Cal State Fullerton professor: Math predicts Obama victory
By PAT BRENNAN
2012-10-30 14:33:33
President Obama will win re-election, with a solid victory in the popular vote as well as the electoral college, if a mathematical forecasting model developed by a Cal State Fullerton professor proves to be correct.
The model, which relies on a variety of state-by-state polls as well as historical polling data, predicts a popular vote total of 51.81 percent for Obama vs. 48.19 percent for Republican challenger Mitt Romney.
The model places Obama's electoral college total at 286, Romney's at 252, with 270 needed to win.
The forecast, produced last week, is unlikely to change unless "somebody does a big blunder" between now and Election Day, said Civil and Environmental Engineering Professor Chandra Putcha – or the massive storm on the East Coast affects voter turnout significantly.
Putcha, who used a similar model to successfully predict the outcome of the 2008 presidential election, said that such models are usually developed by people with a political science background.
"People in civil engineering typically do concrete and steel," he said. "When you're talking to a civil engineer, you're talking about building a bridge."
But he said the same principles of probability apply, whether you're building bridges, consulting for aerospace companies – as Putcha has done – or trying to forecast elections.
Besides, compared to political scientists' modeling, "an engineering professor is using much more mathematics than they are using," he said.
Unlike some other well-known forecasting models, such as the one used by Nate Silver of the New York Times in his FiveThirtyEight blog, Putcha said he did not factor in economic shifts – or anything besides polling data.
His model relies on state polls, plus historical voting patterns among states, to come up with probable outcomes for each state.
He then applies probability calculations, using a spread of polling results for each state or, in the language of his field, a "standard deviation."
The polls used in his model include YouGov, Rocky Mountain Poll, Rasmussen Reports and Public Policy Polling.
In 2008, he factored challenger and incumbent strength into his model, but this time around used only polling data.
He does not attempt to account for attitudes about the economy, gay marriage, war or other issues.
"All that, I feel, is getting into the polls," he said.
And he says his forecast is an assessment of probability based on data, not an attempt to predict the election with certainty.
"Only God can say who will win," he said.
The election, he said, is "very close. One of the closest I've seen."
Nationwide, such attempts at probabilistic forecasting are about evenly divided, said Jay DeSart, an associate political science professor at Utah Valley University.
"I think there are maybe a dozen forecast models out there for elections," he said. "They've been split. About half say Romney is going to win, the other half say Obama is going to win."
DeSart's model, done with Thomas Holbrook, a government professor at the University of Wisconsin-Milwaukee, has produced numbers similar to those of Putcha, whom he has met; both contributed chapters to a book on election forecasting in 2010.
The DeSart and Holbrook model gives Obama 51.39 percent of the popular vote, Romney 48.61, and gives Obama 281 electoral votes to Romney's 257.
Putcha said he will be more likely to publish a paper on his method if his forecast proves right, though he might do so even if it fails.
He declined to say whom he favors in the election.
"I don't want to get into that," he said. "For me, it's the mathematics. It doesn't matter whether I like the output, or don't like the output; that is what the numbers are saying."