Beyond who is leading in the survey : A critical examination of the sampling methodology in The Gambia Participate survey
By Dr Alieu SK Manjang
I have had the opportunity to read the full report of the survey released by The Gambia Participate. While many have rightfully questioned several aspects of the survey, including its funding, the selection of potential presidential candidates, and the exclusion of others, these issues may not necessarily invalidate the survey results but raise important questions about their external validity. In this article, I focus specifically on the methodology, and more particularly, the sampling method: how were the 1,556 respondents distributed across Banjul, Kanifing, Brikama, Kerewan, Mansakonko, Kuntaur, Janjanbureh and Basse identified and reached among more than half a million eligible voters?
Polling and surveys inevitably have limitations, as their findings reflect the opinions of the selected group of respondents interviewed. However, this survey is particularly concerning because it does not clearly explain how this relatively small segment of respondents was reached within each geographical area. Unlike the first round of the survey conducted in 2025, which was also limited in its methodological explanation, the current survey provides little additional information about its sampling procedure. The section on methodology on page 7 is arguably the most critical component of any scientific study, yet it does not sufficiently explain how the sample was constructed.
For the purpose of this discussion, I will assume that a stratified random sampling technique was adopted, as was apparently the case in the previous survey, to capture variations in political attitudes across regions. In that survey, stratification was reportedly based on geographic location—urban and rural—and Local Government Area (LGA) boundaries to ensure proportional representation. This raises several methodological questions: Why did the researchers opt for stratified sampling instead of cluster sampling, given the heterogeneous nature of the Gambian population? On what specific characteristics were the strata defined? Within each stratum, what makes the units comparable? More specifically, which villages and districts were included in the sample, and what similarities were assumed among them?
The survey presents respondents’ gender, education level, employment status, age, and LGA as variables. Yet the extent to which these characteristics alone determine voting preferences in The Gambia is questionable.
In many parts of The Gambia, voting preferences within a compound, village, or district can be highly predictable in favour of, or against, a candidate regardless of the gender, education level, age, or employment status of individual voters. At the same time, a neighbouring compound within the same village, a nearby village within the same district, or an adjacent district within the same LGA may favour different candidates, irrespective of these demographic characteristics.
This points to two important factors that appear insufficiently addressed by the survey. First, ethnicity may be another characteristic influencing voter behaviour. This could help explain variations in candidate popularity across regions, constituencies, and villages—not necessarily because of geography itself, but because of the relative demographic composition of particular communities. The omission of ethnicity therefore limits the survey’s ability to explain some of the variations observed in its findings.
Second, this highlights the heterogeneous composition of regions, constituencies, and villages. It raises the question of whether cluster sampling might offer a useful alternative for capturing variations in voting attitudes. For example, randomly selecting 100 respondents in Basse, Kombo South, or Old Yundum constituencies would not necessarily guarantee representation of the different population groups residing in those constituencies. The same concern applies to the random selection of wards and villages. If such heterogeneity is not adequately accounted for, questions naturally arise regarding the external validity of the findings.
Furthermore, since the survey report does not appear to present geographical locations at a sufficiently detailed level, it would have been useful to identify the constituencies and villages selected, together with the number of respondents drawn from each. This would provide greater transparency and allow readers to better understand the geographical implications of the findings, particularly given that different political parties and candidates have identifiable areas of electoral strength. Limiting the presentation to the number of respondents in each LGA is insufficient, especially when only a small proportion of the population in each area was sampled.
In conclusion, surveys of public opinion can provide valuable insights into political attitudes in The Gambia, but their credibility depends heavily on methodological rigour and transparency. Researchers must recognise that audiences engage with survey findings differently: while some readers focus primarily on the results, others are more concerned with how those results were produced. This requires careful selection of sampling methods, a thorough explanation of procedures, and sufficient transparency to allow independent assessment of validity.
Finally, raising these methodological concerns should not be interpreted as an attempt to invalidate this or any previous surveys conducted by the same institution. This critique is specifically directed at the methodology and transparency of the theses surveys and should be understood within that context.
