Why Gambian polling organisations should publish the full methodology, not just the headline numbers
By Dr Ebrima Ceesay
As The Gambia moves towards the 5th December 2026 presidential election, opinion polling is likely to become an increasingly important part of the political conversation. It is also likely to become increasingly controversial.
In The Gambia, whenever a poll is published showing one political party or candidate performing strongly, supporters of that candidate are often quick to embrace the findings. Those whose preferred candidate performs poorly are equally likely to question, challenge or reject the results.
Sometimes the arguments become heated. The problem is not necessarily that Gambians are unreasonable or incapable of accepting survey evidence. It is that when the methodology behind a poll is not sufficiently visible, people have little basis on which to judge whether they should have confidence in the numbers.
That is why I believe the time has come for polling organisations operating in The Gambia to adopt a much higher standard of methodological and data transparency as we approach the December 2026 presidential election.
This is not a call directed at any one polling organisation. It is a call to all organisations conducting and publishing political opinion polls in The Gambia. And it is not a demand that pollsters reveal confidential information about individual respondents.
It is a request that they publish enough technical information for independent researchers, journalists, political scientists and an interested public to understand how the numbers were produced.
The headline figures are not enough. When a polling organisation announces Candidate A – 35%; Candidate B – 28%; Candidate C – 17%; and Undecided 15%, the natural reaction is to ask “Who is leading?”. But that is not the only question that matters.
A more important question for anyone interested in the reliability of the poll is: “How did you arrive at these numbers?”
The following questions arise: What was the sample size? Who was interviewed? How were they selected? Where were they interviewed? Were they registered voters, voting-age adults or members of the general population? How were respondents distributed across the country? Were the results weighted? If so, why and how? What was the response rate? How many people refused to participate? When was the fieldwork conducted? What exactly were respondents asked? How were undecided voters treated? What is the estimated margin of error? And what limitations should readers bear in mind when interpreting the findings?
These are not unreasonable demands. They are basic questions that allow the public to distinguish between a carefully designed survey and a set of numbers whose reliability cannot easily be assessed.
I am therefore making a modest proposal: Whenever a polling organisation publishes an important national political poll ahead of the December election, I would encourage it to publish two documents rather than one.
First, the public findings. This would contain the accessible information intended for the general public such as headline voting intentions; candidate support; undecided voters; key demographic findings; regional findings where appropriate; major conclusions; and a clear explanation that the survey represents a snapshot of opinion rather than a prediction of the election result.
Second, a full technical report should be made available. This should contain a methodological appendix providing sufficient information for independent scrutiny. In my view, this should ideally include at least the following.
Sample size: The exact number of completed interviews should be clearly stated. If different questions were answered by different numbers of respondents, that should also be disclosed. A statement such as “we interviewed approximately 1,500 people” is less useful than providing the exact achieved sample.
Sampling frame: Pollsters should explain what population the survey is intended to represent. For example, all adults; voting-age adults; registered voters; or likely voters. This distinction matters enormously in an election survey. A survey of the general adult population is not necessarily equivalent to a survey of registered voters.
Sampling method: This is perhaps one of the most important pieces of information. Pollsters should explain how respondents were selected. Was it probability sampling; stratified sampling; cluster sampling; systematic sampling; random household selection; face-to-face interviewing; telephone interviewing; online sampling; or a combination of methods?
If a multi-stage sampling design was used, the different stages should be explained. The public does not need a statistics degree to understand this information. It simply needs to be provided in clear language.
Geographical allocation: For a country as geographically and politically diverse as The Gambia, it is important to know how the sample was distributed. Pollsters should ideally show the number of respondents interviewed by Local Government Area; constituency; or another clearly defined geographical unit. If some areas were deliberately over-sampled, that should be disclosed and explained.
Respondent selection within communities: Knowing that 200 people were interviewed in a particular area does not tell us how those 200 people were chosen. Were households randomly selected? Was one individual selected from each household? Were interviewers instructed to follow a particular selection procedure? Were respondents simply approached in public places?. These details can affect the representativeness of the resulting sample. This is precisely why sample size alone is not enough.
Weighting and adjustment: If the raw survey data were weighted, pollsters should explain whether weighting was applied; which variables were used; what population benchmarks were used; whether weights were applied for age, gender, geography or other characteristics; and how much the weighting changed the results. Weighting is not inherently a problem. Indeed, it can be an important part of producing a representative survey. But readers should be able to understand what was weighted and why.
For the benefit of the reader unfamiliar with some of these statistical terms, weighting and adjustment is a statistical correction applied to survey data after collection to ensure the sample accurately represents the broader target population.
Fieldwork dates: Every published poll should clearly state when the interviews were conducted. This is especially important during a rapidly changing election campaign. A poll conducted in early August and published at the end of August is not measuring opinion on the day of publication. The fieldwork dates therefore need to be prominent.
Questionnaire and question wording: I would strongly encourage polling organisations to publish the actual questions asked. This is particularly important because wording can influence responses. For example, there is an important difference between asking “If the presidential election were held today, who would you vote for?” and “Who do you think will win the presidential election?”
Those questions measure different things. Publishing the questionnaire would allow researchers and journalists to understand precisely what the percentages represent.
Treatment of undecided voters: If a poll reports, for example, 17% undecided, readers should be told exactly how that figure was obtained. Were respondents allowed to say undecided; don’t know; refused; none; would not vote?
Were these categories combined? Were undecided respondents subsequently allocated to candidates? If so, according to what methodology? This is particularly important when undecided voters represent a substantial proportion of the sample.
Margin of error and confidence intervals: Pollsters should provide the estimated margin of error where it is statistically appropriate to do so. But this should be accompanied by a plain-language explanation.
For example, “The estimated margin of error is approximately plus or minus X percentage points, assuming the sampling design and statistical assumptions described above”. That final qualification matters. A margin of error is not a magic number that makes a poll accurate. It primarily captures sampling uncertainty under particular assumptions. Other sources of uncertainty can arise from non-response, question wording, weighting, respondent selection, turnout and changes in voter preference. Where a conventional margin of error is not appropriate because of the sampling design, pollsters should say so rather than presenting a misleading figure.
Response and refusal rates: This is another area that deserves much greater attention. How many people were approached? How many agreed to participate? How many refused? How many interviews were incomplete?
Non-response can matter because people who participate in surveys may differ systematically from those who refuse. A transparent poll should therefore provide information about its response rate or, where conventional response rates cannot be calculated, an appropriate alternative measure of participation.
Funding and commissioning: There is nothing inherently wrong with a poll being commissioned or funded by an organisation. But readers should know who commissioned it. If a political party, campaign, media organisation, civil society organisation, business or other institution paid for the research, that should be disclosed. This does not automatically make the poll biased. It simply allows readers to understand its provenance and assess any potential conflict of interest.
The pollster’s identity and contact details: Every published survey should clearly identify the organisation responsible; the individuals or research team responsible where appropriate; its institutional affiliation; and contact details through which methodological questions can be directed. If an organisation describes a report as an “independent” analysis or review, the basis for that independence should be sufficiently transparent for readers to assess it.
A full technical appendix: Perhaps the simplest solution is for every serious national poll to include a technical appendix. The main report can remain accessible to the general public. The technical appendix can then provide the detailed information needed by researchers.
In my view, it could contain survey objectives; target population; sampling frame; sample design; sample size; geographic allocation; respondent selection procedure; fieldwork dates; interview mode; questionnaire; response and refusal rates; weighting procedures; demographic weighting benchmarks; treatment of missing and undecided responses; margin of error/confidence intervals where appropriate; design effects where applicable; funding and commissioning arrangements; limitations of the survey; and full tables and cross-tabulations where appropriate.
That would represent a significant improvement in the quality of public debate.
Transparency protects pollsters too: There is another reason why I am making this appeal. Greater transparency is not simply about satisfying critics. It can actually protect polling organisations themselves.
When a poll is published without sufficient methodological information, political supporters who dislike the findings can easily claim “The poll is fake.” “The figures were manipulated.” “The pollster is supporting the other party.” “The sample was deliberately chosen to produce those results.”
Some of these allegations may be completely unfounded. But when the methodology is not sufficiently visible, it becomes much harder for the polling organisation to answer them convincingly.
By contrast, if the pollster publishes a detailed technical report showing how respondents were selected, how the sample was distributed, how weighting was conducted and how the margin of error was calculated, critics can challenge the methodology on its merits. That is a much healthier form of political debate.
Transparency does not mean revealing respondents’ identities. Let me be clear about another point. I am not suggesting that polling organisations should disclose personally identifiable information about respondents. Confidentiality and respondent protection must remain paramount. Nor am I suggesting that raw datasets containing identifiable information should be placed in the public domain.
There are perfectly legitimate privacy and ethical considerations. The objective is methodological transparency, not the disclosure of respondents’ personal information. Where appropriate and ethically permissible, anonymised or aggregated datasets could eventually be made available for academic scrutiny. But even where raw data cannot be released, a detailed technical report can still provide enormous value.
Indeed, the Gambia deserves a higher standard of election polling. The 5 December presidential election will be one of the most consequential political events in The Gambia. Polling organisations therefore have an important responsibility. They are not merely producing percentages.
Their numbers can influence political campaigns; party strategy; media coverage; public expectations; voter perceptions; and the broader national conversation. That responsibility requires a corresponding commitment to transparency.
I therefore believe that the publication of headline findings should no longer be regarded as sufficient for serious national election polling. If a polling organisation is confident in its methodology, publishing that methodology should strengthen rather than weaken the credibility of its work.
If there are limitations, those limitations should be acknowledged openly. No survey is perfect. No methodology eliminates uncertainty. And no pollster can know with certainty what voters will do several months before election day. But the public deserves to know how the numbers were obtained and what degree of confidence can reasonably be placed in them.
I would therefore encourage all organisations conducting political polling ahead of the 5 December Presidential Election – whether established polling firms, research organisations, civil society groups, universities or other institutions – to consider adopting a common minimum disclosure standard.
It should not matter whether the poll shows Barrow 40%, Darboe 20%, Bensouda 15% or Barrow 20%, Darboe 35%, Bensouda 25%. The methodological standard should be exactly the same. A poll should not become more credible simply because it produces numbers that we like. And it should not become less credible simply because it produces numbers that we dislike.
The methodology should be judged independently of the political preferences of the reader. That is the culture we should be trying to build. The public debate needs evidence, not competing assertions.
As we approach 5th December 2026, I expect polling to become increasingly prominent. I also expect arguments about polling to become increasingly intense. That is not necessarily a bad thing. Disagreement is a normal part of democratic politics.
What would be unfortunate is if the debate becomes a contest between competing assertions “Our poll is correct” versus “Your poll is fake”.
We can do better than that. Let us instead ask: What was the sample? How were respondents selected? What was the fieldwork period? What questions were asked? How was the data weighted? What is the margin of error, and under what assumptions? What are the limitations? Who commissioned and funded the survey?
And, ultimately: Can an independent researcher (like me) examine the methodology and reach his or her own informed assessment?
Those are much more constructive questions. In conclusion, my appeal to Gambian polling organisations is therefore a simple one: Please give us the numbers, but also give us the methodology behind the numbers.
Publish the headline findings but accompany them with a full technical report. Show us the sample. Explain the sampling frame. Tell us how respondents were selected. Provide the fieldwork dates. Publish the questionnaire. Explain the weighting. Show the geographical distribution. Report response rates. Explain how undecided voters were treated. Provide the margin of error or confidence intervals where statistically appropriate. Disclose funding and commissioning arrangements. And include a clear statement of the limitations of the research.
Doing so will not eliminate political arguments over polling. It probably will not stop supporters of losing candidates from disputing unfavourable findings. But it will give journalists, academics, political scientists and ordinary Gambian citizens something much more valuable than competing claims of credibility: evidence that can be independently examined.
As The Gambia heads towards the 5th December 2026 presidential election, that is a standard of transparency worth striving for. Better polling transparency will not tell us who will win the presidential election. But it will help ensure that when polling results are debated, we are arguing about evidence rather than simply arguing about numbers.