Election 2024: Do Polls Matter?

Rrecently wrote a piece for the Centre for International Governance Innovation in which I described the likely outcome of the 2024 elections as follows – “by conventional wisdom, the economic pinch points felt under Ghana’s ruling New Patriotic Party (NPP) government, poor assessments of government governance and policy performance as captured by the Afrobarometer survey (in 2019, 2022 and 2024), strongly suggest Ghana’s main opposition National Democratic Congress (NDC) candidate, John Dramani Mahama, has the edge.”

Notice the caution in my assertion. The pre-election surveys and polls of those whose work I trust put me in a state of caution.

Here are two examples. Global Info Analytics predicted a win for former President John Mahama. Prof. Smart Sarpong predicted a win for Dr Mahamudu Bawumia although short of the 50 per cent plus one vote needed to avoid a run-off.

With the election now over, there is the temptation to engage in two types of conversations – a) who was right and wrong thus deserving of our collective praise or condemnation; and b) whether any faith can be placed in polls.

These are important conversations to have. However, I would like to engage in a different type of public conversation first by raising this question– how should citizens consume information from pre-election polls and surveys?

Consuming polling data  

Understand the context. Polls, as part of electoral politics, are simply designed to check the “pulse” of voters during an election season. The “pulse” check captures how voters are feeling on important matters of governance and public policy, as well as their voting intentions at a particular point in time. They are never designed to predict with certainty what voters plan to do.

The reality is this – voting behaviour can be very unpredictable. What pollsters try to do is project likely outcomes based on what voters tell them. If voters are predictable, the pictures of large crowds at campaign stops of Dr Bawumia shared on social media should have matched the election results.

Recognise the politics of polls. Whenever I teach data analysis, I tell my students numbers have political meaning.

What I mean is accepting or rejecting polling data sometimes depends on whether it reinforces predisposed narratives about an election. As witnessed during the release of polls by Prof. Smart Sarpong and Mr Mussa Dankwa, NPP partisans were more likely to embrace the work of the former, while NDC partisans the latter.

This is not to say partisans are incapable of independently judging the credibility of a poll or survey.

They are and the public posturing is sometimes done out of political mischief. Partisans prefer good news (polls predicting a win) to bad news (polls predicting a loss).

Partisans

Avoid personal attacks on pollsters. Related to the above point is this – while I understand the response of partisans to polling data and the political mischief that comes with it, there were too many moments during the campaign where I cringed at the hostility with which some responded to their work.

The rhetoric was vile and unprofessional. It was also worrying to see some use their media platforms to engage in personal attacks.

The age of social media also amplifies these unfortunate responses to pollsters.

The name-calling is uncalled for. It is understandable to find the results of poll unacceptable. It is outright unacceptable to read motives and engage in personal attacks.

Understand polling data within the context of its methodology.  I can’t speak for all but only for the two examples I used in this op-ed. These two made public their identity (name and institutional affiliation) and more important, the poll’s methodology.

I am baffled why some researchers or pollsters, who publicly share their election projections or likely outcomes willingly, are unwilling to share their methodology.

The ethics of this work dictates both public sharing of methodology and findings. Back to my main point. When you situate the results of a poll within the methodology used, it helps to understand its findings.

If after the event, the actual outcomes digress from the prediction, it calls for a re-examination of the assumptions and methodology used and how it can be improved for future use.

Concluding thoughts

Mr Mussa Dankwah of Global Info Analytics deserves a lot of commendation, primarily for what set his work apart- multiple polls conducted over almost a two-year period.

 

The writer is the Project Director, The Democracy Project