In Business Day on Monday UCT's Anthony Butler wrote a perceptive article explaining how the African National Congress has rescued itself from the fairly desperate situation it found itself late last year. There is a real possibility, he notes, that the ANC "will run away with the forthcoming elections" and even gain the symbolically and constitutionally important two-thirds majority.
Certainly, the deluge of ANC advertising is drowning out the campaigns of many of the other parties. The ruling party has recovered its internal cohesion for the campaign, and does seem to have all the momentum going into the final weeks of the election. The sense one gets from the media is that the ANC's support stands somewhere in the mid 60s. And, the Western Cape is the only province where the ANC is expected to be overtaken by an opposition party in terms of popular support.
But when it comes to objectively judging how the different parties are doing there is actually very little evidence to go on. The two sources of information we do have are opinion polls and by-election results. There are problems with both of these.
Unlike in the United States we do not have regular opinion polls in the run up to the election. The ANC and DA conduct their own tracking polls, the results of which they leak to the media when it suits them. While these may give some indication of shifting trends in voter support, it is not clear how accurately they can predict the ultimate election result. In a fractured country such as South Africa opinion surveys are time consuming, expensive and difficult to do properly. Only two companies - Markinor and Markdata - have the kind of track record which inspires confidence in their reliability. Neither have recently published the results of a survey of party political support.
A relatively new polling outfit, Plus 94, conducted a survey of registered voters between March 11 and March 21 2009 - the results of which were published last week. According to a press release, "The study was conducted in all 9 provinces using face to face, in-home interviews. A total of 3500 homes were covered among registered voters." 60,8% of those interviewed said they planned to vote ANC, 15,9% said they would vote DA, 14,9% COPE, 1,8% UCDP, 1,6% IFP, and 1,5% ID. The poll seems to have been conducted in a somewhat hurried fashion, and some of the results, particularly at provincial level, are dubious. The IFP's estimated support in KwaZulu Natal (7%) is surely far too low, COPE's estimated support in Mpumalanga (19,9%) and Limpopo (33,7%) much too high.
Even the most rigorously conducted surveys can go wrong, if respondents are reluctant to honestly say who they will vote for. This has been a perennial problem for pollsters when it comes to evaluating the IFP's support. Shortly before the 1999 elections Markinor conducted a survey which found that the ANC enjoyed the declared support of 49% of those polled in KwaZulu-Natal. Only 17% of respondents said they would vote IFP, with 14% undecided. In the event, the IFP received 41,9% of the vote in the province, as opposed to the ANC's 39.4%. Another Markinor poll conducted before the 2004 elections found that 48,7% of respondents said they would vote ANC while 24,9% said they would vote IFP, with 13,2% undecided. This poll was more accurate when it came to measuring ANC support; it went on to win 46.98 % of the provincial vote on April 14. But it still massively underestimated IFP support: that party received the support of 36.82% of the electorate in KZN. It will be interesting to see whether the April 22 election will expose similar problems when it comes to accurately measuring COPE's support.