been declining since the 1930's, but the extent of the change in transportation was greater than expected. For all consumer units, total food declined as a share from 24 percent to 20 percent even though expenditures for food away from home were increasing. While for the total population this was partly the result of the decrease in family size, the decline was significant for all family sizes. The increase in the transportation share from 15 to 21 percent is associated with more cars per family as well as increased use. This trend held across the board for all types of families. The fact that the share for transportation now equals or exceeds that for food, is a graphic confirmation of the importance of the automobile to the American lifestyle.

In addition to published tables, we make available micro data tapes. However for the tapes the problem of the two surveys has not been solved -- separate tapes are produced for each survey. For many users who wish to analyze a particular component such as gasoline, there is no difficulty -- the interview tape is self-contained with all its associated characteristics. For researchers wishing to have a micro record of all expenditures, however, there is a problem, for matching individual diaries to individual interviews is much more complicated than matching by group. I don't know of anyone who has undertaken this activity as of now.

In BLS, expenditure data also have been used in the construction of the 4-person family and retired couple's budgets. These are hypothetical constructs that estimate the amount of income that is required to live at three specified levels of living -- lower, intermediate and higher. The 1960-61 survey was used as a basis for determining the long list of goods and services which constitute a significant portion of the budget estimates.

There has been a good deal of criticism of the methodology of the current budgets and at the very least, the market basket should be updated to reflect 1972-73 data.

A committee of experts was appointed to advise BLS on revising the budget program. The committee's report is now being reviewed. The major recommendation was to establish the level of adequacy as the median expenditure of the 4 person family as being a reflection of the prevailing American standard of living. If the recommendations in the report are adopted, the Bureau's new Continuing Consumer Expenditure Survey will provide the basis for establishing this standard in the future.

I have briefly described three data series published by BLS which measure levels of, differences in, or changes in what is popularly known as living costs, all of which use Consumer Expenditure Survey data. Furthermore, because all three series provide data for selected areas as well as the U.S., they are widely used to satisfy the great demand for information on differences in living costs among areas. Since the differences in what each is designed to measure are not always clear to users the data are frequently misinterpreted. I will briefly describe the differences in concept among the series.

First, the family budgets, because it is the series most widely used for interarea comparisons of living costs. These costs represent differences in cost of living among areas at a point in time. The budget market baskets are not the same in all areas -- there are variations in many of the components. These have not been changed since 1967, but the consumption costs are updated through time by the components of the CPI for the particular area. Therefore, only the change in prices since the base period affects an area's relative position in subsequent periods.

The CPI measures the change in prices over time in each area and is often considered to be a measure of the change in the cost of living over time. A more rapid increase in the CPI between any two periods of time in one area versus another does not mean a higher cost level in the second period but simply a more rapid escalation of prices in one area. However, if one area's CPI increases more rapidly than another over a long enough period, its living costs will eventually reach a higher level even if those costs started at a lower level.

The published tables from the Consumer Expenditure Survey for selected metropolitan areas describe what people actually spend in those areas. The differences in expenditures for an area result from different characteristics of the populations as to income, age, ratio of homeowners and renters, price levels and taste. Thus, we are measuring what people do in these areas rather than how living costs vary.

In addition to the basic difference in what these series are designed to measure, there are differences in definition of particular components, most notably in the treatment of homeownership cost. This component has recently been receiving a good deal of attention in connection with the CPI. While homeowner costs are a very large component of living costs, the concepts of measurement and comparability are difficult and there does not appear to be a consensus as to the "correct" way to measure such costs.

Since the family budgets are supposed to represent an established family, they include the cost of owning a home purchased six years earlier, so that mortgage payments and interest costs are calculated for a mortgage obtained six years earlier than the budget year. This clearly is not relevant to someone just moving into the area or a young family purchasing a home for the first time. The CPI, on the other hand, includes as a weight in its market basket the net purchase of home (in the base period) and the interest costs for half the term of the mortgage for those consumers entering into mortgage agreements in the reference period. These are then moved by changes in house prices and interest rates. These changes in the housing component clearly are not relevant to homeowners who purchased home years before. In our publication of data from the expenditure survey, we present still another concept of homeownership costs, this time in terms of the total household current account and balance sheet. The current costs for all homeowners for interest, taxes, insurance and maintenance, are included in current consumption but principal payments are considered repayment of a loan and are shown as a reduction in liabilities with other loan repayments. This concept of consumption costs might not suit someone who wanted to know how much they needed to earn to cover current outlays.

Users need also to be aware of the concepts and coverage of the expenditure survey data in making comparisons with other data sources. The comparison most frequently made is between the survey data (CEX) and Personal Consumption Expenditures (PCE) in the national accounts. Such a comparison should not be made without adjusting for differences in concept which affect the shares, the aggregates and the trends over time. Significant differences exist in the housing, automobile, health care and financial components. For example, rent in PCE is space rent, in CEX contract rent; homeowner costs in PCE are estimated from rental equivalent values, in CEX an outlay approach is shown; new automobiles are gross in PCE, net of trade-ins in CEX.

These cautionary words are intended not to discourage but to enhance understanding of the data and thereby help make analytical results more reliable. While the surveys provided essential information for government policy making and statistical activities from the start, marketing interest in the data has been more recent -- from the 1960's, when Life Magazine funded an extensive publication produced by the Conference Board based on the 1960-61 data. For the 1972-73 survey we have sold tapes to over 300 universities, individual companies and other research workers in addition to thousands of copies of the published bulletins. We anticipate even wider use of the results of the continuing survey than of the past sporadic surveys.

The Continuing Consumer Expenditure Survey is not in the field and we are looking forward to getting the first tapes from the Census Bureau. Our review and publication plans are being formulated now. We plan to release regular quarterly and annual publications based on the integrated diary and interview data. However, because of the infrequency with which individual items are purchased and the smaller sample size, the quarterly expenditures will be highly aggregated and the number of characteristics quite limited. On an annual basis, both the detail and the number of classes of classifying characteristics can be increased. Deriving reliable data may require aggregating two or more years of data. In addition to providing timely data on consumption patterns, the survey data will eventually be very useful in evaluating the need for and timing of revisions of the CPI market basket.

We also plan to make public use tapes but specifications of content or frequency are not yet available.

It is not until we review the data that we will be in a position to recommend changes in the questions or the methodology. We hope to introduce a research panel on which proposed major changes will be tested but that is for the future. The goal is to provide useful, reliable and timely information.

----------------------------------------