No Easy Fix for Bogus Respondents in Online Opt-In Polls
a day ago
- Bogus respondents are a major threat to online opt-in polls, answering dishonestly to finish surveys quickly for rewards.
- Three methods were evaluated to identify and purge bogus cases: trap questions, automated prescreening, and matching to voter files.
- Trap questions and prescreening similarly improved data quality by reducing yea-saying, problematic open-end answers, and primacy effects.
- Matching to voter files slightly increased error by disproportionately removing valid respondents (e.g., those who declined to share contact info or were unregistered).
- All three methods modestly increased overestimates of Democratic support in the 2024 election, as bogus respondents tended to claim voting for the winning candidate (Trump).
- Bogus respondents are more likely to claim demographic characteristics like ages 18-29, Hispanic, and male.
- No single method provides a surefire solution, and each has trade-offs in accuracy and bias.