Commercial Locating Database Efficacy for Telephone Surveys of Low-Income Populations (Presentation)

Publisher: New Orleans, LA: American Association for Public Opinion Research Conference
May 19, 2017
Kim Mook and Sarah Forrestal

Key Findings:

  • Commercial locating databases can vary in coverage by demographic variables; in this study of low-income households with children, we found differences in locating database coverage by household race and ethnicity but not by head of household age or gender.
  • In this study, phone numbers on the frames were more likely to be accurate than numbers provided by a locating database.
  • Phone numbers that were provided by more than one source, as long as the sample frame was one of the sources, were associated with greater accuracy than numbers from the sample frame alone.
Low-income populations are often surveyed for the study of federal assistance programs. Locating identified program participants can be challenging due to mobility, variable employment status, and phone numbers that cycle in and out of service. To address locating challenges in a telephone survey of low-income households with children (n=11,496), we used two common commercial databases that aggregate proprietary and public data sources about individuals and households. Project grantees provided sample frames constructed from either program administrative records or parent consent forms. We submitted households’ contact information (name, address, and phone number) to the databases to obtain additional phone numbers prior to data collection. We compared 1) database hit rate to the frame, and 2) phone number source and quality, defined as successful respondent contact. We hypothesized that databases would vary in demographic representativeness and phone numbers provided by multiple sources would be better quality. Compared to one frame, results show databases were less likely to return hits on Hispanic heads of households but differences were not found by age or gender. Our investigation of the relationship between phone number quality and phone number source(s) showed that phone numbers provided by multiple sources were more likely to be of good quality; phone numbers on the frames, regardless of the frame source, were more likely to be good than numbers provided by a database. Phone quality varied by source within a database, indicating sources should be prioritized. In this presentation, we will recommend guidelines for prioritizing contact information provided by multiple sources in surveys of low income households to maximize efficiency and sample representativeness.