Enhanced demographic-economic data sets

Survey of Current Business, Nov, 1988 by Roger Herriot, Chester Bowie, Daniel Kasprzyk, Sheldon Haber

Enhanced Demographic-Economic Data Sets

THIS paper explores the possible development and uses of data sets that combine demographic data--both survey and population census--with economic census and administrative information. It describes the 1984 Survey of Income and Program Participation (SIPP) and various pilot projects to augment the SIPP data with information about the establishments and firms for labor force analysis and with tax return information for income studies. The ability to add industry or labor market variables is also discussed.

The idea of augmenting survey data with information from other sources is not new. Such microdata record matches have a number of uses.

. They can add information that cannot be collected from survey respondents--information, such as the amount of the employer's contribution for the respondent's health insurance or of a respondent's contribution to social security over a worklife.

. They can add "contextual" variables about the area in which a person lives or works--variables such as a city's unemployment rate or a neighborhood's racial composition.

. They can provide direct comparisons for evaluation of the accuracy of a respondent's answers to the survey questions--for example, the amounts of wages or social security receipts reported by respondents can be compared with the amounts on administrative records.

. They can be used as weighting controls to calibrate the survey and to reduce the variance for many items.

. Finally, they can be used to replace a respondent's answers in some situations to improve accuracy or to model estimates using both survey and administrative data.

The SIPP was the first Census Bureau survey designed from the beginning to facilitate such matching activities. The SIPP, which began in 1983, was preceded by an 8-year development program--the Income Survey Development Program (ISDP). With respect to matching survey data to administrative records, the philosophy, attitudes, and plans of the ISDP strongly reflected the experience gained in a 1973 exact match study (Scheuren et al. 1975). A review of the work of the ISDP with regard to the use of administrative records can be found in Kasprzyk (1983) and Griffith and Kasprzyk (1980). SIPP design features

The primary goals in designing the SIPP were twofold: (1) To improve the reporting of income and program-related data in a way that would allow the analysis of changes over time at a microlevel, and (2) to accommodate the collection of a large quantity of information in a flexible manner that allowed some information to be collected more frequently than other information. These goals were met principally by using a survey design in which the same people are interviewed more than once.

Persons (15 years old or older) in households selected for a sample panel are interviewed about their income and other topics once every 4 months for approximately 2 1/2 years. These sample persons are interviewed at new addresses if they move, and any other persons that they move in with, or vice versa, are also interviewed. In this way, a highly detailed record is built up over time for each person and household in a sample panel. This design minimizes the need for sample persons to recall most of the information for more than a few months, and it reduces the number of questions asked in one interview.

To enhance the estimates of change, particularly year-to-year change, a new sample panel is introduced every year rather than at the conclusion of a panel. Consequently, two, or sometimes three, panels are in the field concurrently. The overlapping panel design allows cross-sectional estimates to be produced from a larger, combined sample that is about double in size when two panels overlap and about triple with three overlapping panels.

The reference period for the primary survey items is the 4 months preceding the interview; for example, for the February interview, the reference period is the preceding October through January. When the household is interviewed again in June, the reference period is February through May. To create manageable interviewing and processing work loads, the sample households within a given panel are divided into four subsamples of nearly equal size. These subsamples are called rotation groups, and one rotation group, or one-fourth of the sample, is interviewed each month. Thus, it takes 4 consecutive months to interview the entire sample. This 4-month period of interviewing is termed a "wave." SIPP content

Each interview is planned to take about 30 minutes, and it includes content that is divided into three main groups of questions--control card items, core items, and topical module items.

The control card is used to list every person residing at an address and to record basic social and demographic characteristics (e.g., age, race, sex, and educational attainment) for each person at the time of the initial interview. At subsequent interviews, changes in these characteristics are recorded on the card, as well as the dates when persons enter or leave the household. Some information relating to the housing unit or household also is collected (e.g., number of units in the structure and tenure).