Skip Header

Working Paper Number SEHSD-WP1989-03 or SIPP-WP-81
Reynolds Farley and Lisa J. Neidert
Component ID: #ti1844122905


Although data from the Survey of Income and Program Participation (SIPP) provide extensive information about the characteristics of blacks, we have yet to determine whether the sample size permits a detailed analysis of racial differences, or whether the characteristics of the black population, as estimated from SIPP, correspond to those same characteristics as measured in other demographic surveys. To explore these issues, we studied data from the fifth and sixth waves of the 1984 panel of SIPP, and compared them to data gathered in the March, 1985 Current Population Survey (CPS); that is, the Annual Demographic File. We selected those rotation groups from the two waves of SIPP which were interviewed during March of 1985, while the interviewing for the Current Population Survey was done primarily during the first week of March. This report compares SIPP and CPS data. (For information about the design of SIPP, see David, 1985).

The first section provides information about the sample size for blacks from one wave of SIPP. This will allow potential users to know whether the sample is sufficiently large to permit the testing of hypotheses about racial differences. Using unweighted data, this report presents sample counts for blacks and whites classified by age, sex and region of residence in March, 1985.

Many users of SIPP data will likely be interested in the marital status, labor force status and educational attainment of adults. To provide additional information about the sample sizes for the population classified by these characteristics, persons age 15 and over were categorized by race and sex into four age groups. For each group, we present information about the sample size for five marital status, three labor force and five broad educational attainment categories. For  comparative purposes, we show similar sample counts from the March, 1985 CPS. Although unweighted counts are used in this section, we realize that most investigators will analyze weighted data, or will attach adjusted weights to each observation so that the weighted sample counts are equal in size to the actual sample sizes, a procedure which may lead to more nearly appropriate tests of statistical significance.

The second aim of this report is to determine whether the characteristics of blacks and whites, as reported in SIPP, correspond to those reported in CPS. We focus upon three important variables: marital status, educational attainment and labor force status. In this section, weighted data will be analyzed.

The final component of this comparison of SIPP and CPS data seeks to determine whether demographic and socio-economic relationships are similar in the three different sources. Two types of relationships are explored. First, a model was fit which took the log-odds of unemployment for labor force participants as its dependent variable. Independent variables were age, educational attainment, marital status and region. These models used the weighted sample data and were fitted separately for four race-sex groups, but were restricted to respondents age 25 to 64 in March, 1985.

Second, using the same groups and a similar age range, a model was fit which related a logged function of monthly earnings - for those who reported earnings - to educational attainment, age, region and place of residence. A major difference between the two data sources is that SIPP reported earnings for a one-month period at the beginning of 1985, while CPS reported an individual's earnings for the entire year of 1984.

Back to Header