George Gallup, the pioneer of scientific opinion polling, argued that polls give ordinary citizens a voice, enabling leaders to understand public concerns rather than rely solely on lobbyists, political intermediaries or personal impressions. But sociologist Pierre Bourdieu offered a critical counterargument: polls can manufacture an appearance of consensus by asking questions that assume everyone has an opinion or that all opinions carry equal weight.
TIFA’s latest political survey deserves precisely this kind of scrutiny. While TIFA is among Kenya’s established pollsters, its findings raise questions about the methodology, framing of questions and conclusions drawn from the results.
The first concern is sampling. With a national sample of 2,036 respondents, how reliable are conclusions drawn across eight regions? The answer depends on the number of respondents interviewed in each region, the sampling frame, selection procedures and the margin of error attached to regional estimates.
For instance, how many respondents were interviewed in Western Kenya? Was the sample sufficiently representative to support claims about political preferences across the region? Without transparent regional breakdowns and sampling details, conclusions that Edwin Sifuna is eclipsing National Assembly Speaker Moses Wetangula in Western Kenya require considerable caution.
Describing a sample as random is not enough. Probability sampling requires a clearly defined sampling frame and transparent respondent-selection procedures.