LETTER: Apologies are in order over attacks on pollsters

I think it is time for those who criticised market research institutions in the run-up to the elections to extend them an apology. Both Ipsos and the Social Research Foundation (SRF) were remarkably accurate in their national findings.

Published by Business Day on April 26, Ipsos estimated ANC support of 40.2%, the DA 21.9%, the EFF 11.5% and uMkhonto weSizwe (MK) 8.4% of voter share.

In the case of the SRF (May 27), based on a 60% voter turnout, it was ANC 42.3%, DA 23.4%, EFF 11.5% and MK 11.8%. The actual figures were ANC 40.1%, DA 21.7%, EFF 9.4% and MK 14.8%. No-one can say there were hugely surprised by the final result.

Survey research will almost always have some degree of error as, unlike a census, it depends on a sample to be drawn. The April 2024 Ipsos survey relied on data from only 2,545 registered voters — only 0.00009% of the total. Yet its findings were remarkably accurate.

The reason for the validity of their results is the sound survey methodological principles that they use. One key principle of scientific survey research is to draw a representative sample – not a large sample. Research institutions know that samples do not have to be large to be representative. If, in the future you hear someone criticising survey results on the basis of sample size, close your ears. A questioner at the HSRC media conference broadcast on Newzroom Afrika on Friday morning is a classic example. In survey research, larger does not necessarily mean more accurate.

Market research is not for the faint-hearted in an environment of voter uncertainty, vast, some almost inaccessible rural areas, some degree of illiteracy, voter turnout uncertainty/abstentions and voters who one cannot be sure that they are not lying to fieldworkers. So, thank you to Ipsos and the Social Research Foundation for a job well done.

C Boshoff
Via email

JOIN THE DISCUSSION: Send us an email with your comments to letters@businesslive.co.za. Letters of more than 300 words will be edited for length. Anonymous correspondence will not be published. Writers should include a daytime telephone number.