Polling Methodology Explained: How NigeriaPolls Designs Polls

Every poll starts with one clear research question. Are we measuring presidential approval? Party preference? Attitudes to fuel subsidy removal? The question decides the population to sample, the wording to use and the analysis to run.

Defining the Target Population

The target population is the group whose opinions a poll sets out to measure. For most national polls it is all Nigerian adults aged 18 and above. A pre-election poll may narrow it to "likely voters", meaning adults who are registered and say they plan to vote. A youth poll may target 18 to 35-year-olds. Defining the population precisely matters because it decides who is called and how results are weighted.

Questionnaire Design

A poorly worded question can move results by 10 percentage points or more. Questionnaire design therefore follows a fixed sequence:

1. Drafting: questions are written from the research objective and a review of earlier surveys.
2. Neutrality review: more than one analyst checks each question for balance, clarity and leading language.
3. Translation: questions are translated into Hausa, Igbo and Yoruba by native speakers, with back-translation to confirm the meaning holds.
4. Pre-testing: a small number of one-on-one test interviews confirms that respondents understand each question as intended.
5. Revision: wording, answer categories and skip logic are fixed before the poll goes into the field.

The Order Effect

The order of questions can influence answers. A respondent asked about the economy first may rate the president lower than they otherwise would. To limit this, question blocks are rotated across respondents and sensitive demographic questions such as income, religion and ethnicity go at the end.

Balanced Alternatives

When asking about candidate preference, all major candidates are read in rotated order, so no name gets the advantage of being listed first. That advantage is a well-documented phenomenon called the primacy effect. An explicit "undecided" option is always offered, because many Nigerian voters have not made up their minds months before an election.

A poll is only as good as its sample. If the sample is biased, the results are wrong however sophisticated the analysis. NigeriaPolls draws its samples from a random-digit-dial (RDD) frame of more than 11,000 Nigerian mobile numbers, built to cover all 36 states and the FCT.

What Random-Digit-Dial Means

In random-digit-dial sampling, phone numbers are generated from the valid number ranges of each network rather than taken from a directory. Because the numbers are generated and not looked up, people who are not listed anywhere still have a chance of being selected. Numbers are then drawn at random from the frame, so selection does not depend on who is easiest to reach.

The frame was built to cover every state, not just Lagos and Abuja, so that state-level and zonal results can rest on a usable base.

Stratification, Explained

Stratified sampling divides the population into subgroups (strata), such as states, and samples within each. It prevents the geographic skew you get when a simple random draw over-represents the biggest cities. NigeriaPolls uses the state as its primary stratum so that every state and the FCT are represented.

Choosing One Adult

When a call connects, the interviewer confirms the respondent is an adult before any question is asked. A phone number can be shared, so household-selection methods such as the Kish grid exist in the survey world to avoid interviewing whoever happens to answer. The Kish grid is a pre-assigned table that tells the interviewer which adult in a household to speak to. It is the international standard for household surveys and is used by Afrobarometer and the World Bank.

Sample Size

The sample size of every poll is published with the margin of error it implies. As a general rule, a national sample of about 1,500 adults gives a margin of error near plus or minus 2.5 percentage points at 95% confidence. The guide on sample size and accuracy shows the arithmetic.

NigeriaPolls polls are conducted by telephone. Phone fieldwork is fast, reaches all 36 states and the FCT from one operation, and reaches places that are hard or unsafe to travel to. Mobile phones are how most Nigerian adults are reached today, which is why a mobile-number frame gives national coverage.

Repeat Attempts

A single call is never treated as final. Selected numbers are called at different times of day and on different days before they are written off, because people who are hard to reach often think differently from people who answer the first time. Every contact attempt, refusal and non-contact is logged so the response basis can be reported honestly.

Language and Consent

Interviews are offered in English, Hausa, Igbo, Yoruba and Pidgin, matched to the respondent's preference. Verbal informed consent is obtained before any question is asked, and answers are never linked to a name in published results.

Coverage and Its Limits

A phone poll reaches people who own or can use a mobile phone. Those who do tend to skew somewhat younger, more urban and better educated. Weighting to official benchmarks corrects the known gaps in age, sex, state and urban or rural residence, and each release states clearly who the poll covers.

Security and Access

Some parts of Nigeria are hard to reach on the ground because of insurgency, banditry or communal conflict. A telephone frame reaches residents there without sending anyone into danger. Where network conditions leave coverage in an area thin, the release says so.

Data Validation

Before analysis, every completed interview is checked. Duplicate numbers are removed, interviews that finish implausibly fast are flagged, and incomplete or inconsistent records are excluded or reviewed.

Weighting

Even a well-drawn sample drifts from the population. Younger men answer phones differently from older women, and urban residents are easier to reach than rural ones. Weighting adjusts each respondent's influence so the sample matches known population figures. NigeriaPolls weights to official benchmarks from the National Bureau of Statistics (NBS) and the National Population Commission (NPC) on:

• Age group
• Sex
• State and geopolitical zone
• Urban or rural residence
• Education level, where a reliable benchmark exists

The weighting variables and the benchmark sources are listed in each release, so any reader can see what was adjusted.

Descriptive Analysis

Analysis starts with weighted toplines for every question, followed by a check that the numbers hang together: do the answers on related questions point the same way?

Cross-Tabulations

Results are broken down by state, zone, age, sex, education and urban or rural residence. Because each subgroup is smaller than the national sample, its margin of error is wider, and the release flags any breakdown that rests on too few interviews to report.

Reporting Standards

Every NigeriaPolls release carries:

• A plain-English summary of the key findings
• The exact wording and order of every question asked
• Field dates and the sampling frame used
• The response basis and the sample size, with the margin of error it implies
• The weighting variables and benchmark sources
• The sponsor, if the poll was commissioned

1. Call Logging

Every dial, outcome and interview length is recorded, which makes unusual patterns visible and gives each release a response basis that can be audited.

2. Supervisor Review

Completed interviews are reviewed by a supervisor, and a share of them is re-checked against the call record before data is accepted.

3. Recording Where Permitted

Where respondents consent, calls can be recorded for quality review. Recordings are used for training and verification only.

4. External Validation

Poll results are compared against official data and against past election outcomes wherever a fair comparison exists. The 2023 poll accuracy retrospective shows how that kind of check works in practice.

1. Sampling Error

Any poll of a sample carries a margin of error. A result of 45% with a margin of plus or minus 3 points means the true figure is probably between 42% and 48%. Small leads inside the margin are not leads.

2. Non-Response Bias

People who refuse to take part may differ from those who agree. Repeat attempts and weighting reduce the effect but cannot remove it, which is why the response basis is always published.

3. Social Desirability Bias

Some respondents give the answer they think is acceptable, especially on religion, ethnicity and vote choice. Neutral wording, guaranteed confidentiality and an explicit undecided option help limit this.

4. Coverage Error

A phone poll speaks for people who can be reached by phone. The release states the population covered, and weighting narrows the gap to the wider adult population.

5. Timing

A poll is a snapshot of the fieldwork dates, not a forecast. Opinion moves with news, prices and events, so field dates are always printed.

6. The "Shy Voter" Phenomenon

In some elections, supporters of a controversial candidate decline to say so. Comparing results across different question styles is one way pollsters test whether this is happening.