Did you know that 83% of statistics are made up on the spot? This quip is as old as South Africa’s unemployment problem, but as with all clichés, underpins an element of truth. A more telling fact, though, is that 92% of people believe the statistics they read on blogs and in newspapers. Especially if you cite a reference (Fourie 2013). Which makes publishing (especially in print, where mistakes are more difficult to correct) a powerful medium to affect change, for good or bad.
For these reasons, it is problematic when public and private institutions begin to spoon-feed journalists, who are time-constrained and perhaps a bit lazy, rubbish statistics. After the release of the first estimates of the Census 2011 data (which was, understandably, widely publicized), some commentators raised objections to what seemed like strange demographic patterns. The immediate response from StatsSA was denial. Late last year, however, they acknowledged their mistakes and fired two staff members for making serious calculation errors. Even so, some suspect, errors remain in the data.
While these errors were unintended mistakes, Servaas van der Berg, professor in Social Policy at Stellenbosch University, writes in a recent blog post about the absurd unemployment statistics that are published by a private firm, Adcorp. Adcorp’s Employment Index estimates that South Africa has an unemployment rate of 5%, which is considerably less than the standard unemployment estimations of between 25% and 33% for the narrow definition of unemployment. It is also much lower than most of the rest of the world. Van der Berg points out that these errors are not only problematic because they are useless, but that they can actually influence peoples’ behaviour and result in the adoption of incorrect policies. He cite’s the high graduate unemployment that Adcorp calculates (above 50%!): if this is true, who would want to go to university if the general unemployment rate is only 5%? These numbers simply do not reflect reality.
But it is Van der Berg’s second contention – that such statistics may receive enough media attention to influence policy – that is even scarier to contemplate. And it may not be too far away: after a quick search, I’ve found references to Adcorp’s rubbish statistics in several of South Africa’s leading dailies: The Citizen (14 Aug 2012, p. 22), The Star (11 Oct 2012, p. 11), The New Age (11 Jan 2013, p. 13), Volksblad (11 Jan 2013, p. 8), Business Day (11 Oct 2012, p. 2). Only a few media reports reflect the tension between the generally accepted StatsSA data and the Adcorp estimates: the Mail & Guardian (12 Apr 2012, p. 7) and an especially well-written article in the Financial Mail (2 Mar 2012, p. 43).
The wrong Census 2011 statistics can have serious implications for how national government appropriate funding to the various provinces, for example. As Tom Moultrie, UCT demographer, notes in a remark on one my previous posts: “If you were an MEC in the Northern Cape, how would you plan rationally, if StatsSA’s own estimates suggested that the population of your province lay somewhere, with 95% certainty between 984 000 and 1.307m? (That’s a 14% range either side of the point estimate of 1.146m!).” Similarly, the calculation of incorrect unemployment figures may suggest to government that their policies were successful when in reality they were not, or vice versa.
South African editors and journalists are powerful agents and should be more diligent in their assessment of credible news sources. Adcorp’s Employment Index should be off their list.