The old regime has been blamed for a large discrepancy in the Census 96 figures. But the man who built the apartheid model defends his numbers. Gaye Davis and Mungo Soggot report
THE old guard in charge of counting South Africa’s population has crossed swords with the new in the wake of this week’s revelation that the country has four million fewer people than expected.
Predictions that the first post-apartheid census would notch up a total of 42-million were upset when preliminary results of Census 96 put the figure at 37,9-million.
Blame for the massive discrepancy has effectively been placed on reliance by previous census-takers, during the apartheid years, on over-estimation of the rate at which black South Africans procreate.
Central Statistical Services (CSS) chief director in charge of demography Pali Lehohla said a demographic model used for the 1991 census, on which the prediction of 42-million was based, exaggerated the fertility rates of black women.
But the man who devised the model, Stellenbosch University’s Professor JL Sadie, this week defended it as “built on facts” – and he expressed incredulity at Census 96’s preliminary findings.
“Something is wrong,” he told the Mail & Guardian. “The figure cannot be this low.” It was difficult for him to comment on the credibility of the census without access to all its results, but he believed the figure would have to be adjusted upwards.
But while census officials agree an adjustment is necessary, they say it will be no more than 2% – so there will still be a yawning gap between the final census- based estimate and prior predictions.
Census officials said this week the 1991 census was substantially adjusted upwards to match the results of Sadie’s model. While it was accepted practice to check census results against a model to determine accuracy, it was “highly unusual” to adjust the results to match the model. They believe fertility rates applied in the 1991 model were “too high”.
CSS officials were themselves taken aback by the preliminary results and used three other methods to check them. They are adamant that Sadie’s fertility rates were over-estimated.
However, Sadie said this week: “I can’t find any explanation in fertility rates. They may have declined more rapidly than we assumed, but that still wouldn’t explain a discrepancy of four million.
“The population development programme must have been more effective than we thought – but it couldn’t make all that difference.”
He dismissed as “absolute nonsense” any suggestion that the previous census officials were blinded by apartheid-era assumptions that black people “bred like rabbits. If you haven’t got another argument, you blame it on apartheid – it’s the ideology of the 90s.”
Sadie feels the reason for the discrepancy is more mundane: “I believe they experienced the same problem of every census: the central office cannot guarantee that thousands of workers in the field do their job properly.”
But Lehohla shot back, saying the post- enumeration survey revealed an acceptable undercount of 6,8% which had been corrected. Even if controls over fieldworkers were inadequate, it would not explain a four million discrepancy.
“It is his [Sadie’s] model that over- projected the population,” Lehohla said. “The census had shown rapid urbanisation – a major cause of declining fertility.”
Apartheid-era censuses have been criticised for using a pastiche of data and methodologies, leaving large holes in information about the country’s vital statistics. During 1991, doubts emerged about the inflation by 4,8-million of the 1985 census total.
Dubious population estimates were then built on this shaky base: as there were no official data for black births, they could have been wildly inaccurate, said CSS head Dr Mark Orkin.
Sadie had used data from the 1970 census to estimate the black South African birthrate, while that for whites, coloureds and Indians was based on 1980 figures.
“He was dealing with the largest part of the population for which there was the least amount of information available and projecting forward by the longest period. His modelling was sophisticated, but the available data were risky,” Orkin said.