- Data collection strategy (AAPOR A1)
- A survey of Delaware Democratic primary electorate — revised 17-question poll. Collection: Text message invitation to a self-administered web survey.
- Wording shown before the questions (AAPOR A3)
Invitation text (wave 1), 26,715 sent: "Hi {first_name}, We are looking for insight from {community} voters. Take a {survey_minutes} MIN RedBox Research poll: {survey_url} Reply STOP to end; HELP for help."
Invitation text (wave 1), 26,686 sent: "Hi {first_name}, We are looking for insight from {community} voters before {election_day}. Take a {survey_minutes} MIN RedBox Research poll: {survey_url} Reply STOP to end; HELP for help."
Invitation text (wave 1), 26,665 sent: "Hi {first_name}, We are short on responses from {community} voters. Take a {survey_minutes} MIN RedBox Research poll: {survey_url} Reply STOP to end; HELP for help."
Invitation text (wave 1), 26,659 sent: "Hi {first_name}, We are looking for insight from {community} voters on {election_reference}. Take a {survey_minutes} MIN RedBox Research poll: {survey_url} Reply STOP to end; HELP for help."
Words in braces were filled in for each recipient when the text was sent.
Introduction: the frozen questionnaire has no welcome screen, so the survey opened on its first question.
The survey was self-administered on the web: there was no interviewer or AI script.
- Sample type (AAPOR A5a)
- A probability sample: every voter was drawn at random, with a recorded seed, from the voter-file universe named in the disclosure. No response was generated by AI (no synthetic respondents were used).
- Sample frame and its date (AAPOR A5b)
- Voters were drawn from the AFWH voter file, built from the Delaware official state voter registration file dated July 8, 2026 (received July 26, 2026), for 186,560 drawn voters.
- Coverage (AAPOR A5c)
- 106,725 of the 186,560 drawn voters (57.2%) had a textable number and were sent an invitation; the other 79,835 could not be reached by text and are counted as not attempted (AAPOR 3.23).
- Quotas (AAPOR A5e)
- Sample draws were sized to 103 target cells. No interview was stopped or screened out because a cell was full (no case carries a quota-full or screen-out disposition).
- Incentives and contact (AAPOR A5i)
- No incentive was offered. Contact: a single invitation text; no reminders were sent.
- Interviews by frame and mode (AAPOR A8a)
- Completed interviews by frame: Delaware Democratic Primary Poll 2026 — revised-link continuation 523; Delaware 2026 Democratic Primary — L2 hybrid augmentation 637. By mode: web 1,160.
- Margin of error (AAPOR A8c)
- The margin of error is shown because the sample was drawn at random; with a low response rate it holds only as far as weighting corrects nonresponse, which is why leads are judged against the total-error band below.
- Quality checks (AAPOR A10)
- Each invited voter received a unique survey link; a link yields at most one interview. Quality flags recorded: 44 duplicate submissions. Flags are measured, not used to drop cases: all 1,160 interviews completed by the status date are in the weighted estimates.
- Total error (AAPOR A11)
- Sampling error is not the only error: published primary-election polls have missed the final margin by an average absolute error of 9.0 points, so a lead is called only when it exceeds a 90% total-error band that adds that non-sampling error to sampling error. For two candidates near 50%, the band is ±19.4 points on the lead, against a sampling-only 95% margin of ±6.6 points.
- AI use (AAPOR A12)
- AI in data collection: none; no AI interviewer or chatbot was used (Text message invitation to a self-administered web survey). AI in processing: no language model coded, weighted or wrote any published result; every published number and sentence is produced by deterministic code from the recorded responses.
- Panel (AAPOR B1)
- No panel: respondents were drawn fresh from the voter file for this poll.
- Dispositions (AAPOR B5)
- Final dispositions and AAPOR outcome rates are listed above: RR1 0.6%, RR3 0.6% (e = 1.0000).