Study finds PPP was the most accurate pollster in 2012

A study conducted overnight found liberal-leaning pollster Public Policy Polling was the most accurate predictor of the 2012 presidential election.

PPP projected a 2-point Obama victory and put him at the critical 50 percent mark, 50 to 48 percent over Romney.

As of early Wednesday, President Obama was leading the national popular vote by about 2.7 million votes, taking 50.1 percent support against Mitt Romney at 48.4 – a difference of 1.7 points.

ADVERTISEMENT
The analysis by Costas Panagopoulos, Associate Professor of Political Science at Fordham University, also found that none of the 28 major polling outlets he analyzed showed bias toward either candidate in the national polls.

“For all the derision directed toward pre-election polling, the final poll estimates were not far off from the actual nationwide vote shares for the two candidates,” he wrote.

Rounding out the top five were YouGov, Reuters-Ipsos and Purple Strategies. The Reuters-Ipsos and YouGov surveys were not included in the averages of some polling aggregators because they're Web-based polls.

The bottom five national surveys were conducted by Rasmussen, Gallup, National Public Radio, National Journal and Associated Press-GfK. All projected either one- or two-point victories for Romney nationally.

Gallup and conservative polling outlet Rasmussen were frequent touchstones of critics throughout the election cycle who claimed the polls were biased against Obama. The final surveys from both showed Romney ahead 49 to 48 percent.

Gallup's poll received a lot of attention because Romney’s lead among likely voters had been outsized in comparison to other polls. For nine out of the last 12 days before it suspended polling because of Hurricane Sandy, Romney led Obama by between 5 and 7 points in the poll.

And Rasmussen was the focal point in a controversy earlier in the cycle when the Romney campaign and other Republicans said polls showing Obama with a significant lead over their candidate were inaccurate. They argued many mainstream polls skewed in Obama’s favor because of sample sizes that based 2012 turnout projections on 2008, when Democrats — and Hispanics, blacks and young voters in particular — turned out in record numbers.

One website, www.unskewedpolls.com, began re-weighting the mainstream polls to closer track the demographic assumptions of Rasmussen’s conservative polling outlet. The website, which had been updated daily with the latest results, has not been updated since official election results began coming in.

While all the pollsters agreed the national race would be tight, the narrative throughout the cycle was that the election would be decided in a handful of battleground states.

Here, the results were more scattered, although looking at an average of polls was largely predictive.

There were 11 battleground states identified as toss-ups in the RealClearPolitics average of polls, and RCP’s polling averages accurately predicted the outcome in 10 of those states.

And Nate Silver, whose the New York Times’s FiveThirtyEight blog was a frequent target of criticism, got every state right.

“This is probably a good time to link to my book,” Silver tweeted early Wednesday morning.

Silver’s electoral vote prediction was slightly off, but that’s because he weights the electoral count to the likelihood of a state fulfilling his forecast. With states like Florida and Virginia too close to forecast with high confidence, Silver’s model hedged the overall electoral count to account for potential variances.

But still, he was close – forecasting 313 electoral votes for Obama and 225 for Romney. The president appears headed for a 332 to 206 victory — assuming his lead in Florida holds.

One of the reasons Silver’s electoral count fell short was his forecast only showed Obama with a 50.3 percent chance of carrying Florida. The Sunshine State, with its eclectic ethnic, social, and geographic divides, is notoriously difficult to poll, and is the only state RCP mis-forecast. The RCP average of polls showed Romney with a 1.5-point advantage heading into Election Day.

Rasmussen didn’t fare as well, predicting Romney victories in Colorado, Florida, Iowa and Virginia, and showing a tie in Wisconsin, which Obama won easily.

In Ohio, the most closely watch swing state, Obama won by 2, and heading into Election Day, Rasmussen was the only outlet that didn’t show Obama with a lead, forecasting a 49 to 49 tie. Rasmussen forecast the same in Wisconsin, which Obama won by nearly 7 points over Romney.

Rasmussen showed Romney with a 1-point lead in Iowa in its final survey, and Obama carried the state by nearly 6. In Colorado, the firm predicted a 3-point Romney victory, and Obama took the state by almost 5.

The American Research Group was also one of the few to show Romney ahead in both Iowa and Colorado.

Still, PPP was not perfect all around. It was the only outlet that didn’t show Romney with a lead in North Carolina – its final forecast predicted a 49 to 49 tie -- and Romney carried the state by 2.