TBO.com: Tampa Bay Online, The Tampa Tribune and The Tampa Times - breaking news and weather.
Thursday, Nov 27, 2014
Local News

Skewed-polls debate over, most called election right

By
Published:   |   Updated: March 14, 2013 at 12:45 AM
TAMPA -

This year's argument over "skewed polls" in the presidential race appears to have been settled.

The polls were skewed — not in favor of President Barack Obama, as Mitt Romney backers argued, but slightly skewed in favor of Romney.

For the most part, however, national and state polls were close to the election outcome, as in most recent presidential years, according to polling experts.

One example: a late Florida poll by Quinnipiac University Polling Institute that called Florida for Obama by a 1-point margin, and predicted a 13-point victory by Democratic U.S. Sen. Bill Nelson over Republican challenger U.S. Rep. Connie Mack IV.

Derided at the time by the Mack campaign and others, those figures turned out to be almost exactly on the money, according to vote counts as of Friday.

Nationally, "the polls accurately forecast the winner, collectively and in most of the battleground states," said Mark Blumenthal, creator of the Pollster.com website that's now part of the Huffington Post.

"The argument that they were overstating Democratic (turnout) was just utterly false. To the extent that there was an error at all, they understated Obama's standing slightly."

Costas Panagopoulos, polling expert at Fordham University, identified 28 well-known polling organizations as the most accurate, based on the national vote count as of Wednesday.

They projected an average Democratic win of 1.07 percentage points, just a point or so under the actual margin of about 2.2 percent.

Most were news organizations and nonprofit institutes, some affiliated with universities.

They include:

Polling averages of some websites did a good job of forecasting the outcomes nationally and in Florida.

On Nov. 5, the Real Clear Politics website average of national political polls showed Obama up by 0.7 of a point, and the Florida average as a 1.5-point Romney lead.

Nate Silver's FiveThirtyEight.com polling blog, published by the New York Times, showed Obama ahead 50.8 percent to 48.3 percent nationwide and Romney leading by less than a point in Florida.

In fact, as of Friday, Obama led nationally by 50.1 percent to 48.4 percent, and in Florida by 49.9 percent to 49.2 percent.

But some polling flubs also showed up, including the first incorrect Florida presidential race prediction in eight elections by a well-respected Mason-Dixon Polling & Research.

Its final poll done Oct. 30 to Nov. 1 for the Tampa Bay Times and Miami Herald showed Romney leading Obama by 6 points and Nelson up by only 5 points over Mack.

During the campaign, conservatives said polls showing Obama ahead were "skewed" because polling samples included too many Democrats.

They said pollsters were using "turnout models," or projections of who would vote, based on the 2008 election, when enthusiasm for Obama drove the Democratic turnout.

Based on that argument, Mack's campaign revised the polling outcomes to reflect greater numbers of Republican voters and predicted a narrow Mack win.

Other Republicans did the same with national polls and predicted a Romney win.

But pollsters, including Quinnipiac, replied that they weren't using any partisan turnout model.

Rather than choosing a polling sample based on a Democrat-Republican breakdown, said Peter Brown of Quinnipiac, they seek a sample based on a breakdown by characteristics, including geography, age and race.

They then report what the respondents say when asked which party they identify with.

"Party identification is a changing measurement," depending on the mood of the electorate, said Brown. "Voting behavior, and whether people see themselves as Democrats or Republicans, can change without any change in their registration."

In 2010, Blumenthal said, Democrats made the same argument Republicans made this year, citing polls that appeared to have too many Republicans.

Democrats suffered image problems that year, in part, because of the passage of the Obama administration's health care reform act, and sustained heavy election losses.

"It's possible that the same people who called themselves Republicans then didn't want to describe themselves as Republicans this year," Blumenthal said.

"There's a pretty clear pattern; it's not anything new this year. When partisan supporters of a candidate see a poll they don't like, they want to attack the poll."

Mason-Dixon, however, did use a partisan turnout model, based on voter registration and past elections, said Coker.

He blamed the wrong call on Obama's late surge from news reports about his handling of Hurricane Sandy.

The final Quinnipiac poll, however, was done earlier than Mason-Dixon's final poll, and accurately forecast the outcome.

Coker said he didn't foresee the large turnout by young and Hispanic voters, or the strength of the Obama campaign's voter turnout effort.

Asked whether he'll continue to use the same techniques, he said, "You learn from your mistakes."


wmarch@tampatrib.com (813) 259-7761

Subscribe to The Tampa Tribune

Comments