Stats SA draws the line on population after new headcount challenge

Agency stands by its methodology after fibre company’s data suggests huge discrepancy

A jubilant crowd at the final. 'Nearly the whole of SA came to a standstill when we faced Tunisia for the Africa Cup of Nations final.  It brought the entire country together,' Tovey writes.
Novel techniques cannot replace transparency and replicable statistical methods, says Stats SA. (Per-Anders Pettersson/Getty Images)

Stats SA has firmly rejected a private fibre operator’s claim that South Africa’s population could be far higher than official figures, saying novel techniques cannot replace transparency and replicable statistical methods.

“Stats SA has noted Fibertime’s public claim regarding a population estimate above 95-million; however, the official population estimate as of mid-2025 is 63.01-million,” Solly Molayi, acting deputy director-general, population and social statistics, told Business Day.

“While innovative technologies such as aerial imaging or AI-based extrapolations may offer supplementary insights, they cannot replace transparent, independently replicable scientific methods required for the production of official estimates.”

Solly Molayi, acting director-general, population and social statistics. (Supplied )

Molayi’s comments come a day after Fibretime, a township-focused fibre operator founded by Alan Knott-Craig, published an analysis arguing that operational samples from its rollout — more than 250,000 homes visited and planning data for over 2-million households — combined with drone imagery and AI modelling, point to a much larger population, with a mid-range estimate topping 95-million.

If Fibretime’s assertion, which is framed as a question mark, were proved, it would rewrite the denominators that underpin economic and fiscal analysis overnight, forcing business leaders and governments to rework budgets, grant formulas and sovereign risk models. But Stats SA’s insistence on full methods and data means the figures remain unverified and unsuitable for repricing risks.

Fibretime’s number cruncher, Magnus Rademeyer, said repeated physical counts in townships routinely find far more homes than official records indicate, and that under-registration of births in informal settlements is a major driver of any gap.

Stats SA said it calculates the mid-year population estimate using the cohort-component approach, which relies on the most recent data on births, deaths, net migration and other inputs. Those estimates are anchored to the 2022 national census, Stats SA’s first digital census, and are produced using standard demographic techniques that draw on census results, vital statistics and administrative records.

Molayi’s comments were echoed by former Stats SA chief Pali Lehohla, who refused to accept Fibretime numbers at face value.

DR PALI LEHOHLA
Pali Lehohla.

Lehohla demanded data and method before he could treat the claim as a serious challenge to the official series, which is built on a dwelling listing followed by door-to-door enumeration and, from 2022, computer-assisted personal interviewing to collect and manage responses.

“The guys who are [installing] fibre also say there are structures on the ground; no doubt when they count structures they must come to the same number of structures as Stats SA,” Lehohla told Business Day.

“There’s no reason why they cannot come to the same number of structures as Stats SA. First, they must tell us how many structures they found. Now, suddenly, they estimate that the population should be 95-million. The question is, where they counted the structures, where did they put their fibre? Pound for pound, what is the population they estimate there? What is their method of estimation? What is the average size? Because I’m sure if they count the structures, they must come to the same number of structures as Stats SA.”

Molayi said: “Stats SA is always willing to engage with users of official statistics and will consider any alternative evidence if accompanied by full methodological documentation and anonymised data for rigorous independent assessment.”

Lehohla also acknowledged that new technologies such as AI and aerial imaging are efficient tools that flag structures and prioritise field checks.

“AI is very good; it’s something worth considering. AI and imagery can quickly flag likely structures and direct enumerators where to check, which speeds up fieldwork. However, they cannot replace the full set of parameters you need for a population estimate,” he said.

The disagreement over official statistics is not the first. Independent experts and research bodies have questioned aspects of the 2022 census and its post-enumeration adjustments. The Medical Research Council and academic commentators have described serious inconsistencies and raised caution in using census outputs for fine-grained planning.

Separately, private sector figures have publicly challenged other headline statistics. For instance, former Capitec CEO Gerrie Fourie questioned the unemployment measurement earlier this year, prompting a public rebuttal from Stats SA and intensifying scrutiny of official figures.

Fibretime’s assertion is a stress test for an underfunded national statistics agency. Last month Business Day reported that the Statistics Council — a 25-member panel that signs off on national statistics — warned that high vacancy rates and a shrinking operational budget were undermining its functioning.

Would you like to comment on this article?
Sign up (it's quick and free) or sign in now.

Comment icon