HUFFPOLLSTER: Why Did Eric Cantor's Polling Get It So Wrong?

06/11/2014 05:58 pm ET | Updated Jun 16, 2014

We take a closer look at why two partisan polls missed Eric Cantor's surprise defeat. Economic optimism is growing. And New Yorkers are divided on ferrets. This is HuffPollster for Wednesday, June 11, 2014.

WHAT HAPPENED IN VA-07? - No one saw it coming. You see that phrase often in the news coverage of House Majority Leader Eric Cantor's (R-Va.) upset loss to tea party candidate Dave Brat. One big source of the surprise was that two polls conducted in the final weeks before the election gave Cantor double-digit leads. When the votes were counted, however, Brat defeated Cantor by an 11 percentage point margin (56 to 45 percent).

Early polls showed Cantor far ahead - An internal Cantor campaign poll of 400 likely Republican primary voters conducted two weeks before the election (May 27 to 28) by John McLaughlin of McLaughlin & Associates gave Cantor a 34 percentage point lead (62 to 28 percent), with 11 percent undecided. A second survey conducted for The Daily Caller a week later (June 2) among 583 likely Republican primary voters by the GOP firm Vox Populi Polling showed a much closer race, but still had Cantor leading by 11 percentage points (52 to 40 percent) with 9 percent undecided. [WashPost, Daily Caller]

Don't trust 'internal' polls? - That's a lesson some take away. Harry Enten: "This race is another example of why you shouldn’t trust the internal polls put out by candidates...I’ve written about the bias of internal surveys released to the public. We should assume that internal polls are biased and misleading — unless we have a good reason to think otherwise. In this case, we knew what we were getting with Cantor’s polling firm, McLaughlin & Associates. As demonstrated by Daily Kos Elections, McLaughlin has gotten many races wrong in the past two years. In Virginia, it had Mitt Romney and U.S. Senate hopeful George Allen winning their 2012 races easily. Both Republicans lost by a significant margin." [538, McLaughlin's track record via Daily Kos Elections

...but evidence the campaign believed them too - Enten's advice to be wary of publicly released surveys sponsored by campaigns and political parties is warranted, given the way their selective release creates statistical bias favoring the sponsoring party. Yet polls with partisan sponsorship have shown little collective bias in the Republican primaries so far this year, and some accounts suggest that Cantor's campaign believed their big lead in the McLaughlin poll foretold victory. "Cantor's friends are FURIOUS," The Washington Post's Robert Costa reported via Twitter. "[They] said he was told by consultants that he was up 20-30 points, didn't need to worry..." Similarly, an unnamed Republican consultant from Virginia told Roll Call that the Cantor campaign misread an unexpectedly high turnout as good news on election day, since "the models that they had indicated a higher turnout was going to be better for them. They completely missed it." [HuffPollster, @costareports, Roll Call]

Was methodology an issue? - Both the McLaughlin and Vox Pop poll drew their samples from official lists of registered voters, randomly selecting voters with prior histories of voting in Republican primaries to be called. The McLaughlin poll used live interviewers while the Vox Pop poll used an automated, recorded voice methodology. Beyond that, McLaughlin & Associates has disclosed little about their methodology. The automated Vox Pop poll did not attempt to identify the selected voter in each household, instead interviewing the person that answered the phone and using a screen question to toss out any that said they were not registered to vote. As such, the Vox Pop poll may have inadvertently included many voters with little or no prior history of voting in Republican primaries (11 percent said they were "probably not voting" in the primary, 30 percent identified as independents, 5 percent as Democrats). [Daily Caller]

McLaughlin 'tries to explain': Shane Goldmacher: "In an email to National Journal, McLaughlin, whose firm has been paid nearly $75,000 by Cantor's campaign since 2013, offered several explanations: unexpectedly high turnout, last-minute Democratic meddling, and stinging late attacks on amnesty and immigration. 'Primary turnout was 45,000 2 years ago,' McLaughlin wrote. 'This time 65,000. This was an almost 50% increase in turnout.' Translation: McLaughlin's estimate of who was a 'likely Republican' voter was way, way off the mark. But Cantor's total number of votes still shrunk, even as the total number of primary voters went up dramatically in 2014. He secured 37,369 primary votes in 2012 and less than 29,000 this year, with 100 percent of precincts reporting." [National Journal]

Democratic meddling? Dan Balz has more on McLaughlin's theory: "In an interview, McLaughlin said the size of the turnout suggested that many Democrats and others who haven’t voted in previous primaries had shown up on Tuesday. 'We won’t know until they produce the list of who actually voted,' he said. 'Chances are a lot of them didn’t have a history of voting in Republican primaries.' He said in his polling, Cantor was winning Republicans by a margin of about seven to two. Among non-Republicans, Cantor was ahead 46-40 percent. 'Because we were surveying off lists of past primary voters, there were very few Democrats in that,' he said. Asked whether he and others in the campaign had anticipated such a surge in turnout, McLaughlin said, 'Polls don’t predict turnout. You’re trying to make assumptions of who usually votes… Here Eric got hit from the left and the right and it created a large turnout.'" [WashPost]

Or just Brat winning over independents? Vox Populi pollster Brent Seaborn, in an email to HuffPollster: "Based off of previous primaries in Virginia’s Seventh District, it appears there was an 'Operation Chaos' type scenario where Brat successfully mobilized thousands of traditionally non-primary voters. There were 18,000 more votes cast on Tuesday than in 2012. And anti-establishment, anti-incumbent independents saw an opportunity to make a statement.”

But no weak evidence for Democratic interference - Scott Clement: "[T]he data cast doubt on whether Democratic cross-over voting caused Cantor's loss...If Democrats showed up in large numbers to vote against Cantor, turnout should have spiked highest from 2012 in Democratic-leaning areas, with Cantor seeing an especially large drop-off in support. In fact, turnout rose slightly more in counties that voted more heavily for Mitt Romney in the 2012 presidential election. Likewise, Cantor saw the biggest drop-off in support in Republican strongholds...Turnout expert Michael McDonald of the United States Elections Project found similar results analyzing precinct-level data Tuesday night, reporting GOP primary turnout was lowest in the most Democratic-leaning areas of the state. [WashPost]

What about timing? - The McLaughlin poll was conducted two weeks prior to the election, while the Vox Pop poll was fielded one week out. Vox Pop's Seaborn: "We believe that David Brat was able to close gap pretty significantly over the last week of the campaign. Undecided voters likely broke heavily Brat's way plus the enthusiasm and momentum were on his side for turnout. Our poll of past Republican primary voters, a little more than a week before the election, showed Cantor facing the toughest race he has ever seen. We also had him losing with independents and Democrats to Brat. It is likely that those individuals could have turned out to vote on Tuesday for Brat. It will be interesting to see how many new Republican primary voters turned out in this election."

Did you vote in the VA-7 primary? We want to hear from you -- drop us a note at pollster@huffingtonpost.com.

A dispatch from @PollSonnets:
Internals showed the leader far ahead,
And though much closer one poll found the race
None predicted that lead to turn to shreds
And egg to dwell upon many a face.

Victim of the upset of the season:
Majority Leader Eric Cantor.
The polls were wrong, what might be the reason?
Perhaps the work of some dark enchanter.

Maybe immigration made some wary.
Could be, too, crossover votes took a hit.
Then there's risk in many a primary:
Failure to project the electorate.

But one thing's clear now to all our readers,
Republicans need a new House leader

OBAMA APPROVAL REMAINS IN MID-40S - Julianna Goldman: "President Barack Obama’s favorability ratings hit the lowest point of his presidency in a Bloomberg National Poll, with just 44 percent of Americans saying they have positive feelings about him. That drop in personal popularity has become a broader drag on the public’s perception of his performance in office. Obama, already given poor marks on the economy and health care, also gets low ratings on two recent political firestorms: the management of the Department of Veterans Affairs and the deal that freed the last U.S. prisoner of war in Afghanistan, according to the poll. His job-approval rating fell to 43 percent -- near the level at the end of 2013 after the botched rollout of the Affordable Care Act, his signature health-care law." [Bloomberg]

ECONOMIC OPTIMISM GROWS - Steve Liesman: "After seven long years of sharp recession and tepid recovery, the CNBC All-America Economic Survey found some key measures of U.S. public opinion finally regained their precrisis levels. Ninety-one percent of Americans now believe their home prices will either be stable or rise over the next year, the highest since March 2007. And for the first time in three years, Americans picked real estate ahead of gold as the best investment. Respondents also said they will spend about as much on a summer vacation as they did before the recession, following a sharp drop during the financial crisis." [CNBC]

HUFFPOLLSTER VIA EMAIL! - You can receive this daily update every weekday via email! Just click here, enter your email address, and click "sign up." That's all there is to it (and you can unsubscribe anytime).

WEDNESDAY'S 'OUTLIERS' - Links to the best of news at the intersection of polling, politics and political data:

-St. Leo University finds a close race in Florida. [St. Leo]

-Mark Mellman (D) reviews the factors that Senate forecasting models can miss. [Mellman Group]

-New Yorkers are fans of carriage-drawn horses, divided on ferrets. [Quinnipiac]

CORRECTION: An earlier version of this article misstated the release of the Vox Populi poll as occurring "two weeks out" from Election Day. The poll was fielded on June 2 and published by the Daily Caller on June 6.

Suggest a correction