Dr Assan Jallow
The recently released findings of the survey titled “Public Perceptions and Leadership Prospects Ahead of The 2026 Presidential Election,” by the Gambia Participate through CepRass has drawn intense debate in The Gambia and outside its shores of which many had dismissed the results and ostracised the researchers purely based on political expediency and not on the arguments of facts driven by data to advance the discourse on scholarship of objectivity in the realm of explorations or research investigations.
Caveat: I am not doing any data analysis here, but instead sharing theoretical and informational knowledge to help those in the debate of questioning the results to build the foundational understanding behind the workings of polling and the governance of statistical sampling.
First, what throws off many people when a study uses a 1% margin of error (MoE) most often is the question of how surveying just 0.16% of the population (1,556/962,000) can give you an MoE of 1%? I get it, but it is noteworthy to equally note that a 1% MoE does not ipso facto mean that there is a 100% chance the actual value is within 1%. However, a 1% MoE from a sample of 1556 from a voter population is highly reliable if the proper methodological mechanics were appropriately designed and applied at the time the respondents took the poll. This suggests that a sample can be nationally representative if the sampling bias, measurement bias, non-response bias and weighting scales are noted in such studies.
Second, data is the new oil, and sometimes, if we wrongly interpret or use false positives in our data analysis, it could result in tilting or biasing the data with inaccurate or misleading results. Understanding your data, particularly the concept of internal consistency, is crucial. This means ensuring that the items in the scale accurately reflect the construct we are attempting to measure. For instance, when we talk about construct validity, we are referring to how well the items in a scale represent the underlying concept. This includes content validity, which ensures that the scale covers all aspects of the idea, and convergent and divergent validity, which confirm that the scale correlates with other measures as expected. Criterion validity, concurrent validity, and predictive validity are also important aspects to consider (Adams & Lawrence, 2019). This knowledge empowers us to make informed decisions and ensures the reliability of our findings.
Friends, there is no monopoly or miracles in findings of studies, as the only established miracle is generated through the use of appropriate measures to get the meaningful results. Having said that, the reliability and validity of a study are critical components, as they determine the trustworthiness and accuracy of its findings. Reliability ensures consistency across measurements, while validity refers to how accurately the results confirm that the research truly measures what it intends to measure, and this is examined through internal and external validity (Adams & Lawrence, 2019). Accordingly, these provide confidence that conclusions drawn from the data are meaningful and scientifically sound. However, the authors further argued that “results of a study cannot be valid unless they are reliable and that results of a study can be reliable but not valid,” suggesting that reliability alone is not sufficient to demonstrate validity. Central to this theme is that the consequences of unreliable or invalid data analysis can be severe, leading to incorrect conclusions and potentially harmful decisions. This underscores the importance of these principles in both research and personal decision-making.
Finally, one needs to understand the methodology used for determining the sample size. This is key in the critiques. Based on the mentioned thoughts, I would advise those disputing the findings of the survey polls to reach out to the researchers for the data to do a replication of the study, or conduct a similar survey to see what results they will arrive at.
