TWO DISTINGUISHED scholars from UCT’s centre for Actuarial Research (CARe), Professor Robert Dorrington and Professor Tom Moultrie were attributed the headline “Census 2022 results are incoherent and implausible” in the report they prepared for the Medical Research Council.
Their executive summary of their report for the Medical Research Council titled, “The 2022 South African Census, a technical report”, is a common script and predictable.
Earlier censuses carried a similar mantra and I have similarly experienced that in the three censuses and two large scale community surveys that I led. The two professors were part of the expert group for the Statistics Council.
The expert group consisted of up to seven PhDs in demography and statistics. The Statistics Council itself had a membership of 15 people, half of whom would be PhDs in statistics and demography.
Statistics South Africa (StatsSA) in these endeavours held fort with a minimum of 10 PhDs qualified in demography, statistics and urban and regional science. These would be in addition to an army of masters and junior degrees that would have worked on the field operations, data processing, data analysis, post-enumeration survey and quality assessment of the census.
The summary and essence of the critique by the two UCT professors reads thus: “Uncritical use of the 2022 census results for policy and planning purposes would be unwise, pending a full review of the 2022 census results. We are working to produce an alternative set of population estimates at the census date that will resolve – as far as possible – the inconsistencies identified.”
The two professors arrogate to themselves the authority to correct the data. This is not new.
On its part StatsSA has always as a matter of principle opened the anonymised census unit record data of an expanded sample to scholars to dip their fingers in Thomas’s wounds and quench their intellectual curiosity. For instance, in their book edited by an expert team of African scholars, a landmark edition was produced on the basis stated above.
Professor Tukufu Zuberi, Dr Amson Sibanda and Dr Eric Udjo were editors of “After Robben Island, The Demography of South Africa”, published in 2005 by Routledge. In it are eminent scholars who treated different topics on the demography of South Africa.
StatsSA in these endeavours held fort with a minimum of 10 PhDs qualified in demography, statistics and urban and regional science besides an army of masters and junior degrees that would have worked on the field operations, data processing, data analysis, post-enumeration survey and quality assessment.
That the undercount is high is not at stake, what is at stake is what the best procedure to adjust for the undercount is? The two professors claim superiority of methods above those of the Statistics Council and its independent experts, the StatsSA team and the base instrument for adjusting thay they used in the main is a statistical procedure - the post-enumeration survey (PES), which besides discussing quality issues in great detail is used for adjusting the counts.
It is a desirable goal to keep the undercount to a minimum of say under 5% – a dreamland for South Africa's entire history of census undertakings.
The following categories of error measures are deployed:
1. Not counted in the census, but counted in the PES (this component is a substantive source of undercount and often accounts for at least about 80% of the undercount).
2. Missed in the PES but counted in the Census.
3. Missed in the PES and missed in the Census.
4. Counted in both the Census and the PES. Out of the quadrants is established the percentage under or over count as the sum of the product of these quadrants. If we accept that the undercount whatever its magnitude is based on this science-based calculation, then the arguments invalidating the census do not hold.
Dorrington for much longer of the three decades and Moultrie over two thirds of the three decades have done incredible theoretical work in demography. As scholars their views cannot be taken lightly.
But theoretical underpinnings cannot trump the combined force of theory and practice, which is what StatsSA brings to the party as not only an intellectual hub in theory of measurement, but a distinguished well of practice.
From a praxis value, South Africa would not have been in better hands than in the hands of StatsSA. Space does not allow for the intellectual run-ins that StatsSA has had with both these professors. They kept StatsSA on its toes for the past three decades in every census we undertook from Census 1996 to Census 2022. It would have been an anomaly if they left Census 2022 “intellectually unchallenged”.
As a former statistician-general (SG) imbued in equal measure in theory and practice, I agree with the current statistician-general and his team who undertook the PES and derived the undercount, which by any measure is high. But this does not take away the validity of the science of adjustment.
The size of the Indian population is not an anomaly. I observed this in the 2016 community survey. Its genesis is a misclassification of the Pakistani and Bangladeshi population. Second, the foreign component of the South African population has been the fastest growing and possibly not accounted for in the modelling by the two professors.
As a reminder their ASSA model estimates on mortality were refuted by the body of work in the “Demography of South Africa” publication and were further refuted by the 2006 Community Survey. The last refutation of their model fertility estimate was empirically unseated in the 2011 Census through the successful triangulation of fertility data from longitudinal studies of the demographic surveillance sites (DSS) of South Africa.
I released the 2011 census data against their protestation that asserted similar intellectual superiority of application of methods that seemingly proved to have sought o satisfy their important, but not practical, intellectual curiosity. In the end the DSS acquitted the intellectual and practice decision I took as the SG then. Equally the intellectual and practice decision the SG took in releasing the numbers on the advice of Council will point to competence of intellect and practice.
Intellect not only is important and essential in the development of nations, however, it cannot trump joint force of intellect and practice – it cannot trump praxis.
There is yet an issue the two professors raise. This is one of the Treasury forcing its hand on when the census could have been undertaken. If the information of the two professors is true, the financing overreach by Treasury would constitute a violation leash because the Treasury has no intellectual nor administrative competence on work of the SG and the Statistics Council. And that the two professors said it is important so that we should keep the unfettered and usurpation of power of Treasury over issues the Treasury are not competent in.
In 2002, I suffered a similar fate of consumer price index (CPI) where an uninformed Treasury refused to finance the October Household Survey from which we measured the housing component of CPI. The consequences were disastrous.
Dr Pali Lehohla is a Professor of Practice at the University of Johannesburg, a research associate at Oxford University, a board member of Institute for Economic Justice at Wits and a distinguished Alumni of the University of Ghana. He is the former statistician-general of South Africa.
BUSINESS REPORT