CSIR shows its world class prediction expertise in Elections 2019

Reading Time: 2 minutes

Three hours after the first voting district in rural Eastern Cape released its results on Wednesday evening,  the Council for Scientific and Industrial Research (CSIR) accurately predicted the number of votes the top three parties would get to, to within 0.8% of the final result announced on Saturday evening. The CSIR made the prediction with only 1098 of the total number of 22 925 voting districts declared.

The SABC and the CSIR have collaborated on predicting election results since 1999. The 2019 project forms part of the SABC’s mandate to bring South Africans the most up to date and accurate information, especially around national events such as an election.

The CSIR predicted the ANC would receive 56.72% of the vote. The party ended with 57.5%. A variance of 0.78%.

Its prediction for the DA was out by 0.39% and 1.29% for the EFF.

The CSIR based its predictions on less than 5% of the vote available at 02h56 on Thursday morning 9 May.

It also predicted about 17.4-million people would cast their votes, a prediction only 2.8% lower than the official final result.

The CSIR’s project leader working with the SABC’s Election Results unit at the IEC Results Operation Centre in Pretoria, Pravesh Debba, says the council’s election prediction model relies on two core principles relating to voting behaviour, namely:

  1. Voters do not randomly allocate their electoral preferences but are influenced by political, socio-economic and demographic factors, as well as past voting history; and
  2. Changes in voting behaviour between one election and the next are also not random, but are correlated with past voting behaviour, demographic and socio-economic factors.

These two principles combined allow the CSIR team to group voters (or rather voting districts) together based on their past voting behaviour (using a statistical clustering method) and to then expect that any changes to voting behaviour in the new election will be fairly similar within each group.