The goal of the Ask U.S. Panel was to produce representative and reliable statistics on a rapid turnaround for the U.S. Census Bureau and partnering organizations. Access to a pool of prerecruited panel members, including historically undercounted populations, will help researchers at the Census Bureau (Census) and elsewhere better understand public opinion related to federal data collection, including administrative data matching, privacy, and confidentiality, and will facilitate methodological testing; thus, contributing to the public good.
The Census Bureau and RTI International began working together in September 2020 to design the Ask U.S. Panel. The Ask U.S. Panel was to include an entirely new representative, probability sample of U.S. adults who were not members of an existing survey panel, with the possible addition of targeted subgroups to supplement the new panel in the future. Pilot data collection, which included mixed-mode recruitment, was conducted from April to September 2022, and full panel recruitment was not conducted.
The Ask U.S. Panel Pilot evaluated several design aspects to inform the possible development of a future panel, including experimental design results, nonresponse bias evaluation, and an observational study on tablet acceptance. Data were evaluated and edited prior to analyses. Among the 4,503 addresses randomly chosen for the pilot, 822 (18.3% unweighted, 36.6% base-weighted response rate) completed the screener. Both the use of the visible $5 incentive and the tag letter and envelope were shown to have a significant influence (0.05 test level) on the unweighted and weighted response rates. Fifty-eight percent of sampled adults completed the baseline questionnaire and enrolled in the panel. The overall response rate for the topical survey among participants who completed the baseline questionnaire was 28.6% unweighted (37.1% weighted). Nonresponse bias analyses detected statistical significance for census region in both the screener and baseline.
Overall, we found the use of mixed-mode recruitment for a web panel essential for minimizing nonresponse bias. Offering an internet-enabled tablet to sample members who reported no or limited internet access did not prove beneficial.