Jim Day
ADMINISTRATIVE problems and computer glitches have delayed the results of Census ’96 by six months, and pushed costs beyond its R365-million budget.
Preliminary results from the population survey were first expected in January – then February, then March. Now, officials say it is too early even to tell when results may be ready, though they are aiming for sometime in “the first half of the year”.
The Central Statistical Service (CSS) officials who ran the census have still to carry out the initial checks on the basic accuracy of the figures. Outside observers say they are not expecting to see any results until June, with final data not expected until November.
“Why the hell isn’t it out yet?” asked David Everatt, director of a social research non-governmental organisation, the Community Agency for Social Enquiry.
CSS head Dr Mark Orkin says that despite the delay, he is comfortable with the rate of progress.
“We’re certainly not going to compromise the integrity of the data by being hasty,” he says. People are not harassing him for results, he says, although he recognises the need for speed.
The CSS chief director of demographic statistics, Pali Lehohla, says he is confident the results will contain a margin of error of less than 10%. The degree of accuracy is vital to develop effective policy decisions, particularly for rural areas and informal settlements.
However, previous censuses had margins of error of 15 to 20%.
“We’re feeling the pressure to have the results come out fairly fast,” Lehohla says. “But circumstances were beyond our control.”
Those circumstances revolved largely around problems with transferring information electronically. First, the limited ability to send information led to enumerators – the people conducting the surveys in October and November – not getting paid on time. Administrators had to go back to manual operations before they could straighten out the problem by mid-January.
Then similar problems with sending data hampered the collection of census results. Officials have yet to determine whether data was lost due to the transferring problems, but Lehohla believes any loss was minimised by physically carrying census forms from dozens of regional offices to Pretoria to input the data into the CSS computers.
That raw data, as well as data from a smaller survey conducted to cross-check the accuracy of the count, has now been collected, but it will take months to compile any results.
The delays will increase the costs of the census, but Lehohla believes this will be limited to “some tens of thousands of rands”. By far the greatest cost of conducting the census involved payment to enumerators, which will not rise because of the delays.
The census has been plagued with problems. There were questions about the hiring practice used to recruit the 100 000 enumerators, insufficient training, lack of access to certain areas, the inability to count some segments of the population, inconsistent collection of surveys, and concerns – particularly among some white respondents – that some questions on the survey could be biased against them.
But observers say the census was a far more comprehensive attempt at surveying the population than had been tried or accomplished before.