This chapter describes the two components of the Consumer Expenditure Surveys (CE)—the Interview survey and the Diary survey. The Panel on Redesigning the BLS Consumer Expenditure Surveys identified some limitations associated with these two surveys, and these issues are presented in Chapter 5.
As noted in Chapter 1, the current CE are based on the design of 1972–1973 predecessor surveys (Bureau of Labor Statistics, 2008). The CE consist of two different surveys of U.S. households conducted independently, the Interview survey and the Diary survey. Each of these surveys is designed to represent the total U.S. civilian noninstitutional population, using the 2000 Census 100-Percent Detail File augmented by new construction permits and through coverage improvement techniques.
In both surveys, the sampled unit consists of:
(1) all members of a particular housing unit who are related by blood, marriage, adoption, or some other legal arrangement, such as foster children; (2) a person living alone or sharing a household with others, or living as a roomer in a private home, lodging house, or in permanent living quarters in a hotel or motel, but who is financially independent; or (3) two or more unrelated persons living together who pool their income to make joint expenditure decisions. Students living in university-sponsored housing are also included in the sample as separate consumer units. (Bureau of Labor Statistics, 2008, p. 2)
The two surveys are sampled and conducted independently. Expenditure estimates are made independently from the two surveys for various purchased items (goods and services). The concept of having two distinct surveys is fairly simple. The Interview survey was “designed to collect data on the types of expenditures respondents can be expected to recall for a period of 3 months or longer. In general, expenditures reported in the Interview Survey are either relatively large, such as for property, automobiles, or major appliances, or occur on a fairly regular basis, such as for rent, utility bills, or insurance premiums” (Bureau of Labor Statistics, 2008, p. 2). On the other hand, the Diary survey was designed “to obtain expenditure data on small, frequently purchased items, which are normally difficult to recall. These items include food and beverage expenditures, at home and in eating places; housekeeping supplies and services; nonprescription drugs; and personal care products and services” (Bureau of Labor Statistics, 2008, p. 3).
In reality, there is considerable overlap in expense items collected on the surveys, increasing the total response burden. The Interview survey, reaching beyond its original design, also collects information on small, frequently purchased items that may be difficult to recall. For example, it collects expenses for prescription medication, fresh flowers, sewing notions, and the full range of clothing and shoes. It also collects average monthly cost of gasoline and the average weekly cost of buying food in a grocery store. Similarly, the Diary survey, with its open listing sheets, collects many larger items identified as appropriate for the Interview survey. For example, it collects information on the purchase of dishwashers, china, jewelry, and vehicles (Bureau of Labor Statistics, 2011b,c).
The final Bureau of Labor Statistics (BLS) estimate presented in the Consumer Expenditure Survey Integrated Tables for any particular item is based on an estimate from one or the other of these surveys. Creech and Steinberg (2011) describe how the survey source is selected for each published item, and Bureau of Labor Statistics (2009a) provides examples of which source was used for different items for the Consumer Expenditure Survey Integrated Tables for 2009. Some choices are fairly obvious. New refrigerators and living room chairs were estimated from the Interview survey. The Diary survey was used to estimate expenditures for “wine consumed at home” and “lunch at a fast food restaurant.” However, the choice is not always intuitive. For example, bedroom linens were estimated from the Diary survey, while curtains and draperies were estimated from the Interview survey. Under the grouping of “housewares,” silver service pieces were estimated from the Diary survey and other service pieces from the Interview survey. Luggage was estimated from the Diary survey, while smoke alarms were estimated from the Interview survey.
Interview CAPI instrument (2011): http://www.bls.gov/cex/capi/2011/cecapihome.htm
Interview Survey Information Booklet 2011: http://www.bls.gov/cex/current/i_infobook.pdf
Diary Survey Form 2005-10: http://www.bls.gov/cex/csx801p.pdf
Computer Assisted Diary Household Characteristics Questionnaire 2011-12: http://www.bls.gov/cex/ced/2011/cedhome.htm
Diary Survey Information Booklet 2011-12: http://www.bls.gov/cex/current/d_infobook.pdf
The survey questionnaires ask for dollar amounts for services and goods purchased by a household member during the prescribed reference period. They exclude all business-related or reimbursed expenditures.
Survey documents for the CE can be found on the BLS website. Box 3-1 provides the specific links.
Design and Implementation of the Interview Survey
Sampling Frame and Sample Size for the Interview Survey
The sample for the Interview survey begins with a selection of 91 area-based primary sampling units (PSUs). The PSUs may be individual counties, groups of counties, or “core-based statistical areas” (CBSAs) identified by the Census Bureau. Of these 91 PSUs, 21 are metropolitan CBSAs with over 2.7 million people, while 16 are considered “rural.” The rest fall somewhere in between (Bureau of Labor Statistics, 2012) as indicated in Box 3-2.
Each year the Census Bureau selects approximately 15,000 addresses from these PSUs for contact on the Interview survey using the augmented 2000 Census 100-Percent Detail File. The bureau uses a rotating panel design with sampled households contacted quarterly for five quarters. This process results in a usable sample size of approximately 7,100 interviews per quarter.
The 91 PSUs used in the CE sample are classified into four categories:
- 21 “A” PSUs, which are metropolitan core-based statistical areas (CBSAs) with a population over 2 million people
- 38 “X” PSUs, which are metropolitan CBSAs with a population under 2 million people
- 16 “Y” PSUs, which are “micropolitan” CBSAs, defined as areas that have at least one urban cluster of at least 10,000 but less than 50,000 population, plus adjacent territory that has a high degree of social and economic integration with the core as measured by commuting ties
- 16 “Z” PSUs, which are non-CBSA areas, and are often referred to as “rural” PSUs
SOURCE: Bureau of Labor Statistics (2012, p. 5).
Implementation of the Interview Survey
Figure 3-1 displays the flow process of the Interview survey. Work on the Interview survey begins each month with one-third of the quarterly sample. The Interview survey had an overall response rate of 73 percent in 2010. (See “Comparison of Response Rates” in Chapter 5, for a discussion of how these rates are calculated.) Households receive a pre-survey notification letter. The Interview survey is designed for collection through an “in-person” visit by a field representative, and most data are collected in this fashion. However, field representatives are allowed to fall back to a telephone interview and often do. Of completed cases in 2010,1 approximately 17 percent were completed entirely via the telephone, and an additional 48 percent were completed in part over the telephone. The relative number of cases completed over the telephone increases over later phases of the survey. Of interviews completed in quarter 1 (initial interview), only 2 percent were interviewed entirely via telephone and an additional 34 percent had some data collection over the phone. By quarter 5 in the rotation, 22 percent of completed interviews were conducted entirely over the telephone, and an additional 50 percent had some data collected over the phone. Whether via personal visit or over the telephone, the field representative uses a computer-assisted personal interviewing (CAPI) instrument.
1Data on interviews by mode for 2010 come from an internal spreadsheet of costs provided to the panel by BLS.
In the rotating design, approximately one-fifth of the sample is new to the survey each quarter. Only limited data from this initial survey contact are summarized as part of the BLS published estimates. Instead the interview is used to “bound” the time frame for asking future questions on expenditures and to provide baseline data about the household.
The Interview survey is currently designed to collect detailed data on approximately 60 to 70 percent of household expenses. The detailed questions are arranged by major expenditure groupings (such as housing, transportation, clothing, and health care) and ask the respondent to “recall” purchases made for detailed items during the past three months. In order to cover an additional 20 to 25 percent of the household expenditures, the questionnaire also collects three-month average estimates of purchases of food and related items.
The first, second, and fifth interviews of a household deserve additional description. In the first interview with a new household, the survey collects demographic data for the household, inventories major durable goods within the household, and asks for only a one-month recall of expense items. Data from this initial interview are used in only limited ways in BLS expenditure estimates. Instead, data collected in this interview are primarily used for classification of the household, to help prevent duplicate expense reporting in subsequent quarters, and to minimize telescoping (a common tendency in recall surveys to report for a time period beyond the reference period). During the second and fifth interviews, the Interview survey also asks a series of questions to obtain a detailed financial profile. This profile includes income data such as salaries, unemployment compensation, alimony and child support, assets, and investments. These questions use a 12-month recall period.
Proxy reporting is currently used in the CE Interview and Diary surveys, allowing a single household representative to respond for the entire household. The accuracy of data collected from proxies depends heavily on how much the proxy respondent knows about the daily expenditures of all household members. BLS uses proxy reporting in a tradeoff for lower costs and reduced burden. Field representatives attempt to interview the “most knowledgeable” member of the household, who is more often female than male.
The BLS estimates that the average time to complete a quarterly Interview survey is approximately 60 minutes. The panel believes that many interviews are rushed, and so this time estimate may be shorter than what is needed for accurate reporting of expenditures (see p. 83 of Chapter 5, “Motivation in Interview Survey,” for more discussion on this point). The CE Interview survey does not currently offer monetary incentives to respondents.
Sampling Frame and Sample Size for the Diary Survey
Selection of the Diary survey sample begins with the same 91 PSUs selected for the Interview survey. Each year the Census Bureau then draws a separate sample of approximately 12,000 addresses from the augmented 2000 Census 100-Percent Detail File. The effective sample size for the Diary survey is 7,100 interviewed households, producing approximately 14,200 weekly diaries. The placement of diaries is spread equally over the 52 weeks of the year. There are approximately 273 diaries completed each week.
Implementation of the Diary Survey
Figure 3-2 displays a process flow of the Diary survey. Households selected for the Diary survey are asked to keep two sequential one-week diaries of expenditures of household members. Excluded are expenses incurred by a household member while away from home overnight. Also excluded are credit and installment plan payments made during the two-week period. The Diary survey had an overall response rate of 72 percent in 2010. (See “Comparison of Response Rates” in Chapter 5, for a discussion of how these rates are calculated.)
The Diary survey begins with a pre-survey notification letter to selected households, followed by a visit from a field representative to “place” the first week’s diary. As with the Interview survey, the Diary survey is designed for proxy reporting. A single member of the household is asked to keep the diary, recording expenditures for the household and for household members. At this initial visit, the field representative also collects data on a Household Characteristics Questionnaire about the family composition and demographics. BLS uses this information for household classification. BLS also uses this information as a way to associate Diary households with similar households from the Interview survey to analyze comparable expenditures and create integrated tabulations.
The Diary form is left with the person designated as the household respondent. It is a paper-based “self-reporting” form. The form is structured around the “day” of the purchase and by classification of whether the item was (1) food purchased away from home; (2) food purchased for consumption at home; (3) clothing; or (4) other expenditures during the week. The household respondent is asked to list any item purchased and to provide a detailed description of the item as well as its cost.
As the survey is currently designed, the field representative picks up and reviews the diary after the first week and places a second week’s form with the household diary-keeper. The field representative returns at the end of the
second week, picks up the second diary, and collects additional data on the work experience and income for the previous year of individual household members. In practice, both diaries are sometimes placed and picked up at the same time, without the intervening visit. The CE Diary survey does not currently offer monetary incentives to respondents.
BLS provided the panel with an internal spreadsheet containing calendar 2010 survey costs. The cost discussions on the Interview survey and the Diary survey are based on this spreadsheet.
The total “field” cost for the CE in 2010 was $21.2 million. Total field cost includes interviewer salaries/benefits, mileage, training, awards, and related expenses. It also includes Census staff cost in the Field Division. It does not include BLS staff costs or costs for Census employees in other divisions.
The total cost for the Interview survey in 2010 was $17.4 million. This includes five quarters of interviews, dealing with approximately 60,000 cases. The cost per case worked in 2010 was $283, while the cost per completed interview was $487. Interviewing and mileage costs composed 60 percent of the total. Excluding costs associated with noninterviews, the cost for interviews completed entirely or in part via in-person interviewing was $324 each, while those interviews completed entirely over the telephone cost $146 each. The cost per case worked for the first quarter was approximately 14 percent higher than the average of the other four quarters. This is probably due to more screen-outs in the initial contacts and a greater use of telephone interviewing in subsequent quarters.
In fiscal 2010, the total cost for the Diary survey was approximately $3.8 million, or approximately 17.9 percent of the total cost of both CE surveys. At the same time, the Diary survey has approximately 20 percent of the total contacts for the CE surveys.
Table 3-1 provides a summary of the design features and costs of both surveys for comparison.
|Current Consumer Expenditure Surveys|
|Interview Survey||Diary Survey|
|Population||U.S. civilian noninstitutional population||U.S. civilian noninstitutional population|
|Information Collected||Household and personal expenditures: large and regularly occurring expenditures via 3+ month recall||Household and personal expenditures: small, frequently purchased items|
|Sampling Frame||2000 Census 100-Percent Detail File augmented by new construction permits||2000 Census 100-Percent Detail File augmented by new construction permits|
|Sample Design||Multistage area probability sample using 91 Primary Sampling Units; rotating quarterly panel design for five quarters||Multistage area probability sample using 91 Primary Sampling Units; sampled addresses are equally spread across 52 weeks|
|Mode and Field Protocols||Personal interview (CAPI) with decentralized telephone interviewing (CATI) allowed||Initial interview and 2 consecutive weekly diaries; weekly follow-up to retrieve diary|
|Household Sample Size||15,000 addresses||12,000 addresses|
|Average Interview Time||60 minutes (quarterly interview)||Unknown|
|Completed Interviews in 2010||35,843||14,599|
|Overall Response Ratea||73.4%||71.5%|
|Approximate Cost (2010)||$17 million||$3.7 million|
|Approximate Cost per Completed Case (2010)||$476||$248|
|Data Collection Period for Each Household||13 months||2 weeks|
SOURCES: Panel-designed table based on BLS documentation (Bureau of Labor Statistics, 2012) and internal spreadsheet provided to panel.