On October 30 2012, the first sets of Census 2011 data were released. Further data is still to come but a large majority has been presented to the country.
Pali Lehohla, statistician-general and head of Statistics South Africa (Stats SA), says that there are standard practices and processes that are part of defining a successful census such as data quality.
The Stats SA Data Quality Report communicates the processes that were followed to assess the census data. Each variable was analysed according to an international standard to assess accuracy and quality of the information.
The United Nations's standard for statistics considers any variable that shows an imputation rate of less than 10% acceptable and less than 5% very good.
The report shows a summary of imputation rates with the majority below 5%. The Roof Construction Material variable is at 10%. The only variable that is above the 10% standard is Estimated Property Value, at 29.1%.
"The problem with Estimated Property Value is that people's responses must be taken at face value," says Lehohla. "Although there isn't much that can be done at the enumerator phase, we can improve on the credibility of the data through using the banks's market value reports." Lehohla says that all the tests show that the data is good quality.
The Census 2011 undercount is 14.6%. Lehohla says that this is a concern and that the number is high. He notes that the data is adjusted to compensate for the undercount through the post-enumeration survey — but why is it so high?
"I believe it's due to a challenge that we were aware of from the beginning, that people are not located at home throughout the day and travel to and from work," says Lehohla. "It was a real concern but the solutions were not amenable to the South African context. Is it possible in a democracy to call a curfew as is done in Turkey, Nigeria and Iraq? The other option is to declare a public holiday."
Another reason for the high undercount, in comparison to other countries, is that other nations impute persons for missed dwellings where there is evidence that those dwellings are usually occupied. "This reduces the undercount," says Lehohla.
In South Africa, treatment of missed dwellings differs because people in the dwellings are not accounted for but calculated as an undercount.
Lehohla says that a true comparison of undercount in South Africa with that experienced in other countries would be undercount minus imputed persons in whole households missed in the census.
During and prior to the enumeration phase, Lehohla was quoted saying that he hoped for a 2% undercount. He says that initial surveys showed that 95% of people would be willing to participate.
"Post the undercount calculation, we realised that willingness to participate is not the problem but the ability to participate is. When you have priorities such as going to work, good intentions are overpowered," says Lehohla. "We will take these lessons going forward."
When it comes to determining a successful census the two critical areas are data quality and the undercount.
"The undercount is a measure of the success of a census, but it doesn't degenerate the quality of the information. However, it would have been a much better census with an undercount of less than 10%," says Lehohla.
According to him, challenges with the higher undercount really show up with very small area analysis. A 14.6% undercount is acceptable at national and provincial level but not when dealing with statistics for a small area.
Howard Gabriel, chairman of the Statistics Council, concurs: "Obviously we would have liked the undercount to be lower but given the methodology used the census still created a data set that is fit for use especially at policy level. It is only at the lower levels that you need to exercise caution."
Other success factors
Lehohla says that the quality of a census is also about transparency, that it is clear what has been done, how it has been done and what the limitations are.
Public participation is another measure and includes the use of the census results. Dissemination of the data and public access are integral to this measure.
"Dissemination continues to improve and there has been good uptake by the media including healthy debates," says Lehohla. Further dissemination strategies are being planned by Stats SA.
There have been concerns around the age distribution figures which show an unexpected increase in births across all groups. Lehohla acknowledges that the results are more than expected but says that the soon-to-be-released fertility and mortality data will assist in understanding the dynamics.
"There was no reason to hold the census data back before getting explanations. The data is good, but the causes that underlie the increase need to be researched," says Lehohla.
The census figures have been confirmed by independent studies such as birth and death registers adjusted according to the post-enumeration survey and for coverage.
Population distribution has also come under debate because the 2011 mid-year population estimates pointed to KwaZulu-Natal having the highest population count. Census 2011 showed Gauteng to be the largest province with 12-million people.
Lehohla says that the census was done by counting in units of 150000, independent of provincial boundaries, and that the process provides no bias.
"The accuracy of the results has been verified by the Statistics Council, the methodology and quality measures are governed by international standards and the process has been transparent," says Lehohla.
"The discussions around Census 2011 are healthy. It also allows for information to be understood better. Census 2011 has been successful and has provided lessons that we will take into the next census."
Get the census information
The census data can be obtained on the Statistics South Africa (Stats SA) website (www.statsa.gov.za). Stats SA's existing mobile app is set to be updated with an innovative feature — My Ward, My Councillor for more indepth interaction on a ward level. This can be downloaded from the website.
The public has access through media reports, breakfast sessions and printed copies available at district offices. Members of the public can call or email Stats SA for a copy of the census results at
[email protected] gov.za or (012) 310 8600.
Stats SA runs programmes for journalists, including statistics workshops, and provides assistance with interpreting data. A schools programme engages children with the census results and encourages further dissemination to parents. Contralesa has also agreed to disseminate information to the traditional areas and Absa Bank will assist by releasing results using ATM screens across the country.
Stats SA is currently planning further dissemination of the census data beyond what is mentioned above.
Data to be released
Although a large portion of the Census 2011 results has been released, more information is still to come. Work is being done at place-name level and the intention is to release the data at the end of March 2013. Further tables on fertility, mortality, occupations and industry are still to be released. This data goes beyond population count and deals with indepth questions which require further analysis and editing.
There will also be monographs (analyses or descriptive pieces of the data). In this case, the aim is to give additional benefit through interpretation.
This article was supplied and approved by the Mail & Guardian's advertisers. It forms part of a larger supplement.