‘We can’t throw the entire Census 2022 in the bin,' says Stats SA official

Statistics South Africa (Stats SA) has doubled down on its defence of Census 2022, saying its figures are credible and rejecting calls from the academic and scientific world for a recount.

Stats SA officials were forced to mount another defence of the census figures after the data collection agency decided this week not to release key data, including figures on mortality, fertility, employment and household income, on the advice of its technical experts.

The data are important because they would supplement Census 2022 with more information that can be used by researchers, policymakers and government departments — which often use the information to understand the fabric of society and make taxpayer fund allocations to service delivery programmes.

Stats SA’s decision not to release the supplementary data has given demographers more suspicions about the integrity of Census 2022, which was conducted in February 2022, during the Covid-19 pandemic, and officially released to the public in October 2023. The census cost taxpayers R2.3-billion.

In an interview with Daily Maverick on Thursday, Solly Molayi, Stats SA’s acting deputy-director general responsible for population and social statistics, said not publishing the key data was a normal practice by statisticians when reporting and coverage biases were identified.

“When you look at the census, we have released most of the variables and data to the public. And the data we have held back, we wanted to ensure that we have done our quality assurance,” said Molayi.

“That is the reason why we took a bit of time to go through the mortality,  fertility, employment and household income data. This is not unique to the census, but it is also done for other surveys such as the quarterly labour force and income surveys. There are variables that we have not released.”

Molayi said Stats SA’s decision to withhold key information was informed by fewer people who responded to its questionnaires when conducting door-to-door surveys. And if released, the data would give an incomplete picture of trends and “be misleading the public”, he said.

“We are not going to release data because we are collecting data. We go through our processes and one of our processes is to ensure we have quality data.”

Alternative data reliability proposals

In the interim and as an alternative to the key data that were withheld, Stats SA has referred researchers, policymakers and government departments to its mid-year population estimates. The estimates contain fertility and mortality data based on the administrative records — meaning that they are based on the data that usually are supplied by the Department of Home Affairs on trends such as recorded life and death when birth and death certificates are issued.

To estimate trends in employment and household income, Stats SA has referred demographers to its yearly Income and Expenditure Survey and Quarterly Labour Force Survey.

Stats SA has faced intense criticism from UCT demographers Rob Dorrington and Tom Moultrie, who concluded that Census 2022 produced several anomalies, with the pair saying that it was not fit for purpose.

Their central criticism was that there was a huge 31% undercount of the population in the census of 2022, compared to an undercount of 14% in the census of 2011, which was already high by global standards.

They said it was important to understand what went wrong with Census 2022 before planning began for the next census, which is set to be conducted in 2031.

Read more: South Africa’s 2022 census ‘unfit for purpose’ — demographers explain what went wrong

Stats SA’s Molayi said the agency had already admitted to the 31% undercount — something that was discovered through its post-census release initiatives.

“After the census was done and released, we ended up going back to communities to check if there was any household that we had missed. Based on that survey, we have realised that the census has missed households and came up with the 31% undercount.”

Molayi said there would always be limitations to the data collection methods, considering that Census 2022 faced the challenge of being conducted during Covid-related lockdowns.

“When we conduct a survey by going into a household, we ask how many people were born. We do not ask for a birth certificate as evidence. That is not how the survey works,” said Molayi. He added that the criticism around the undercount was unfair because some demographers used their models that rely on birth, death and net migration data as inputs, without directly surveying households.

Molayi said Stats SA planned to continue to refine data that the agency had withheld to determine mortality, fertility, employment and household income trends, not only at the national level but also at a municipal level.

Asked whether Stats SA was prepared to redo the census, Molayi responded with an emphatic “no”.

“We have an understanding of the population. The undercount doesn’t mean we must write off the census and throw it in the bin.

“In the census of 2022, there is a household section looking at data indicating access to water, education, migration trends and people living with disabilities. Those are solid sections. So, we can’t just say that because the four variables [mortality, fertility, employment and household income] have been excluded, therefore we have to throw the entire census in the bin.” DM