1726 164 10 LINKEDIN 334 COMMENTMORE

Republicans out-performed their poll numbers in race after race Tuesday, raising questions about pro-Democratic bias in this year's election's polls — a major turnaround from the pro-GOP bias in 2012's polls.

The series of misses caused at least one political forecaster — Larry Sabato of the widely read Center for Politics' Crystal Ball — to call for changes to the industry Wednesday.

Sabato said on Fox News that he wanted an investigation of polls in Virginia that showed a double-digit lead for Democratic Sen. Mark Warner — who wound up winning re-election by less than a point.

Video Keywords Mark Prior senate majority leader Cory Gardner Kay Hagan Mark Udall Rick Scott Tom cotton Florida governor Mitch McConnell South Dakota Reported season Arkansas senator Kay Dakota incumbent senator incumbent Republican Pat Quinn Charlie Crist Oxford Montana North Carolina West Virginia Colorado White House Wisconsin Kentucky Venus Virginia Iowa Michigan Illinois Ohio Georgia Washington Obama

Republicans took over the Senate from Democrats and captured big wins across the country in a convincing and dramatic victory in the 2014 elections. VPC

Video Transcript
Automatically Generated Transcript (may not be 100% accurate)

"Boy, is that an industry that needs some housecleaning," Sabato said.

In all eight of this year's marquee Senate races, Republican candidates beat the polling averages compiled by Real Clear Politics, and in two of the races — North Carolina and Kansas — Republicans pulled off upset victories. Results were similar for governor races, including Republican wins in three states — Florida, Illinois and Maryland — where polls generally showed Democrats in the lead.

Nate Silver, whose FiveThirtyEight website pioneered the averaging of polls to help produce predictions in major races, reported that the average Senate poll in the past three weeks overestimated the Democratic performance by 4 points. That's a major turnaround from two years ago, when Senate polls in the final weeks of the campaign overestimated Republican performance by 3.4 points.

FiveThirtyEight missed on two Senate predictions. North Carolina's incumbent Democratic Sen. Kay Hagan was given a 69% chance of winning. She lost by 1.7 points. Greg Orman, an independent candidate for the Kansas Senate, was given a 53% chance of victory. He lost by 10.7 points.

Other major polling misses included Senate races in Arkansas, where Republican Tom Cotton won by 17 points after polls showed him with a lead of 5-7 points, and in Iowa, where Jodi Ernst defeated her Democratic opponent by 7 points after most polls showed her with a lead of less than 3 points.

The wave of Republican victories caught many Democrats — who were hoping the polls this year were once again undercounting Democrats — by surprise. Several poll analysts this year had speculated that the factors causing 2012 polls to underestimate Democrats' performances would re-occur — specifically citing the possibility that not enough Hispanics were included in Colorado polling and that whites were overrepresented in polls across the South.

Both problems of underestimating Republican performance in 2014 and overestimating it in 2012 can be traced to the difficulties pollsters have in trying to identify who will vote, according to Evans Witt, CEO of Princeton Survey Research.

Pollsters use a variety of techniques to build a "likely voter" model that excludes those seen as unlikely to cast a ballot. A major component of the models has traditionally been the past voting behavior of the respondents, along with interest in the upcoming election. But Ann Selzer, whose polling firm has won plaudits in recent elections for its accuracy, points to the fabled get-out-the-vote organization of Barack Obama in 2012 as changing the dynamics of who votes.

"The Obama campaign proved that if they can find you, they can get you to the polls," she said.

Witt cautions, however, that pollsters might have been "seduced by the story of the Obama campaign getting uninterested voters to vote." Pollsters took 2012 into account in creating their models of likely voters for this campaign, he said, because "you want to look at the last few elections, and say, 'Does my likely-voter model agree with that?' "

Pollsters who altered their models to give less credence to interest in the election or past voting behavior may have given Democrats too much credit for being able to turn out the vote in a midterm election, Witt said.

Exit polls offer support for the theory that Democrats faltered in getting their supporters to the polls. Young voters, who have overwhelmingly voted Democratic, made up 19% of the vote in 2012's exit poll, but were only 13% of Tuesday's total. Hispanics dropped from 10% to 8%, and women from 53% to 51%.

Autoplay
Show Thumbnails
Show Captions
1726 164 10 LINKEDIN 334 COMMENTMORE
Read or Share this story: http://usat.ly/1u1y3Yx