On the outcry about the 2022 South African Census
Let's re-run the Census and move on
In the last week there have been dramatic newspaper headlines about problems with the 2022 South African Census. This coincides with the release of an initial dataset, and the announcement that certain information will not be released at all.
For those who prefer to get my views in visual or auditory form, here’s an interview I did on this issue today:
A national census aims to count everyone within a country’s borders and collect basic information about them, usually through a household head or representative. That data is very valuable in figuring out things like how fast the population is growing and a whole range of other demographic trends and statistics. Along with income data it should provide the definitive baseline of these variables for the year in which it is run. For example, national population estimates in-between census years will still use the Census for their estimates. It is also used to inform formulas that decide how public money gets distributed across provinces and municipalities. And it is used as a reference point in various ways for other household surveys. So there is no doubt it is very important.1
However, a census is very challenging to implement and also quite costly compared to surveys run by statistical agencies that just aim to sample a small proportion of the population they are interested in. For example, the Quarterly Labour Force Survey, on which all the employment and unemployment statistics are based, only surveys 30,000 households. In comparison: the 2022 Census claimed to have counted/surveyed 17.8million.
Statistics South Africa (StatsSA) appears to have been somewhat underfunded before, during and after the 2022 Census. As always the National Treasury is keen to distance itself from responsibility and denies that its funding decisions were part of the reason for the problems, but these denials are rarely credible. Perhaps even more importantly, the Census was delayed due to Covid-19 lockdown restrictions and conducted in a period when there numerous, largely unprecedented challenges. As one of the initial press releases noted:
Census 2022 was affected by unprecedented challenges, including riots, ongoing COVID-19 lockdowns and climate change issues such as flooding in some parts of the country.
The most recent drama has been caused by StatsSA announcing that it will not release data on employment, income and certain demographic variables (mortality and fertility). It continues to claim, however, that the ‘numbers’ are otherwise robust.
A contrary view has been put forward by two academic demographers, Dorrington and Moultrie, who were then commissioned by the South African Medical Research Council (SAMRC) to produce a full report on their concerns.
As it happens, I have engaged with the work of these two demographers since my undergraduate days when I took an interest in demographic claims linked to debates about HIV/AIDS in South Africa. Based on that experience I would say these academics are certainly technically competent, but also not averse to allowing quasi-political considerations affect their interpretation and communication of certain analyses. They are also connected to networks in the media and academia that will amplify their claims uncritically. (And, in the case of the former SAMRC president Glenda Gray, fund them to do such work). For that reason, it is initially difficult to assess whether the extent of their concerns are proportionate or somewhat exaggerated.
Regardless of such concerns, these authors also note the role of logistical challenges and underfunding:
Although logistical arrangements for the 2022 South African census were impacted severely by the repeated waves of SARS-CoV-2 infections, the National Treasury – unfortunately – declined a request to reallocate the census budget from the 2021/22 financial year to a subsequent financial year. This and other complications probably contributed to certain problems identified with the estimates produced from the census.
Frankly, the mere fact of such a large undercount was already enough to persuade me that there were likely to be serious problems with the Census that could fundamentally undermine its usefulness. Having skimmed the ‘SAMRC report’ I am reasonably convinced that such problems do exist.
What this creates is a situation where the baseline data on the country is at worst unreliable and at best will be the subject of debate and disagreement for the next decade. There may be beneficiaries from this, such as those who may be contracted to fulfil the promise to resolve the problems through statistical methods [highly doubtful in my view], but overall the implications for the country are definitively negative.
Fortunately, there is an easy - albeit not quick - solution: re-run the Census.
Some commentators may throw their hands in the air and shout about ‘wasting public funds’. I have an evergreen example in response to such concerns: South Africa is currently wasting more than R7billion per year on an ineffective tax incentive that is supposed to create employment for young people but in fact appears to be mostly just a transfer of public money to the profits of large companies. (See this academic paper and this more accessible article). A new Census would likely not cost more than R4billion and, when it is done effectively, happens once every ten years. Even better, the Census process creates a large number of temporary jobs.
The money is there, the need is there, and there are additional benefits like temporary job creation: just re-run the Census and let’s move on.
26/08/2024 Update: the two demographers at the University of Cape Town who have been most (publicly) critical on this issue have an op-ed in News24 today arguing against rerunning the Census. I do not find their arguments at all persuasive. For example, they say that planning the Census always takes 4-5 years, and that it would first be necessary to analyse everything that went wrong with the recent one before re-running it: neither of which is true. They provide no credible alternative way forward, simply referring to it as ‘fraught and complex’. An issue I hint at above is that one of the likely outcomes if the Census is *not* re-run is that the next ten years will create a ‘cottage industry’ of academics and researchers analysing all the problems with the data and promising that they can fix them; these authors would be among the primary beneficiaries of that situation. As I also said in the SABC interview above, this is a decision that needs the input of not just a wide range of academics and researchers, but policymakers and members of the public. It is an ideal task for Parliament and I hope the relevant parliamentary committee will consider convening a discussion with StatsSA, that is open to public submissions, on this soon.