Count us out

There were howls of disbelief when the Independent Electoral Commission announced on March 4 that it had reached its target of 70 per cent voter registration by dint of a reduction in the number of South Africans aged 18 and over eligible to vote. Originally the IEC had talked of there being 26.3 million potential voters and had even mentioned figures of up to 26.9 million, though this was later said to have included 16 and 17-year-olds who would be too young to vote on June 2. But on March 2 the IEC’s chief executive officer, Mandla Mchunu, revised the potential electorate down to 25.5 million. Then following the meeting on March 4 between the representatives of the IEC, the department of home affairs and Statistics South Africa (the renamed Central Statistical Services, responsible for the census) it was announced that there were only 24.6 million South Africans aged 18 and older on February 28 who were entitled to vote.

This reduction of 900,000 from the figure Mchunu had given only two days before, let alone from the higher figures issued before that, was not explained. The meeting then cut the figures by a further 2.2 million by excluding permanent residents who are not citizens (581,000), those abroad on a permanent basis (97,000), convicted and awaiting trial prisoners (150,000) and those without a bar-coded ID (1,358,000). This produced a figure of 22.4 million — to which were then added the 400,000 17-year-olds who would turn 18 before June 2, giving a final total of 22.8 million.

This was a curious result for when, in the run up to the 1995 local elections, local authorities were asked to submit figures for the maximum number of potential voters in their areas, this produced a total of 26,496,796 voters. The CSS said this was too high and insisted instead on a figure of 23,363,205. Now four years later, Statistics South Africa (SSA) has agreed on a figure over half a million lower than that — although SSA estimates that South Africa’s population is growing at over 850,000 a year.

It is not the first time that these figures have been massaged. As Graeme Gotz showed in his report on the 1995 local elections, Buying in, Staying Out, (Centre for Policy Studies 1995), as those elections approached and registration stayed disappointingly low, great pressure was exerted on municipal authorities to reduce their estimates yet again. “Second-guessing the original estimates”, wrote Gotz, “opened the door to projected potential voters being used indiscriminately as a fine-tuning variable to achieve an acceptable picture of registration. In the process the suposedly correct CSS figures were in turn second-guessed and discarded where inconvenient.”

Last November, Mchunu unwisely committed himself to the view that a registration of less than 80 per cent would mean that democracy was “limping”. As it became clear that registration was going slowly, the embarrassing possibility loomed that the inevitable ANC victory would appear flawed by a low turnout on a low register. The ANC’s opponents would be able to claim that despite winning the election, the government had suffered a humiliating drop in popular confidence and consent. This would be a most inauspicious start for Thabo Mbeki’s presidency and a situation that he would want to avoid at all costs.

To those IEC-watchers who noted some time ago the commission’s great sensitivity to what might embarrass Mbeki, it came as no surprise that there should have been a sudden rush to reduce the size of the eligible electorate, thus exaggerating the rate of registration and, ultimately, the levels of political consent and confidence in a Mbeki government. Nor did the sight of census officials taking part in a meeting which somehow bargained the size of the electorate down by 2.7 million from the figure given by the IEC’s chief executive only two days before do much to enhance the image of Statistics South Africa (SSA). However, those who had observed the curious census of 1996 were not entirely surprised.

This was to be the first “democratic census”, a stark contrast to the 1991 apartheid census, vilified because it had excluded the homelands and used sample surveys and aerial photography to count areas regarded as inaccessible. Typically these were squatter camps. Particularly abhorrent was the idea that squatters should be counted in just the same way that conservationists establish a herd count — flying over them in light planes and applying a multiple (of just under 5.5 people per shack). In fact the evidence suggested that this was not an inaccurate means of estimation. Its final figure, including an estimated 6.8 million for the homelands, was 37.8 million. The Development Bank of Southern Africa (DBSA), which carried out its own population estimating process, concurred almost exactly, with a figure of 37.6 million for 1991.

The 1996 census was a heavily ideological exercise of a different kind. Originally it was announced that there would be no classification by race, sex or class. On second thoughts SSA said that such classifications would be used, but only in order the better to track inequalities and to chart the demise of racism and sexism. There would be no aerial photography and not only would the census questionnaire be printed in all 11 official languages but, theoretically at least, every household involved could choose which of the 11 languages it would like to be interviewed in. The enumerators — some 100,000 of them — were almost all African and were hastily trained in what was clearly an employment-creation exercise. This notion achieved sufficient currency for enumerators to meet considerable intimidation from groups of unemployed who had wanted jobs as enumerators. Questionnaires were burnt, two enumerators were raped, six more died in car accidents, five had their cars hijacked, one had his car shot at and another his car set alight. Despite all this — and a failure to meet the census deadline — the census was celebrated as “the great counting” with a big handing-over ceremony to President Mandela.

Anyone with experience of such matters quailed at the thought of what 100,000 barely trained enumerators might achieve. When SSA came out a year later with a total count of only 37.9 million there was a roar of disbelief, despite SSA’s assurance that this figure already included an upward “adjustment” for undercounting. The DBSA had already estimated the 1996 population at 44.4 million and other estimates had been in the same ballpark. In rapid succession critics showed that the SSA figure was inconsistent with the 1994 election result, with the school population and with the typical sex ratio of middle-income developing countries. The Congress of South African Trade Unions, convinced that black workers were being seriously undercounted, sent a delegation to present its grievances to SSA, which obligingly revised its estimate up by 2.7 million to 40.6 million. Most informed critics continued to believe that this was still at least one or two million too low.

The 1991 census put the white population of South Africa plus the non-independent homelands at 5,068,110. With an expected growth rate of 0.8 per cent (DBSA) or 0.67 per cent (SSA) the Institute or Futures Research estimated the 1996 white population at around 5.2 million and the DBSA at 5.35 million. SSA, on the other hand, came up with a figure of only 4.4 million. This was by far the most dramatic figure in the census. Either whites had been undercounted by up to a million (nearly 20 per cent) or else the 1991-96 period had seen white emigration of up to 200,000 a year (nearly 600 a day), figures way beyond the reckoning of even the most pessimistic observers at that point. Without much doubt inexperienced black enumerators had simply not counted large numbers of whites. SSA said that not enough whites had come forward to act as enumerators, ignoring the fact that it had clearly signalled that such jobs were part of a new South Africa empowerment exercise and, as such, were not for whites. This apparently large underestimate of the white population meant that the census was useless in determining the extent of the mainly white braindrain of professional and managerial talent.

Despite the promise of a “great counting” in the end SSA’s figures owed a great deal to continuous adjustments: the first to get to the 37.9 million figure, the second as a response to trade union and other criticism and most recently in March to help the IEC reach a politically desirable reduction in the potential electorate. Other problems have continued to surface. It is still unclear how the original census could find an excess of 1.5 million females over males in the population, particularly given the heavily male migration from neighbouring African countries. The minister of agriculture, Derek Hanekom, pointed out that SSA’s figures suggested a shrinkage in the agricultural workforce that was far too rapid and large scale to be believable.

Disquiet also grew over the accuracy of some of the economic data provided by SSA, for example over the national balance of payments. In late 1998 the auditor-general issued a report critical of the way SSA collected and disseminated various other types of economic data, suggesting that its methods for calculating such basic figures as gross domestic product and the consumer price index were flawed and could have misleading results. SSA indignantly insisted that the auditor-general’s report was based on the 1996 situation and that these shortcomings had since been remedied — but even this rebuttal turned out to be wrong, for the report had been based on an audit carried out in 1997.

It took the gyrations in the figures for the electorate, however, for many to realise that there was a worrying fluidity to even the most fundamental official figures. Subsequent developments have hardly restored confidence. Mchunu announced that 18.34 million people had registered to vote, taking the registration figure up to a triumphant 80.55 per cent. This final increase had occurred because an extra 400,000 names had been “discovered in the system” and added to the previously released figures. However, 296,266 names were to be struck off the register — 31,327 because they had died, 39,498 because they did not have bar-coded IDs, 109,441 more because they were not South African citizens or were not on the population register, and another 116,000 for unspecified reasons. Quite how all these people had been registered in the first place is not clear. Meanwhile, several dozen of the “dead” have indignantly come forward and had their identities confirmed. Some 36,000 of those struck off the register have been reinstated on appeal, adding to the sense of chaos.

Even the resulting figure of 18.08 million registered was not final, because last month the Constitutional Court ruled that denying the vote to prisoners was unconstitutional. Inevitably, angry comparisons were drawn at the notion that convicted rapists and murderers should have the vote while, for example, the 581,000 permanent residents should not. Even more remarkable was the disenfranchisement of South Africans travelling abroad — businessmen on foreign trips, students acquiring degrees and skills abroad, both to the country’s ultimate benefit, or even touring sportsmen such as the national cricket team, actually representing South Africa abroad. For the Constitutional Court decision meant that assassins such as Clive Derby-Lewis and Janus Walusz were to be allowed to vote while national heroes like Hansie Cronje and Jonty Rhodes were not. This contrast was so unsustainable that further concessions to allow at least some South Africans who are abroad on June 2 to vote became inevitable.

The sheer fluidity of the baseline figures provided by SSA, the department of home affairs and the IEC, as well as the occasional contradictions between them, mean that it is never possible to say beyond all doubt that any particular figure is right or wrong. Indeed, while much of the criticism of the IEC has suggested that its figures for the size of the potential electorate are too low, the opposite may be the case. After all, the IEC’s final figure of 22.8 million potential voters allowed for only 1,358,000 disenfranchised through lack of a bar coded ID. But there is strong survey evidence to back up the claim by the DP and the NNP that the real figure for those disenfranchised is around 4 million.