Overview of the Consumer Expenditure Surveys

N. Gail Hoff, Bureau of the Census
[ to cite ]:
N. Gail Hoff (1981) ,"Overview of the Consumer Expenditure Surveys", in NA - Advances in Consumer Research Volume 08, eds. Kent B. Monroe, Ann Abor, MI : Association for Consumer Research, Pages: 245-250.

Advances in Consumer Research Volume 8, 1981      Pages 245-250

OVERVIEW OF THE CONSUMER EXPENDITURE SURVEYS

N. Gail Hoff, Bureau of the Census

[This paper cannot be copyrighted. It may be reproduced.]

GENERAL

The United States Government has recently begun a statistical program to provide current data on consumer expenditures of the population. Until now, there have been periodic surveys but no continuing program to provide such data on a regular basis. This gap in the government's statistical program has hindered many important analyses of the interrelationships of income and expenditures for various population groups, has created a lack of means to keep the Nation's Consumer Price Index more current, and has caused a void in studying and measuring the effects of a broad range of important policies and programs on the population such is inflation, taxes, energy, housing, medical care, and so on. After many unsuccessful attempts in the past, the Bureau of Labor Statistics (BLS) received appropriations in fiscal year 1979 to develop and conduct an ongoing Current Consumer Expenditures Surveys Program. The project has been given high priority by the Secretary of Labor. The program consists of three interrelated surveys; the two expenditure surveys are designed to provide the expenditure data for revising the expenditure item weights and for studying family buying habits; the Point of Purchase Survey provides data on the outlets at which people shop, to be used in updating the outlet samples for the BLS consumer pricing surveys.

In addition to producing the data necessary to update the CPI market baskets of goods and services and to revise the CPI weights, there are many other uses of the data. Some of these uses are to (1) provide a continuous flow of data on income and consumption patterns for use in economic analysis and policy formulation, (2) provide one of the basic data inputs required to revise and update the BLS family budget program as required by the Comprehensive Employment and Training Act of 1973 (CETA), and (3) provide a flexible consumer survey vehicle that will be available for use by other Federal Government agencies.

The Census Bureau is responsible for the survey design, data collection and processing activities for the program.

The surveys were last conducted during the period of 1972 through 1974.

This paper will present an overview of the operational aspects of each of the surveys and point out any significant differences with the earlier surveys.

Sample Design

Data in the two expenditure surveys are collected from a national probability sample of households designed to be representative of the total civilian noninstitutional population. A sample of the primary sampling units (PSU's) was selected, 86 of which have been previously defined and selected by the BLS for the CPI revision. These 86 PSU's represent the urban part of the United States with the additional 16 PSU's added to represent the rural population. Only the 86 urban PSU's are used in the Point of Purchase Survey.

The unit for analysis for all three surveys is the set of eligible individuals comprising a consumer unit. A consumer unit consists of all members of a housing unit or other type of living quarters who are related by blood, marriage, adoption, or some other legal arrangement. The consumer unit determination for unrelated persons is based upon financial independence.

The sample design for the Quarterly Interview Survey will be a rotating panel scheme. Interviews with each sample unit will be conducted quarterly over a period of five consecutive quarters (15 months) and then dropped from the survey. The first interview for a household will serve to "bound" data for expenditures and to obtain information on the household composition and characteristics of its members. Data collected in the initial survey quarter will not be included in developing the regular series of expenditure estimates. Data collected during the second through fifth interviews will form the basis of the estimates derived from the quarterly survey. After the fifth quarter in sample, the panel will be dropped from the survey and replaced by a new incoming panel. The quarterly sample will be divided into three equal parts with each part designated for interviewing in a particular mouth of the quarter and every 3 months thereafter while in sample. The monthly sample consists of approximately 2,860 assigned units.

This scheme differs markedly from the 1972-73 survey, partly because it is a continuing survey rather than a one-time survey. In 1972-73, approximately 23,000 addresses were selected in 216 PSU's. Half the sample was introduced in January 1972 and half in January 1973. Each yearly sample was in the survey for five consecutive quarters.

The survey methodology for the Diary Survey requires each selected sample unit to keep two 1-week diaries of expenditures over two consecutive weeks. The earliest possible day for placing a diary with a household will be predesignated so that each day of the week has an equal chance to be the start of the reference week. The annual target sample size at the United States level for the Diary Survey is 5,470 completed interviews with a total estimated workload of 7,700 sample units. During the last o weeks of each year, the Diary sample will be supplemented to twice its normal size (about 850 of the above designated units) to increase the reportings of types of expenditures unique to the holiday season. The basic difference between the current Diary survey and the one conducted in 1973 and 1974 is again a one-time survey versus a continuing one. One additional difference is in the assignment of a first possible placement day. In the previous survey, each sample unit was assigned a week for placement, with the effort made to place the Diary as early as possible in the sample week. Results of the 1973-74 survey showed that the amount of expenditures reported was highest during the first day of the recording period regardless of what day of the week the first day was and tended to taper off towards the end of the 7-day period. By attempting to control the first day of the recording period, each day of the week has an equal chance of being selected as the first recording day. This should spread any reporting bias resulting from calendar day placement among all days of the week.

Preliminary results of the relationship between the earliest placement day assigned for a diary case and the actual day the diary was placed show that 62 percent of the diaries are placed by the third day and 72 percent are placed by the fourth day. This indicates some difficulty in placing a diary on a specific day.

The sample for the Point of Purchase Survey is a clustered systematic national sample of housing units in 86 SMSA's or ZIP code areas across the country. These areas will be treated essentially as PSU's. The program will be designed so that the entire 86 areas in the sample will be updated every 5 years, approximately one-fifth of the PSU's will be surveyed each year. In the first year, 18 of the PSU's will be surveyed. Each year for the succeeding 4 years, a different fifth of the areas will be sampled. In the sixth year, the cycle will begin again. Through the 5-year cycle, 29,000 units will be surveyed. The only difference between the 1974 survey and the current survey is the spread of the PSU's across 5 years. This was done mostly for financial reasons.

QUARTERLY INTERVIEW SURVEY

Data Collection

The Quarterly Interview Survey is designed to collect data on certain types of expenditures which respondents can be expected to recall fairly accurately for periods of 3 months or longer. In general, expenses reported in the survey are relatively large, such as purchases of property, electrical appliances, and vehicles, or are expenses which occur on a fairly regular basis, such as rent, utility bills, or insurance premiums. The purpose of the Diary Survey is to collect information on the small, inexpensive items which respondents cannot be expected to recall even for relatively short periods of time. Most of the items for which estimates will be derived from the Diary Survey are not sufficiently covered, or are not included at all in the Quarterly Interview Survey. Data collection began in October 1979.

Each occupied Quarterly Interview Survey sample unit will be interviewed once per quarter for five consecutive quarters. In the first interview, information will be collected on socioeconomic characteristics of the household, an inventory of major durable items, and expenditures for selected goods and services during the reference period. Since the initial interview is meant to serve as a bounding interview, the reference period for expenditures is kept to 1 month in order to reduce the amount of interview time. Since duplicate reporting is most likely to occur in one period closest to the interview, the 1 month reference period should provide an adequate period for pounding purposes. A uniform questionnaire with a 3-month reference period has been designed for interviews 2 through 5. Replacement households in interviews 2 through 5 will be unbounded and have a 3-month reference period.

The exact reference period which will be used in all interviews after the first will be from the first of the month, 3 months ago, up to the time of interview. Although this allows the reporting of expenditures for slightly longer than a 3-month period and creates overlaps in reference periods, it is preferable to collect purchases made during one month of interview at this time since recall will be the shortest and most accurate of these purchases. Because of the intentional overlap in reference periods, it was necessary to develop a technique for bounding the data in order to control for duplicate reporting of expenditures in subsequent interviews. Bounding also prevents the reporting of expenses in a current reference period which actually occurred during an earlier period. As stated previously, the first interview's data will be used primarily to bound data reported in subsequent interviews. To prevent duplicate reporting from one quarter to the next, a method was developed for the pretest which has been implemented for the ongoing survey. For most purchases reported, the date of purchase and a description of the item is obtained. At the completion of the interview, the field office will transcribe the information reported for selected items onto designated areas of the next quarter's questionnaire for the sample unit. The interviewer then has the previous quarter's responses available for bounding purposes at the next interview.

The inventory technique is also being used in the Quarterly Interview Survey to improve the accuracy of reporting. An inventory of certain items, such as owned property, major household appliances, automobiles, and insurance policies possessed by the household, is obtained at the initial interview. The inventory serves as a launching point in subsequent interviews, reminding the respondent of any additional or replacement purchases of such items made since the time of the initial interview.

In addition to the initial interview questionnaire and the uniform quarterly questionnaire for the second through fifth interviews, there will also be annual supplemental questions which will be asked at the fifth interview. These questions will obtain more detailed income information for the past 12 months--cash receipts, taxes, contributions, occupational expenses, and detailed mortgage payment data. Another of the differences in methodology between the 1972-73 survey and the current one is the way the data is collected. In the 1972-73 survey, the reference period was not constant as it is now, but varied for different kinds of expenditures.

In the 1972-73 survey, the first interview collected information on socioeconomic characteristics of the household, an inventory of major durable items, and expenditures for selected goods and services purchased since the first of the year. The second, third, and fourth interviews focused primarily on relatively inexpensive items, such as utilities, clothing, small durable goods, household furnishings, vehicle operating expenses, and the like. Data on income received during the preceding year also is collected during the second quarter's interview. The fifth and final interview obtained data on income, assets, and liabilities for the survey period, as well as expenditure data for most items covered in the survey.

The recall period for reporting expenditures varied according to the size and frequency of the expenditure class. Frequent and relatively inexpensive items, such as utilities, clothing, and vehicle operating expenses, were collected at quarterly interviews. Slightly more expensive items, such as small kitchen appliances and furniture, were collected using a 6-month recall period. For very expensive and infrequent purchases, such as real estate, automobiles, and major appliances, the recall period was 12 months.

One of the reasons for such a change in the collection methodology is the estimation procedure to be used. In 1972-73 only annual estimates were produced, with all data collected referring either to 1972 or 1973. In the current survey, "quarter of occurrence" estimates will be produced. This will require that data collected in a 5 consecutive-month period be used for such estimates. For example, expenditure data for the October through December quarter will be collected in the November, December, January, February, and March interview months. For this reason, data must be tabulated by the particular month of the reference period in which the expenditure was reported to have occurred. For some characteristics, this will not be possible as expenditures are not reported by month but are provided either in aggregate figures for a 3--month period or by an average expenditure pattern over time such as monthly or weekly costs. The BLS is exploring different methods of allocating the reported 3-month aggregate expenditures or the expenditure pattern data to individual months within the reference period.

One of the primary concerns in the Quarterly Interview Survey is the accuracy of the expenditure data reported. It is unlikely that any respondent could accurately report all of the detailed information requested without referring to records of some type; however, pretest interviewers estimated that only 50 percent of all respondents used any records at all during the interview. Encouraging respondents to keep records for the quarterly visits in the ongoing survey will continue to be an important part of the program. Attractive calendars and files will be offered to respondents; interviewers will be trained to persuade respondents to keep records; and a letter will be sent before each quarterly visit to remind respondents of the type of expenditures they will be asked about.

Data collection is under the direction of the Census Bureau's permanent professional field staff, operated through 12 regional offices throughout the United States. Each staff member is thoroughly trained prior to Beginning his/her work on the survey. Initial interviewer training consists of 3-1/2 days of classroom training plus self-study training. A 2-day classroom training session was held prior to the start of the second quarter with subsequent self studies and classroom sessions being provided as needed. In addition, a 1-1/2 day supervisor's training session was held prior to interviewer training

Data Processing

In 1971-75, all clerical and computer processing was performed at the completion of the five quarters of interviewing at a centralized location. Since quarterly estimates are a necessary part of the ongoing survey, a different method of processing had to be developed. Although a number of alternatives were discussed, the most feasible was decentralized processing in our 12 regional offices.

When the questionnaires are returned to the offices, they are clerically coded for household and consumer unit relationship, make and model of automobile, trip destination, and detailed codes for type of housing repairs. In addition, all outlets (insurance companies, schools, etc.) are coded uniquely by name. Upon completion of the clerical processing, the data are keyed using a Nixdorf data entry system and transmitted to Washington via communication lines where it passes through a very detailed computer preedit. Inconsistencies, errors, and identification of missing questionnaires are transmitted back to 'the regional offices for reconciliation by the field staff through office review or interviewer follow-up. Corrections are then keyed and transmitted to Washington and cycled through the computer preedit. This continues until errors identified by the preedit no longer appear. Once a panel month's preedit is complete, data necessary for bounding is transcribed to the next quarter's questionnaire. The current quarter's questionnaire is sent to our processing office in Jeffersonville, Indiana, for microfilming and storage.

Industry and occupation coding is completed via computer listings sent to Jeffersonville; the codes are keyed and transmitted to Washington to be merged into the main data file.

The data then goes through a rather complex computer edit. In addition, demographic and work experience items are imputed when missing or invalid. The file is then run through an adjustment to net out business and reimbursed expenses; adjusted to always include sales tax and then weighted.

For interviews 2 through 5, an inventory management system is being developed in order to show expenses for items identified using the inventory technique for which payments are still being made at the same level.

The BLS will be provided with both monthly and quarterly weighted tapes beginning in June 1981, for the first calendar quarter of 1980.

DIARY SURVEY

Data Collection

The purpose of the Diary Survey is to obtain expenditure information for small, relatively inexpensive items which are subject to severe recall loss and for which the Quarterly Interview Survey collects only cumulative expenses or does not cover at all. Expenditure classes not sufficiently covered in the Quarterly Interview Survey include food and beverages, personal care products and services, small household supplies, and a few other relatively inexpensive items.

The survey methodology for the Diary Survey calls for each selected sample unit to keep 1-week diaries of expenditures over each of 2 consecutive weeks. The earliest possible day for placing a diary with a household is predesignated so that each day of the week has an equal chance to be the starting day of the reference week.

Respondents are asked to report all expenditures made during the recording period, not just those which will be used for CPI revision purposes. The main reason for this is to take away from the respondent the decision making about which items to record in order to reduce the chance of error due to the respondent's confusion over whether to record an item. It was felt that this additional recording would not increase the respondent's burden for reporting, since the vast majority of the items reported are those of interest to the survey.

The interviewer makes three visits to the Diary household over the 2-week period. At the first visit, the interviewer conducts a brief interview with the household respondent which includes obtaining the household composition and selected demographic characteristics of each member. The interviewer then explains to the respondent how to keep a diary, the type of detail that is needed, and where to record the information. The interviewer then makes an appointment for 1 week later, At that time, s/he returns to the household and reviews the first diary for completeness and legibility, updates the household composition, and asks a few selected questions regarding common types of expenditures. The third appointment is made, again for 1 week later. When the interviewer returns for the third visit, the second diary is reviewed and selected questions on income and work experience are asked.

If the interviewer arrives to pick up the diary on the appointed day, and the diary has not been completed, the expenditures for the week are obtained through recall. The interviewer would mark a box on the front of the questionnaire to indicate the Diary was completed by the recall procedure. However, if the Diary is not picked up on the appointed day, the expenditures cannot be obtained through recall.

Because the weekly Diary workload is so small, interviewers in the Quarterly Interview Survey also work on the Diary Survey. A 3-day training session was held prior to the start of the Diary Survey and included concentrated instructions on concepts common to both the Quarterly and Diary Surveys as well as training for the Diary Survey. This was done to shorten the Quarterly classroom training which has traditionally been 5 to 7 days.

A l-day supervisors' training session was held prior to interviewer training. Any additional supervisor or interviewer training will probably be combined with that for the Quarterly Interview Survey.

Data collection for the Diary Survey began in September 1979, using the product-structured diary which is composed of a similar set of adjoining pages for each day of a 7-day period. The pages are structured by product category headings to aid the respondent in recalling and recording purchases made during the week.

Data Processing

When Diary questionnaires are returned to the offices, they are checked in and reviewed for completeness and consistency. Any problems are immediately referred back to the interviewer. The questionnaires and diaries are then shipped to a centralized processing location (Jeffersonville) for coding and data keying. The data is then transmitted to Washington where a computer preedit is performed to identify errors resulting from coding and keying. Any remaining errors are corrected and rekeyed.

A computer edit for consistency is then performed along with the imputation of missing or invalid demographic or work experience data. As in the Quarterly Interview Survey, adjustments are made to include sales tax. However, since the respondent is asked only to report personal expenditures, no adjustments have to be made for reimbursed expenses. The data is then weighted. Monthly data tapes will be provided to the BLS beginning in February 1981.

POINT OF PURCHASE SURVEY

Data Collection

Data collection procedures are similar to those used in the 1974 survey, with modifications to the item checklists and to the recall periods for certain expenditure items. The survey is conducted over a period of 4-6 weeks beginning in April of each year. Each consumer unit is interviewed once. The field work takes place under the direction of the Bureau of the Census' permanent field staff operated through the 12 regional offices and the Washington field staff.

Each staff member is thoroughly trained prior to beginning work on the survey. Interviewers will receive 2 days of classroom training plus 8 hours of self-study training materials. Formal training is supplemented by on-the-job training as necessary to ensure job performance, at a level of established standards.

The interview itself consists of two major parts. In the first part, the interviewer elicits information on the demographic and socioeconomic characteristics of the household. Then the interviewer conducts the second part of the interview, which consists of the checklist and questionnaire administered together. The checklist, a list of consumer goods and services, is sectioned into a number of different recall periods. The respondent identifies the items on the checklist which the household has purchased in the stated recall period. Then, the interviewer requests information about all outlets at which a particular type of purchase was made. The name and location (including the closest intersecting street) of each store at which items were purchased and the costs of the items are collected. The interviewer then proceeds to the next items with the same checklist/questionnaire procedure, until the interview is completed.

There are two separate checklists, but only one is administered to a given consumer unit. The checklists are split to reduce the reporting burden for each consumer unit. Unlike for the 1974 survey, there is considerable overlap between the two checklists with respect to the items included. The recall period on each of the checklists varies from 1 week to 5 years depending on the relative frequency and expense of the purchase. A shorter recall period is used for the more frequently purchased items, such as foods and gasoline, whereas a longer period is employed for less frequently purchased and more expensive items, such as major equipment items or vehicles.

At the end of the day's interviewing, the interviewer codes as many of the store names obtained during the day as possible. An alphabetized, precoded computer list of retail, wholesale, and service outlets in the area is provided for this purpose. When possible, the interviewer matches the exact store name and address obtained in the interview to one on the list and codes the outlet accordingly. Those outlets which cannot be coded from the list are coded in a centralized operation at the Bureau's Jeffersonville operation's office.

Data Processing

In the pre-computer processing, all store names not coded in the field are coded. A card file is created to code stores which do not appear on the coding list. The processing of the final Store Code List comprises a major portion of the data processing. BLS has requested that outlets retain the same code from one year to the next. This involves matching codes for two lists compiled 5 years apart, since each PSU is to be interviewed only once every 5 year.

The remaining pre-computer and computer processing is similar to that performed in the 1974 survey. After check-in, general and specific coding and verification, data keying, pre-computer editing, and reject processing, the computer processing will take place. This involves reformatting and editing, matching coded stores back to the original list to obtain additional geographic detail, and weighting the final records. A microdata tape is delivered to BLS by September 30 of each year.

PRELIMINARY RESULTS

Currently, the only results available from the Quarterly Interview Survey or the Diary Survey are field response rates, inasmuch as the computer processing system is still in the developmental stages. The Point of Purchase Survey is conducted only for internal use by the BLS, so again only data on the collection activities are available.

Quarterly Interview Survey

Response rates in the first 9 months of the survey are quite pleasing and compare well with the 1972 survey. Tables 1 and 2 show response rates by month and by quarter for the current survey and the 1972 survey respectively.

Although the initial response race of 90.2 percent for the first quarter is somewhat lower than 1972, by the end of the third quarter, the difference is less than one-half percent, in the third quarter, the response rate of 88.6 percent is only 0.4 percentage points lower than 1972. However, it should be noted that the current survey has rotation groups at various interviews, i.e., first through third interviews, whereas in 1972, all were in their third interview. But given the general decline in response rates in the past few years, the current survey would appear to be doing well.

Diary Survey

Response rates during these first several months have been considerably lower than expected. Although they are comparable to those obtained in the 1972 survey, they are considerably lower than those obtained in the 1973 survey. (see Table 3)

During the first 4 months (September-December), the cumulative response was 78.2 percent, with the rates ranging from 93.4 percent to a low of 59.0 percent among the 12 regional offices. The highest response rate for any week during this period was 64.9 percent for Week 10, and the lowest rate was 58.2 percent for Week 17.

Since the sample is an annual sample, the weeks start again with the beginning of each year and response rates are cumulated on a calendar basis. As of Week 21 for 1980, the cumulative response rate is 82.8 percent with the highest rate reported being 90.9 percent in Week 12 and the lowest rate reported being 69.4 percent in Week 21. There is an overall improvement of about 4.6 percentage points in the 1980 work to date over the rate obtained in the 1979 period, with most of this improvement being noted during the last several weeks.

One of the main reasons for the improvement in the response rate is the increased emphasis and stressing the importance of the Diary Survey to the regional office staff. In addition, interviewers are requested to contact the offices earlier in the placement week when they encounter refusals, that the supervisors must make a better effort to convert refusals, and that we keep constant watch over the rates, getting back to problem offices on a timely basis. Probably the most important effect of our efforts with the field staff has been their placing more effort on contacting households within the allotted time.

When the supervisors were brought together in February, special emphasis was put on the noninterview problem. Supervisors from offices with low Type A rates spoke on methods they use to deal with the problem.

Point of Purchase Survey

The Point of Purchase Survey has just completed its fourth year. A comparison of response rates for 1974 and 1977-80 are shown in Table 4. The 1980 response rate of 94.9 percent is only 1.6 percentage points lower than when first conducted in 1974. Almost all of the change can be observed in a corresponding increase in one refusal rate.

TABLE 1

RESPONSE RATES BY MONTH AND QUARTER - QUARTERLY INTERVIEW SURVEY

TABLE 2

RESPONSE RATES BY MONTH AND QUARTER - 1972 CONSUMER EXPENDITURE SURVEY (QUARTERLY)

TABLE 3

DIARY RESPONSE RATES BY QUARTER - COMPARISON OF 1972 WITH CURRENT SURVEY - DIARY SURVEY (WEEK 1 PICKUP)

TABLE 4

RESPONSE RATES BY YEAR - POINT OF PURCHASE SURVEY

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