During the 2011 local elections in May, the CSIR election forecasting team's precision was again demonstrated by its early national predictions. After 1% of the national votes were in, the CSIR forecast that the ANC would get 62% of the votes, and after 14% were in, that the party would get 62.8 %. This number stayed constant and was, in fact, the final percentage released.
For the DA, the CSIR predicted 26% after 1% of the votes were in, and 24.2% after 14% of the votes. Again the number did not change, with it being the final official percentage announced.
"The accurate predictions again confirmed our confidence in the forecasting model that the CSIR has developed and updated over time," says Hans Ittmann, spokesperson for the CSIR's election forecasting team.
Unlike political analysts who comment in advance on possible trends, CSIR scientists use rigorous statistical modelling based on the first actual voting results that are received to do forecasting.
One of the new challenges that the CSIR faced team during the past elections was to take into account votes for the high number of people who had registered as independents, but actually represented a political party.
"We experienced some usual glitches as the communication with the database is not always impeccable. But that makes life interesting and prevents the coverage from being routine. A few elements of our program could also be tested only once real data came in, which meant we had to make last minute changes in our software, in real time," reports Dr Jan Greben, project leader of the CSIR's election forecasting team.
In addition to Greben and Ittmann, the CSIR election team consisted of Dr Chris Elphinstone, Dr Peter Schmitz, Jenny Holloway, Maanda Rasuba and Rosalie de Villiers.
"The CSIR team again showed tremendous dedication during the three weeks running up to the elections. When forecasting started as soon as votes began coming in, it was all systems go, right up to the end of the three-day period," says Ittmann.
The team operated from the IEC headquarters in Pretoria and worked in shifts, around the clock, to make the forecasts available to the nation through the SABC and other media present.
"The CSIR watched the race for votes keenly on a national level for the 10 biggest parties when early forecasting data came in. A few of the metros, like Cape Town, Tshwane and the Nelson Mandela metro, were also of high public interest due to contesting political parties," Ittmann comments.
Very early predictions by the CSIR when 9-11% of the votes were in for these metros, compared to the actual results were:
- Cape Town - DA 54.62%, ANC 36.72% (actual results - DA 60.92%; ANC 32.8%);
- Nelson Mandela metro - ANC 54.49%, DA 39.12% (actual results - ANC 51.91%; DA 40.13%);
- Tshwane - ANC 56.93%, DA 36.59% (actual results - ANC 55.32 %; DA 38.65%).
For plots of the convergence of CSIR predicted versus actual results for other political parties, see graphs:
The CSIR's election forecasting experts were contracted first by the IEC a number of years ago, and since 2000, by the SABC, to carry out election night forecasts for the national and local elections.
* The CSIR also uses its statistical modelling capabilities for supply and demand forecasting relating to service delivery and to forecast long-term energy demand, among others. Read more