ELECTIONS 2019

PODCAST | The Numbers Don't Lie

A limited podcast series exploring the data behind the elections, pulling out the stories we didn't know were there.

15 April 2019 - 12:37
By Scott Peter Smith
The deciphering of election data is a complex task.
Image: Andrei Krauchuk/123RF The deciphering of election data is a complex task.

The numbers don't lie, but often their interpretation does. 

Paul Berkowitz, director of data analysis company Edgis, has taken on the challenge of deciphering electoral data before and after this year's national elections. 

In this introductory podcast we unpack the mechanics of data journalism just a little bit and talk about what kind of data, whether from polls or hard numbers from previous elections, we have available to us in South Africa.

We lay out the kind of stories we will be uncovering using some of his techniques. 

While Berkowitz says "I don't make predictions because I don't like to be wrong", that doesn't stop him from building a kind of probability model. So no predictions on final electoral numbers here, but he is sure to provide some insights not normally seen.

"Sometimes we make an educated guess when the data is limited. There is the science and there is the art," Berkowitz says. "The best practice is to make assumptions explicit," he adds, and to be clear about what we could be wrong about. 

But conclusions are only part of the journey of what data journalism can uncover. 

Keep a watch out for Berkowitz's stories in the coming weeks on the TimesLIVE website or under our Elections 2019 tab.