UI's Election Analytics website has Clinton winning comfortably -- for now

UI's Election Analytics website has Clinton winning comfortably -- for now

URBANA — If the presidential election were held today, Hillary Clinton would win almost twice as many electoral votes as Donald Trump, according to a nonpartisan University of Illinois website that analyzes state polling data.

Of course, things change — especially this year, with large numbers of undecided voters and plenty of instability in the polls, said computer science Professor Sheldon Jacobson, who developed the Election Analytics website with his students.

The website, electionanalytics.cs.illinois.edu, predicts the Electoral College split based on available polling data: As of Tuesday, Clinton would get 359 electoral votes and Donald Trump 179 in a two-person race, with about a 12-vote standard deviation. Candidates need 270 to win.

But the website is also interactive, allowing users to customize results by changing the scenario — excluding certain polls, including third-party candidates, or projecting which way undecided voters might swing. Simply choosing "very strong Republican" changes the Clinton-Trump split to 319 to 219; "very strong Democrat" makes it 387 to 171.

The website applies an intricate weighting system to state polling data to project Electoral College votes, an approach that Jacobson believes sets it apart from dozens of others across the country. It has been one of the most accurate reflections of both the presidential and U.S. Senate races for the past six years, he said.

With this methodology, the group should be able to provide better insights, especially in close races, said Jacobson, director of the UI's Simulation and Optimization Laboratory. And this is the year to test that, because "it's going to be close."

Clinton held a 4-point lead over Trump in Iowa in a poll taken just after the Democratic National Convention. On Thursday, the same poll showed Trump with a 1-point lead. Other polls show similar flip-flops between the two major candidates in "up-for-grabs states," with more instability this year than in any recent presidential election, Jacobson said.

"The question is: Will this instability smooth out over the next three months?" he said.

The large number of undecided voters who don't like Clinton or Trump will be key, with their numbers reaching 15 percent in some polls. The big question is what will happen when they get to the voting booth.

Based on data trends, the election is "going to get closer," Jacobson said.

Jacobson said the U.S. presidential races in 2000 and 2004 suggested that it's difficult to predict the outcome based on the popular vote. In 2000, Al Gore won the popular vote, but George Bush won the election by capturing the Electoral College (after a controversial Supreme Court fight over hanging chads and the like).

So the UI group created a prediction model based on the Electoral College vote.

It applies statistical estimators to state poll results to determine the probability that each presidential candidate will win each of the states (or that each political party will win the Senate race in each state). These state-by-state probabilities are then used in a programming algorithm to determine a probability distribution for the number of Electoral College votes each candidate will receive (or Senate seats that each party will secure).

The site weighs polling data based on different factors. More recent polls have more weight than earlier ones, for example, and a poll with 1,000 respondents would be weighted more than those with 500.

Jacobson and his students have used this method to analyze polling data and predict election outcomes since 2000, and developed a website in 2008. Their track record is pretty good.

In 2008, the Election Analytics team predicted that Barack Obama would take 359.52 electoral votes, and John McCain would get 178.48. The actual numbers were 365 and 173, respectively.

In 2012, the group expected 304 Electoral College votes for Obama and 234 for Mitt Romney. The result: 332 vs. 206.

The group also correctly predicted 10 of 11 battleground states in 2008 and 2012, and 35 of the 36 Senate races in 2014.

Although the site makes political forecasts, it is strictly nonpartisan. Jacobson shies away from any election analysis.

"We're the only true purist academic site," he said. "We don't give you political biases. And we've got really good analytics."

The mission of the project is to advance science and technology and give students an opportunity to apply what they learn in the classroom to a real-world event, the national election, he said.

The site is updated daily as new polls are published, usually about 6 p.m., he said.

The Senate forecasts aren't up yet but will be when sufficient polling data becomes available.

The site gets anywhere from hundreds to several thousand users a day, ticking up when it gets media attention, Jacobson said.

Of course, that doesn't come close to his popular March Madness basketball prediction website, bracketodds.cs.illinois.edu.

"We get more activity on that one in three days than we'll get here in an entire election season," he said.