The Council for Scientific and Industrial Research has stated that data modelling and statistical analysis play a key role in unmasking South Africa’s (SA) voter behaviour and election results.
The Council’s Impact Area Manager, Professor Pravesh Debba, says their election prediction model provides deeper insights into shifting voting patterns and the final electoral outcomes long before the counting ends.
Debba was speaking on the eagerly awaited Local Government Elections.
He says the methodology behind these early projections highlights why mathematical accuracy and transparency are critical for responsible.
“We aren’t doing the verification as such, but, you know, as the data starts coming in, and, you know, what we are trying to do is to observe when we get to a position that the model is stable and giving a consistent result. So, when we are getting a consistent result, then that is when we communicate to audiences, to the journalists, and so on, the prediction at various different levels. So, we look at it at a national level.”
“So, what’s the outcome at a national level? And for that you need 5% to 10% of the voting districts declared. But then we also go down to the provincial level, and at provincial level usually you need a bit more. So, we usually need about 10% to 15% of the voting districts declared.”
–SABC–
