- In the fourth Census data released earlier this month, the largest undercount was recorded which led to many questioning the reliability of the results.
- The undercount was nearly double that of the 2011 Census. All of South Africa's democratic Census data have had double digit figures - something that is a major statistical concern.
- The Census data is used by the national government to allocate funding and create policy plans that benefit citizens.
Since the release of the 2022 Census for South Africa was announced two weeks ago, experts have questioned the data because of the largest undercount since 1996 - the first post-apartheid Census.
Only 69% of South Africans were counted, meaning approximately 19 228 525 people were not counted (based on the 2022 Census). This undercount percentage of 31% is the basis why many question the "veracity" and "accuracy" of the data and "cast great doubt on its credibility".
A Census is a snapshot of South Africa at a particular moment. The reference date for this Census is 2-3 February 2022. The data was collected between February and May 2022, with planning having started several years ago.
Census data informs policy and government decisions, particularly funding allocations for provinces, municipalities and district councils. Therefore, accurate Census estimations impact funding and policy decisions.
What is an undercount, and why is it important?
An undercount is the percentage of people or households that were not counted in the Census. Once the Census data is collected, another round of data is collected but on a smaller scale and is "used to measure the accuracy and reach of the Census" - this is called the post-enumeration survey (PES).
This sample survey picks up how many people/households were missed or counted twice. It also provides the basis for statistically adjusting the Census count to the true population estimate. With some statistical equations using a dual-system estimate methodology, the population of South Africa is then estimated, taking into consideration adjustments from the PES. The difference between the estimated true population and the Census data is the undercount or, very rarely, overcount.
READ | Census fails to count 30% of the population
Statistician-General Risenga Maluleke told SAfm that Stats SA takes the undercount percentage and uses it to come up with a population estimate for the nation. He adds that this sample survey is compared with the data from the Census "record for record, settlement for settlement, age for age, sex for sex" to come up with an undercount.
South Africa has always had an undercount, and this is the norm all over the world. The undercounts were 10.7%, 17%, and 14.6% for the censuses of 1996, 2001 and 2011. The 2022 Census had an undercount of 31.7%. A 5% undercount is considered decent. Developed countries tend to regard an undercount of 10% and above as catastrophic. Worldwide trends for undercounts vary, as some countries do not conduct PES surveys or even a census.
READ | Low pay and racism in the Western Cape bogged down Census 2022, Stats SA says
The population figure of 62 million was arrived at by adding up estimations of the people missed, plus the 24.3 million included in the Census, the PES and other living quarters.
According to then Statistician-General Pali Lehohla, Stats SA had "lost the plot" with the 17% undercount in 2001 and promised a lower undercount in the next Census.
So, is the Census credible?
Maluleke said in the SAfm interview on Monday that: "By all means credible, the credibility of the Census is not only decided by Stats SA but by the statistics council".
The Statistics Council is a legislative body that advises the statistician-general on any issue concerned with the production and use of the official statistics.
For Census 2022, the council hired independent experts to make sure the methods and process were correctly done and followed. The council, however, does not see the results until it is released by Stats SA.
Maluleke said that the statement the council delivered to the nation was "that the Census was fit for purpose". There were no red flags raised. He added "we would not have released the numbers if they were not credible".
In a statement last year, the council said that due to a low response rate in the Western Cape, "the repercussions of prevailing undercount are worrying for planning and service provision in the province". It added that "the risks of a significantly under-enumerated province is that the Census as a whole will be harmed".
READ | SA's population swells to 62 million - 2022 census at a glance
Writing in an op-ed this week the former stats-general said: "However high the undercount might be, does not diminish the validity of the count. This is because techniques of compensation through an independent post-enumeration survey (PES), are deployed to scientifically establish the extent of the undercount and adjust for it."
Many countries do not conduct a PES, and their census data are treated with scepticism. Many experts commend Stats SA for having always conducted a PES.
With that said, there is the possibility that some of the data might be restated, as was done in the 1996 census. Because not all the data has been released yet, as Stats SA usually does over a period of a year in stages, many analysts are waiting for more information before making any conclusions about the credibility of the data.
So, what's the controversy?
In an analysis written by the former chair of the statistics council David Everatt, the undercount "may set a new international record".
He adds that "it is safe to assume that with weighting on this scale - adjusting for an undercount of 31.06% - analysts may identify some confounding results."
Four researchers and analysts who looked at the first batch of results were surprised by some of the data, and questioned the veracity of the data.
Their greatest surprise was that the annual population growth for Johannesburg was 0.8%.
They said:
We can't conclude definitively at this time whether the 0.8% reflects a challenge with data or represents a startling new reality.
Graeme Götz, director at the Gauteng City-Region Observatory, highlights how surprised they were at some of the stats. He particularly highlights the Gauteng-specific data like migration, lower education levels and drop in informal housing.
In a rapid research paper the observatory produced on the Census 2022, it said: "Rather than simply confirming commonly understood trends, the data offers some real surprises."
Götz says that:
So, either Johannesburg numbers are just plain wrong. Or, alternatively, they may not be precisely right, but they say something really interesting. And the direction of change tells us something important is happening in the city and we need to be concerned about that.
Has controversy surrounded census data before?
Yes.
In a paper, Jeremy Gumbo and Clifford Odimegwu evaluated undercounting controversies in the past three South African censuses. The 1996 Census was "criticised as having underestimated children under 5 years and young male adults, as well as overestimating female adults"; while the 2001 Census was "criticised for both underestimating children under 5 years and about 400 000 whites". The 2011 Census was critiqued for "having been rushed and published prematurely and hence the counts were largely inaccurate".
The researchers concluded that "the processes and outcomes from these censuses are largely accurate" because the "adjusted counts were better estimates of the country's actual population relative to unadjusted counts".
For the 2011 Census, an internal dispute, revealed months later, threatened to compromise the integrity of the Census. Lehohla at the time said disciplinary action was taken against two Stats SA individuals because of "methodological and computational irregularities" which, if left unaddressed, "would have provided wrong figures of the Census by province as well as a wrong national total figure".
Other concerns raised for that Census was that there was a large variation in population estimates for some provinces.
