Student Creates Presidential Election Forecast Model Used by CNN

TUSCALOOSA, Ala. – The chances President Donald Trump is reelected in the 2020 presidential election are about the equivalent of rolling a six on a six-sided dice, predicts University of Alabama sophomore Jack Kersting.

Kersting, a 19-year-old native of Richardson, Texas, studying economics, created political forecasting models that gained recognition for their quality and accuracy, and have been used by major media outlets such as CNN since 2018.

“Before that I did the Democratic primary a year ago on June 1,” he said. “I was the most accurate out of all the forecasters. Towards the middle of it and after Nevada, I had Bernie Sanders favored to win, but on the vote projection I had the least error for each state.”

As of June 26, his evolving presidential forecast model had Democratic candidate Joe Biden winning in a landslide victory of 85.4% to Trump’s 14.6%.

And the reason he creates the models? Just for kicks.

“I just do this as a hobby,” he said. “It’s something to do.”

Kersting started political forecasting during his senior year of high school for a statistics class. He started with the 2018 mid-terms and enjoyed it so much he decided to continue creating them.

“This takes coding, which is something I picked up by myself,” he said. “My initial models were all in Microsoft Excel. I would copy and paste them in Adobe Illustrator, which looks decent, but it takes a long time. Now I code everything in JavaScript and HTML.”

As part of his models, each state has a fundamentals and a partisan vote index, which is how much more Republican or Democratic the state is compared to the nation. For instance, Texas, his home state, leans about 11% Republican because its voters selected Trump by 9% margin in 2016 while the nation voted for Democratic candidate Hillary Clinton by a 2% margin.

The models also average job approval rating, national polls that consider registered and likely voters, job approval, the economy, state polling, state similarity, polling from other states and experts ratings.

Kersting said it takes him about 30-40 hours to complete a political forecasting model. Once he completes them, he shares them online in a niche group called “Election Twitter.” They catch on from there.

Kersting doesn’t  agree with all of his models predictions, but he stands on their accuracy.

“I don’t think Biden is going to win by an 80% plus landslide because Florida and North Carolina can switch to red. But the math is pretty good.”