A predictive analysis of polling results administered to Nigerians in anticipation of the 2023 presidential elections.

At Stears, we often say, ‘In God we trust; everyone else must bring data’. 

In 2019, we launched the Stears Election Centre, recently relaunched as Stears Elections, becoming Nigeria's first real-time election results provider. Nigerians loved it and used it to understand data about the past and the present.

But one thing remained. The holy grail of information for decision-makers. Data about the future.

And where better to begin than with Nigeria’s most pivotal and hotly contested elections since the turn of the century?

This is why we conducted polls for the 2023 Nigerian Elections. Now more than ever, Nigeria—Nigerians—need a clear view of the future before we confront it.

The absence of polls in previous elections has made it difficult for decision-makers to prepare for the relevant outcome. We decided to change that.

Looking at other polls, we have seen problems that can only be solved by a data company; the high prevalence of voters unwilling to reveal their chosen candidate when asked during polling interviews. This has prevented other 2023 electoral polls from providing useful predictions of the outcome of the ballot. We desperately wanted to change that, so we built Nigeria’s first-of-its-kind electoral prediction model to identify who these silent voters intend to vote for, inferred from the rest of the polling data they reported.

We are proud to disclose the results of the Stears electoral poll and prediction model, which show the most likely outcomes of the 2023 Nigerian Elections. Over the coming days, we will analyse other parts of the poll not included in today’s results, so make sure to stick around.  

How we conducted Stears' 2023 Electoral Poll

Today, we give a summary that covers our poll's methodology, what makes it unique, the original poll results, our proprietary estimation model that predicts likely voting patterns for silent voters, and the new results after applying the model to the original poll results.

Our objective was to produce the most accurate and predictive poll of Nigerian voters ahead of the 2023 Elections. To do this, there were three things we needed to ensure.

The first is that our sample size is as large enough to be nationally representative. We achieved that: the Stears 2023 Electoral Poll covers 6,220 respondents in total across Nigeria's 36 states, making it the largest telephone-based electoral poll in Nigeria. Furthermore, the poll includes booster samples from strategic states to enable more precise national and state-level predictions. 

The poll included 22 questions (on demographics, voting history, top-of-mind-issues, religion, ethnicity, confidence in INEC, etc.), with an average time of 15 minutes per interview, and was conducted in five major languages: Hausa, Yoruba, Igbo, Pidgin English and English.

The second is that our sample mirrors Nigeria’s voter registration database (voters over 18 years). We used stratified random sampling to ensure the sample is representative of the voter register up to State Senatorial District. And as is best practice in polling, we weighted the sample by gender and geopolitical zone to ensure that the final data exactly mirrors the breakdown of Nigeria’s 2023 voter registration database as of January 2023.

The third reminds us that not all polls are created equal. To be sure that the results of a poll truly reflect the underlying population, a pollster must ensure that everyone in the relevant population has an equal chance of being selected. The best way to do this in Nigeria is via telephone polling, as roughly 80% of adult Nigerians own a basic phone.  

The unique features of Stears 2023 Electoral Poll

Our poll has three unique features.

#1 Statistically reliable state-level electoral analysis and predictions

First, we can provide state-level predictions on the presidential and gubernatorial races. The pre-election polls released so far have focused on national results, even though elections are won at state and regional levels.

It is important to highlight here that although other pre-election polls can report state-level results, sample sizes for each state are too small for these to be statistically significant. In other words, they are more like best guesses.

Our booster samples allow us to generate insights and make predictions for Lagos, Rivers, Kano, Plateau, Benue, and Kwara states, all of which are strategically important in the 2023 Presidential Elections.

Can Peter Obi usurp Bola Ahmed Tinubu in Lagos State? Will Kwara maintain its record of predicting every presidential election winner since 1999? Will voter turnout in Rivers look more like 2003 (2.2 million) or 2019 (700,000)?

You can only get these answers and more from the Stears electoral poll and prediction model.

#2 Multiple predictions based on various turnout scenarios

As The Economist memorably put it, polls are subject to the vagaries of voter turnout. All around the world, much more people express an intention to vote than actually cast a vote. The situation is worse in Nigeria, where the voting value chain is so brittle because voter apathy, voter suppression, and outright rigging are prevalent. All of this means that it is one thing for a voter to resolve to vote, another for them to successfully cast a ballot, and another entirely for that vote to be counted.

All of this means that headline poll results are a popularity contest and likely poor predictors of electoral outcomes. For example, over 90% of respondents in the Stears poll expressed an intention to vote. But voter turnout was just 35% in 2019, and even the most optimistic estimates peg 2023 voter turnout below 60%. So, our predictions account for this risk.

Considering not all candidates will be hit equally by lower-than-predicted voter turnout—for example, candidates with a higher share of new or inexperienced voters are typically more vulnerable—it is necessary to estimate poll results under multiple turnout scenarios.

You will see how drastically the results change when we do that.  #3 Nigeria’s first electoral prediction model

Polls cannot be useful or predictive if they are silent about the preferences of 30-50% of polled respondents. This happens when 30%-50% of respondents refrain from answering questions stating exactly who they will vote for, even though they answer many related questions.

This matters because international evidence indicates that silent voters are unlikely to be evenly spread across candidates, from the shy Tory factor that confused UK pollsters to President Donald Trump consistently confounding US polls.

Whereas a high proportion of silent voters has rendered other polls inconclusive, our proprietary prediction model provides conclusive predictions by estimating the most likely candidate for these silent voters.

For each of the top three candidates, Bola Ahmed Tinubu, Atiku Abubakar, and Peter Obi, we created probit models to estimate the probability that each silent voter will vote (or not vote) for one of the top three candidates. Probit models are the most commonly used for electoral predictions as they are ideal for estimating the likelihood that an observation (e.g. a voter) will fall within a binary variable (vote or not vote).

After building our model, we tested it (to ensure it works) on those voters who had already self-reported who they would vote for. We were very comfortable with the results, seeing over 94% accuracy for each candidate. In short, the model worked.

For those who want to understand more, an intuitive explanation of the model is provided in the appendix of this article, with examples of how the model works.  

Stears 2023 Electoral Poll results

Well done for making it this far. Now, time for the results.

The basic poll results (without applying the model) show Peter Obi holding a healthy lead over his more established rivals. In fact, his share of the vote is roughly Tinubu’s and Abubakar’s combined.

But, before you walk away, remember that these results are simply the starting point, and more is needed.

Look closely, and you will see those silent voters are still larger than any single candidate’s votes, so the poll is inconclusive without the Stears prediction model. Still, it is useful to see the voting pattern across geopolitical zones, so let's explore some more.

Unsurprisingly, voters in the South East (64%) and South South (48%) heavily back Peter Obi. More impressively, Peter Obi is also the most popular candidate in the North Central (26% of voters), ahead of Tinubu (19%) and Abubakar (11%).

Although Rabiu Kwankwaso lags far behind in the national polls, he may be a credible threat in the North West, accounting for 8% of the votes. Perhaps more importantly, Kwankwaso’s presence splits North West votes, with the most popular candidate (Abubakar) getting just 18% of the vote and as much as 45% of North West respondents declining to reveal their preferred candidate.

The poll also shows that Nigerians vote along religious lines.

The numbers are stark, with only 17% of Christians opting for either Tinubu or Abubakar, while 43% support Obi. Meanwhile, Obi gets just 3% of the Muslim vote, even less than Kwankwaso (6%).

Despite this, Nigerians are adamant that religion does not dictate their vote: just 4% of respondents selected religion as one of the three main factors determining their candidate choice.

Make of that as you wish.

Going beyond these basic results, we can provide more precise predictions by accounting for different voter turnout scenarios.

For simplicity, we provide two scenarios here.

High turnout: Includes every respondent that expressed an intention to vote except those that had not yet collected their permanent voter’s card (PVC) and did not indicate that they were definitely going to vote.

Low turnout: Only includes respondents that have their PVC, stated that they are certain about their chosen candidate, stated that they are sure they will vote, stated that they feel very confident INEC’s ability to conduct free and fair elections, and said they feel safe or very safe going out to vote.

The high turnout scenario corresponds to an absurdly high voter turnout of about 85% (Nigerians are very optimistic about their likelihood of voting), and Peter Obi extends his lead over Tinubu.

Something incredible happens in the low turnout scenario: Tinubu and Obi are locked in first place, raising the possibility of a run-off.

Again, this is only partially conclusive due to the high proportion of silent voters—step in the Stears prediction model.  

The Stears model provides more conclusive results

Before we show the prediction model results, it is useful to show which candidates have the most silent voters. Remember, the model estimates three probabilities for each silent voter, so to assign voters to each candidate, we have to set a rule, e.g., the candidate with the highest probability receives that voter.

Our preferred rule is to allocate each silent voter to the candidate with the highest probability, as long as the probability is greater than 40%.

Stears prediction model believes that Tinubu has the most silent voters i.e. people who intend to vote for Tinubu but choose not to disclose this during an interview across the three main candidates.

We can now rerun the analysis using our preferred rule and generate the new poll results.

Even after including silent voters, Peter Obi holds a comfortable lead. In essence, the Stears electoral poll and prediction model affirms that Peter Obi is by far the most popular and desired candidate for the 2023 Presidential Elections. Having said that, to get a more realistic prediction of the outcome of the race, we can apply the same high and low turnout scenarios to our model predictions. When we do so, Peter Obi once again emerges triumphant in the high turnout scenario.

But finally, in the more realistic low-turnout scenario, something significant happens: Tinubu usurps Obi.

For today, that is where we end.

The Stears electoral poll and prediction model indicates that Peter Obi is Nigeria’s most popular presidential candidate and should win the presidential ticket as long as Nigerians follow through on their stated intent to vote. But, if voter turnout looks similar to 2019, then Tinubu will win the election.

stears.com