Says its ground-level data suggests South Africa may be far more populous than assumed.
Could South Africa’s population be underestimated by tens of millions – and actually sit near 95 million rather than the latest official estimate of 63 million?
Township fibre operator Fibertime believes its granular, street-level data may point to a South African population far larger than official figures suggest.
Fibertime, founded in 2022, is best known for connecting households in informal settlements to uncapped fibre for just R5 a day.
ALSO READ: BMR issues population numbers amid uncertainty of population figures
The 95 million question
Through its operations, the company has now taken a growing interest in a different national project – understanding how many people actually live in South Africa.
The question emerged through Fibertime’s network-building process, which relies on detailed household mapping, large-scale aerial imaging and advanced modelling.
According to Magnus Rademeyer, former CEO and now head of insights at Fibertime, the need for accurate planning information is fundamental to Fibertime’s work – and to every commercial and government endeavour in the country.
“This question has been close to my heart for decades,” Rademeyer says, adding that the scale of Fibertime’s capital investment created an opportunity to address and share insights arising as a byproduct of connecting households across the country.
Fibertime’s work on South Africa’s population estimate is being driven in large part by Rademeyer.
He has extensive experience in geographic information systems (GIS) and spatial modelling, having co-founded AfriGIS in 1997 – one of South Africa’s first mapping and location-intelligence firms where he was CEO and managing director for more than 20 years.
A civil engineer by training, Rademeyer was also part of the Independent Electoral Commission’s national voting-district mapping project that created the electoral maps still in use today.
“I have had the privilege of being involved in elections since 1997 through my involvement there as a GIS contractor.”
His involvement in elections, census work, and land-affairs datasets, he says, has shaped his sensitivity to demographic gaps and inaccuracies.
“When you look at a map and the geography, and you do that over time and in different geographies, you definitely develop a feel for it. It tells a story.”
ALSO READ: Census 2022: Population increases to 62 027 503
Fibertime’s data
Fibertime’s fibre-deployment model has given it unusual access to information within South Africa’s informal settlements.
The company has physically visited 265 000 households “not randomly, but in targeted pockets”, says Rademeyer, by entering homes, drilling, installing routers and observing household structures and living patterns.
It aims to have two million households connected by 2028.
Because the company owns its network infrastructure, it can detect how many devices connect to each router, how dense neighbourhoods are, and how residents use bandwidth.
Rademeyer explains that Fibertime runs this information through regression models that incorporate external datasets, including data from the Department of Water Affairs, Statistics South Africa’s 2011 Census and its own planning data sources.
In addition, the company uses drone photography of coverage areas, combined with artificial intelligence (AI) and machine-learning tools, to extract additional demographic signals, such as car counts, roof counts and head counts.
A mathematical formula containing seven constants then adjusts these estimates. These constants account for factors such as dwelling size, income levels, migrant-worker patterns, weekday-only occupancy and household size.
ALSO READ: Census data ‘unfit for purpose’ – experts
‘When the penny dropped’
As network planning progressed, Rademeyer and Fibertime co-founder Alan Knott-Craig Jnr debated the possibility of large undercounts “from the get-go”, beginning in 2023.
“Early 2024, during a helicopter fly-over and subsequent deep-dive of the Bushbuck ridge area, the penny dropped,” says Rademeyer.
“We confirmed that Fibertime will indeed be producing information which could benefit not only ourselves, but could stimulate a debate on the number of people in the country.”
Rademeyer says many of the early signals arose from “on-the-ground immersive” impressions during field visits.
The team generating the population outputs is small – about five people – but, he says, they work alongside a much larger, decentralised field operation.
Thousands of contractors and subcontractors capture data daily as part of the fibre build. Fibertime is currently active in 39 communities, each requiring a mix of local partners, contractors and community members with detailed area knowledge.
ALSO READ: Census 2022’s 2.4 million foreign-born population figure deeply suspect
Data demo site
Through an interactive website, Fibertime has used an informal settlement in Walmer in Gqeberha in the Eastern Cape as a formal demonstration site for its data.
There, Fibertime has made its drone imagery, modelling outputs and raw indicators publicly available for analysis.
Rademeyer says the purpose is to “invite lively debate” and allow researchers, planners and institutions to help validate – or challenge – their findings.
Over time, these insights will be consolidated in a new publication, Fibertime Signal, which the company wants to publish annually, outlining population estimates and the methodology behind them.
ALSO READ: Half a million white South Africans have left the country in 25 years
Why the question matters
South Africa’s official population estimate, released by Stats SA in July, stands at 63.1 million, but the 2022 Census was beset with difficulties.
The post-enumeration survey found a 31% undercount, described by researchers as potentially the highest reported by any country and likely globally unprecedented.
Experts say an undercount of 5% is considered reliable; anything above 15% is regarded as doubtful.
At 31%, researchers warn, planning becomes seriously compromised because population-based funding models skew allocations towards areas where the count was stronger.
Stats SA has attributed the problems to Covid-19 disruptions, logistical setbacks, social unrest, training delays and difficulties with digital data-collection systems.
In the end, it chose not to publish several key datasets, citing reporting and coverage biases.
Researchers at the University of Cape Town’s Centre for Actuarial Research and the Medical Research Council described the census as “deeply flawed”, highlighting anomalies in age and sex distributions, migration patterns and the heavy reliance on post-enumeration survey adjustments.
Stats SA nevertheless endorsed the census as “credible”, maintaining that its methodology met international standards.
ALSO READ: The people of SA by numbers
What happens next
Fibertime wants its “95 million population” figure to serve as a starting point for debate rather than a firm conclusion.
The company is urging collaboration between the private sector, research institutions and government.
“South Africa,” Rademeyer says, “can benefit if we put our analyses together in lively, collaborative debate.”
This article was republished from Moneyweb. Read the original here.
