2020 Election Predictor

Harry Truman holds up a copy of the Chicago Tribune with the now-iconic headline.

Remember all those election predictor models that showed Hillary Clinton with a 98% chance of winning? Those screen-grabs became modern-day equivalents of ‘Dewey Defeats Truman.’

Such as the Huffington Post…

Huffington Post gave Hillary Clinton a 98% chance of winning the election. The leftwing news aggregator claimed Donald Trump had just a 1.7% chance of winning.

And the NY Times…

The New York Times’ Upshot Model gave Clinton an 85% chance of winning the presidency, with Trump at just 15%.

And Five-Thirty-Eight

Five-Thirty-Eight had Clinton winning an Electoral College landslide, garnering 302 ECVs. The much hyped model only gave Trump a 28.6% chance of winning.

Newsweek even went to print with Madam President Clinton on the cover.

The Newsweek cover of Madam President became an instant collectible.

Despite the widespread embarrassment, these same data-centric prognosticators are spewing out the exact same models with the same laughable predictions.

When called out for their bias, they claim that they’re simply crunching the data that’s available. This is pure spin.

Data integrity is of the utmost importance in predictive analytics, and they know it. In computer science there is a basic concept called Garbage In, Garbage Out (GIGO). Its theory says that bad input will produce bad output.

The reputation of Nate Silver took a huge hit, as the vaunted numbers cruncher became a laughingstock on the right.

And that’s the problem with all of these models. The input is flawed which makes the output total garbage.

It doesn’t matter how well designed and/or intricate predictive models such as the Cook Report or Five-Thirty-Eight appear. If the data is garbage, its output will be as well.

This principle applies more generally to almost all analysis and logic – in that an argument is unsound if the premise is flawed.

With that, Politipage will be producing its own model which will take into account all kinds of public and private data, including, but not limited to, public polling (not skewed toplines), 2016 election (pre-polls, exit polls, and official results), focus groups, social media listening, enthusiasm indexes, narrative data, campaign travel maps, economic data, stock market, gas prices, and more.

It debuts on Tuesday, October 14, 2020 and will be updated daily, through Election Day, November 3, 2020.

LEAVE A REPLY

Please enter your comment!
Please enter your name here