Legacy media reporting on the forthcoming November elections is making four basic mistakes about poll data, its interpretation, and forecasts.

The first mistake is that political journalists continue to compare turnout-modelled polling from the Social Research Foundation (SRF) with raw political data from Ipsos to drive a debate about which set of numbers is right.

The comparison is ignorant of what the two sets of numbers measure.

The SRF polling applies a turnout model to determine the election-day decisions of voters. Ipsos, by contrast, reports just raw political support among registered voters. Given that Democratic Alliance (DA) voters are significantly more motivated than African National Congress (ANC) voters, for example, the turnout equation shifts the election data result far from the raw voter support base numbers.

For a political journalist to compare the two as if they are measuring the same thing is akin to an economic journalist comparing real and nominal GDP numbers and then starting a furious debate about which is right.

The second mistake is that the media report that the SRF and Ipsos have predicted that the election result will be this or that. Neither institution has done anything of the sort.

Both have produced data, modelled and raw, at the moment in time when their polls were conducted. Ipsos and the SRF routinely state that their latest figures should not be interpreted as a prediction of the final election outcome.

For a political journalist to present the data as a forecast is akin to an economic journalist taking a first-quarter GDP growth number and reporting that the institution which produced it has forecast that growth for the full year will be the same.

A third mistake is that legacy media journalists don’t account for large swings in support trends taking place over very short periods of time as the ANC starts to come apart.

The SRF, for example, has polled the ANC in Johannesburg at under 20.0% and nearer 30.0% this year. Which number is right? The answer is both — at the times they were collected. What is happening is that, as the ANC comes apart, its voters are looking to the uMkhonto weSizwe Party, DA, and Economic Freedom Fighters, even the South African Communist Party, and then back to the ANC. Depending on the day and week that they are polled, this may yield significant swings in the poll data.

For this reason, in particular, the SRF cautions against taking its current data as a forecast of anything. (The SRF does produce forecasts at times, but these will be labelled explicitly as that.)

The fourth mistake is not to account for overhang. Even if the final voting data could be known, this will not necessarily translate accurately into the balance of power in each municipality.

The reason is that every municipality in South Africa is split essentially 50/50 between ward seats and proportional representation seats. Voters have distinct ballots for each.

The party with the most votes in a ward wins that seat even if its tally falls far short of a 50.0% majority. Given how fragmented ANC support has become in many wards, parties could therefore win seats with perhaps 20.0% or 30.0% of the vote in the ward in question.

The remaining proportional representation seats would be allocated  proportionality.

Hence, to make an extreme example, a party with nearly a quarter of the vote could hold nearly half the seats in a council, while a party with nearly half the votes could end up holding a significantly smaller proportion of the seats.

Simply knowing which party receives the most votes, therefore, does not necessarily tell you which party will control the council. That is par for the course in a ward system.

A final layer of complexity is that the Constitution prescribes that the balance of power in a council must broadly reflect vote proportionality. It is plausible that, in councils with a significant degree of overhang, court challenges may flip the balance of power months after the election has concluded.

Between the ignorance on turnout and raw data, the misstating of predictions and forecasts, the failure to appreciate the extent of swings around the ANC base, and a lack of awareness of the implications of overhang, a lot of legacy media reporting to date has been less than useless in helping to explain what is happening in South Africa’s politics and why.