Yahoo’s Prediction Engine Calls Presidential Race In Favor Of Obama (Sorry, Mitt)

Using smarty-pants economists and computer scientists, Yahoo analyzed historical data and says the math points to the prez taking the White House this fall.


An entire industry has emerged to try to predict who will triumph in any particular political contest. Polling is huge business. But a team of scientists at Yahoo say that, when it comes to the U.S. presidential race at least, who candidates are and how they campaign almost doesn’t matter. What does matter are factors over which they generally have very little control, such as economic indicators and the electorate’s overall political ideology.

By studying the last 10 presidential races, and corresponding economic and social indicators, the scientists have developed a model that predicts that President Obama will win the White House this November (assuming certain indicators don’t change dramatically between now and June).

Their current projection gives 303 electoral votes to the president and 235 to the Republican contender. Yahoo will launch the predictive tool to the public later today.

The tool’s model draws on data like the president’s approval rating, the rate of growth of income, and the ideology of the electorate in individual states.

What’s important to note, David Rothschild, an economist with Yahoo Labs, tells Fast Company, is that that, according to his analysis, when it comes to certain indicators, the rate of change is much more important than the absolute levels. And the most important data point is state-by-state income growth. (The model has to look at states individually, since electoral votes–and not overall popular sentiment–determine the actual outcome of the election.)

So even though overall economic indicators are still comparatively dismal, they seem to be going up in most states, which, according to the model, weighs the race in favor of the current occupant of the Oval Office.


One of the scientists’ most important conclusions, based on their analysis of historical data, is that the candidates’ actual campaigning plays very little role in the final tally. “Variables beyond their immediate control describe the outcome very well,” the team writes in a post which will go live on Yahoo’s political blog, The Signal, at noon PT today. “A brilliant or lucky campaigner is at an advantage, but the net impact of politics and strategy averaged over the past 40 years is just the small variation that this model cannot predict.”

Yahoo says the model has a 88% success rate when compared against historical data, with a margin of error of 3 percentage points.

Rothschild says Yahoo won’t be able to make a final prediction about the race until late spring. That’s when they’ll be able to collect the last relevant piece of data: the president’s approval rating in June. According to their model, changes to the model’s variables after that point tend not to impact the final outcome in November.

For now, the Yahoo model is assuming that Obama’s approval ratings won’t change much between now and June. Of course, in the current age of real-time social media, where the news cycle moves at the speed of light, and public opinion can fluctuate correspondingly, who knows if that’s still a safe bet.

[Images: The White House]

E.B. Boyd is’s Silicon Valley reporter. Twitter | Google+ | Email


About the author

E.B. Boyd (@ebboyd) has holed up in conference rooms with pioneers in Silicon Valley and hunkered down in bunkers with soldiers in Afghanistan