Methodology disclosure
Last updated May 11, 2026
centpoll publishes a methodology statement with every public-facing poll. Our default template is aligned with the AAPOR (American Association for Public Opinion Research) Transparency Initiative disclosure standards. The Service generates this statement automatically from the campaign settings at field close — you can edit before publishing.
Minimum disclosure standards
Every poll fielded through centpoll discloses, at minimum, the following:
- Survey sponsor and underwriter
- Exact wording and order of every question asked
- Population the sample is intended to represent
- Sample source (L2 voter file, panel, or client-provided list)
- Number of complete responses, broken down by mode
- Field dates and time-of-day distribution
- Method of contact (P2P, live caller, IVR, online panel)
- Base used for percentages (registered voters, likely voters, all adults)
- Sampling error / margin of error at the 95% confidence level
- Weighting variables and target distributions
- Response and completion rates, computed per AAPOR RR3
Weighting
We use iterative proportional fitting ("raking") to weight responses to match the voter file on age, gender, race, party registration, and region. For likely-voter screens we add a propensity score derived from past-three-cycle turnout. Weighting recipes are saved per workspace and reused for tracking polls so trend lines are comparable across waves.
Margin of error
Margin of error is computed at the 95% confidence level on the weighted base, with a design effect derived from the weighting distribution. We do not report margin of error on sub-samples smaller than n=100.
Quality controls
Every completion passes through automated checks: an attention trap, a length-of-interview floor (we reject sub-60s completions), a straight-lining detector for matrix questions, and a bot fingerprint based on device and behavioral signals. Flagged records enter analyst review before they count toward quota.
Open data on request
We will release anonymized respondent-level data on reasonable academic request, subject to the sponsor's approval. Contact research@centpoll.com.