Variables appended to ABS frames: Has their data quality improved?

Abstract

Address based sampling (ABS) has become current state-of-the-art methodology for conducting household surveys by mail, telephone or web in the United States. One potential advantage of ABS frames is that additional information about the sampled households can be appended and leveraged for data collection and analytic purposes. The appended data come from many sources and are of variable quality and completeness. The goals of this research were to evaluate data quality of demographic and socioeconomic variables provided for recent ABS samples from one vendor, and to examine their potential usefulness for sample design, including oversampling. We report on the completeness of the appended data as well as their concordance with data reported by respondents to two recent large ABS household surveys, one that invited households to complete the survey online and another that was mail only. Based on the quality assessment, we also examine the utility of the appended variables for oversampling. Our general conclusions are that the quality of select appended variables has improved such that the Hispanic origin, Hispanic surname, and presence of age group 65+ variables may be used to efficiently oversample these subgroups. However, this is not the case for oversampling other subgroups through appended variables for home tenure; those with head of household whose educational attainment is high school or less; low income households; households with children; presence of age groups 18–24, 25–34, and 35–64; or households based on the number of adults in the household.