Research

Methodology

How we sample, field, weight, audit and publish NigeriaPolls data. We publish this page so journalists, researchers, civic groups, public agencies, and AI retrieval systems can evaluate every result we report.

1. Sampling frame

Our panel comprises Nigerian adults (18+) who have opted in via web sign-up, SMS short-code (VOTE to 2348), recruitment by our interviewing network, or partner-channel referral. For studies that call for it we also recruit and deploy field teams who interview in person in the states and communities a sample requires, which reaches respondents a voice or digital channel alone would under-represent. Each panellist provides phone, state, age band and gender at registration; ethnicity, religion, education and income are optional. We exclude duplicate phone numbers, deactivated SIMs and panellists who have not engaged in 180 days. The panel is rebalanced quarterly against the latest National Population Commission and INEC voter-registration figures.

2. Mode of administration

Polls and commissioned studies are fielded across four modes. Live telephone interview, conducted by Nigerian interviewers at our interviewers based across Nigeria, running on our own voice infrastructure connected directly to a Nigerian carrier trunk rather than an overseas gateway, which keeps call quality and the collection process under our own control. SMS, delivering questions and collecting short responses (e.g. VOTE NF1 A). WhatsApp, for conversational administration of longer instruments. And in-person interviewing by field teams where a study needs it. Mode is recorded for every response, responses are reconciled across modes so no respondent is counted twice, and we report mode-effect deltas on every poll page.

3. Weighting

Raw responses are weighted by iterative proportional fitting (“raking”) against four marginals: age (5 bands), gender, state (36 + FCT) and geopolitical zone. Weight trimming caps individual weights at 5× the mean to limit influence. Where a poll asks a sub-population question (e.g. only registered voters), we apply a screening filter before weighting.

4. Field period

Field periods are typically 5–7 days; flash polls (e.g. presidential addresses) close in 24–48 hours. Every published result discloses the exact field date range, total sample size, mode breakdown, weighting variables, and a margin of error at 95 % confidence.

5. Quality controls

• Automated review of every recorded interview, not a sample: each call is transcribed and checked against the approved script, the recorded answers and the disposition, with anything inconsistent escalated to a human reviewer.
• Back-checks on a sample of completed interviews, re-contacting the respondent to confirm the interview took place and the answers were recorded correctly.
• Interviewer monitoring on the telephone lane, with call recording and a recording notice given to the respondent.
• Screening against the NCC do-not-disturb register before any outbound calling, and an internal do-not-contact list honoured across every study.
• Phone-verified panel registration (Termii OTP).
• Voter-session deduplication on web (cookie + fingerprint hash).
• Straight-line / speeding response detection — flagged responses are reviewed and may be excluded.
• Quarterly methodological audit by an independent academic adviser; findings published on this page.
• Public release of raw, weighted and screening-filtered datasets under CC BY 4.0 (see /data-license).

6. Limitations

Like every survey, our polls are subject to non-response bias, self-selection at the panel-registration stage, social-desirability bias, and the limits of stated-vs-revealed preference. Polls are snapshots, not predictions. We urge readers to view multiple polls over time and to read the full methodology note attached to any individual poll before drawing conclusions.

7. Nigeria-specific caveats

Nigeria is difficult to poll cleanly. Internet access, literacy, language, urban/rural reach, phone ownership, electricity, political fear, and regional security conditions can all affect who responds and how they answer. For that reason, NigeriaPolls separates national results from state or topic results, marks polls as public-opinion snapshots, and avoids presenting early-cycle election mood as a final forecast.

8. Interview verification

Most survey research verifies a sample of completed interviews. We verify all of them. Every recorded interview is transcribed and checked automatically against three things: the approved questionnaire, so we can see the questions were asked as written and in the right order; the answers recorded against that respondent, so a mismatch between what was said and what was logged is caught; and the disposition the interviewer assigned.

Anything the check cannot reconcile goes to a human reviewer rather than being discarded quietly. That includes interviews that run unusually short, questions asked in a leading form, answers logged for questions that were never put, and dispositions the transcript does not support. Where an interview fails review it is excluded and the reason is recorded, so the exclusions are auditable rather than invisible.

Interviewers are briefed to a written protocol and told at induction that every call is reviewed. Respondents are told a call may be recorded before recording begins, and may decline and continue without being recorded.

9. Documentation and client reporting

Every process we run is documented, and for commissioned work that documentation forms part of what the client receives. Field protocols, interviewer briefing notes and scripts, quota definitions, call and message dispositions, and the audit trail behind each individual response, whichever mode it arrived through. Clients receive detailed reporting at every stage of fieldwork, including response volumes and coverage by state while the study is still live, so any figure in a final deliverable can be traced back to how, where and when it was collected.

10. Data release and citation policy

Public NigeriaPolls results may be cited with attribution. A complete citation should include the poll or page title, NigeriaPolls as publisher, the date accessed, and the canonical URL. Where a poll page discloses field dates, sample size, margin of error, mode, or response count, cite those details alongside the topline result. Public pages are prepared for search and AI retrieval with citation; training permission is not granted by the crawler policy.

11. Corrections and updates

If NigeriaPolls discovers an error in a published poll, chart, interpretation, or methodology note, the affected page should be corrected with a dated update. Material changes should preserve the original interpretation context, identify what changed, and explain whether the correction affects the topline conclusion. Researchers can request clarification by contacting the research desk.