*This story was originally posted on Oct. 11, 2016.
For the last three election cycles, PollyVote has successfully projected the winner of the U.S. presidential race.
University of West Florida political science professor Al Cuzan has been involved with the political forecasting project since it was created in 2004. It was a team project that included founder Scott Armstrong from the University of Pennsylvania and Randy Jones from the University of Central Oklahoma.
“The PollyVote really is sort of, you might say, a metamethod,” said Dr. Cuzan. “What we do is we just combine methods sort of like a ravaging parrot.”
So, ‘Polly the Parrot’ was a good fit for many reasons, including the fact that it made a good mascot for their website pollyvote.com.
Cuzan says their real-time forecasting model utilizes a lot of data such as vote-intention polls, politico-economic models, expert judgement and citizen forecasts.
Well-known poll aggregators such as RealClearPolitics and the HuffPost Pollster update combined polls on an almost daily basis.
According to Cuzan, six different methods are swept in, that we sweep in. The they average all the forecasts within each method, and then average the averages and “that gives us the PollyVote,” Cuzan said..
Thus far, the forecasting results have proven to be very accurate.
“In ’04, in ’08, in ’12, we predicted the winner every single day in the last 100 days leading to the election. We never predicted the wrong winner,” said Cuzan.
Additionally, the PollyVote team went back in time and applied their forecasting methods in the three prior presidential races from 1992 to 2000. Each time, PollyVote predicted the election winner over the last 100 days leading into the election. The margin of victory between their prediction and the actual vote was just one percent.
In the 2016 presidential race between Democrat Hillary Clinton and Republican Donald Trump, PollyVote has been tracking that head-to-head matchup since January.
Since then, Clinton has maintained at least a narrow lead.
In January, the real-time interactive prediction showed Clinton and Trump in a virtual dead heat, 50.3 percent to 49.7 percent.
In April, Clinton’s lead grew to 53.9 percent compared to 46.1 percent for Trump. The prediction was much the same at the end of August. Clinton lost a little ground in early October, but still led with 52.6 to 47.4 percent.
In Florida, PollyVote predicts a narrow victory for Clinton...
“It’s been anywhere between 4 and 6 percent, so it’s been consistent,” Cuzan said. “PollyVote has not predicted a single day that Trump would win the election.”
The PollyVote website also includes an interactive map that shows a state-by-state percentage-based prediction of each candidate’s share of the two-party vote. percentage and by number of electoral votes.
As of our interview in early October, PollyVote projects 341 Electoral College votes for Clinton and 197 for Trump.
Looking specifically at Florida, Cuzan says it’s been going back and forth, right now showing Clinton narrowly ahead. “As so many keen observers have noted, Trump needs Florida; if he loses Florida, it’s just about over,” he said.
Cuzan called Florida a bellwether state for Trump, as it’s an early reporting east coast state. Other critical states for Trump are Ohio and North Carolina.
“Yes, early on if Trump loses North Carolina, the dark clouds are gathering. And, if he loses Florida and Ohio, that’s it. It’s like it’s a flood coming.”
The critical states for Hillary Clinton include Virginia and Pennsylvania.
But, if Clinton and Trump win the east coast states they each need, some western states could swing the results.
“Colorado is very important; and in the polls sometimes they show him winning, sometimes the other one,” said Cuzan. “Right now, I think it’s flipped to Clinton. Nevada is also important.”
You can keep track of the real-time forecasting at pollyvote.com and you can link to it on our website, wuwf.org.
