Chagrined Minnesota pollsters hope to do better next timeMultiple election prognostications, by nonpartisan and partisan polls in advance of Nov. 2's power-shifting election, weren't just off. They were way off when compared with the election results. And now the pollsters are under fire.
By: Dave Orrick, St. Paul Pioneer Press / MCT
Dayton ahead by 12 points.
Oberstar to win by 10 percentage points.
Republicans to pick up eight state Senate seats.
Wrong, wrong and wrong.
On Thursday, Minnesota Public Radio and the University of Minnesota's Humphrey Institute of Public Affairs announced their entire polling methodology will get a thorough review from the editor of the Gallup Poll.
The organizations had released a poll less than a week before the election showing the "Democratic Party's gubernatorial candidate Mark Dayton has opened a 12 point lead over the Republican Party candidate Tom Emmer, 41 percent to 29 percent."
Oops. The results were more like neck-and-neck, with a recount almost certain, and Dayton leading 43.63 percent to Emmer's 43.21 percent.
Republican leaders blasted the poll at the time, and haven't stopped since Election Day.
But the MPR/Humphrey Institute poll isn't the only one whose results were strikingly different from the election results. Of six independent polling operations that conducted eight polls in October, only one -- a SurveyUSA poll conducted for KSTP-TV -- had Emmer within 1 percentage point of Dayton.
Gov. Tim Pawlenty criticized the media two days after the election for being "consistently wrong" in publishing "bad data." Nationally, a bipartisan group of partisan campaign pollsters sent a letter to independent pollsters criticizing them, Politico reported.
To be clear, just about no one in Minnesota -- including Republicans -- foresaw the amplitude of the wave that rolled the GOP to power in both chambers of the state Legislature, with a pickup of 16 seats in the Senate and 25 in the House.
"Let's be fair: The Legislature hasn't had a Republican Senate in years and years and years," said Michael Brodkorb, deputy chair of the Republican Party of Minnesota who headed the GOP's Senate effort. He said the highest range of the GOP's own polls showed perhaps a pickup of 18 seats in the Senate, "so we knew it was possible, but that was also the high end. In some areas, we were pleasantly surprised."
Independent polling is not usually done in local legislative races, but Brodkorb said he suspected the trend of high dissatisfaction with Democratic-Farmer-Labor Party members was still being overlooked, especially by DFL polls. "We were really puzzled," he said, referring to several races where the GOP believed the race was tight but DFL incumbents appeared to act otherwise.
Generally speaking, early summer polls are used to gauge where a candidate stands -- and, assuming he or she appears to have a chance to win, to raise money among supporters.
Late polls can actually affect the election, some say.
In his criticism, Pawlenty stopped short of saying Emmer would be leading in votes were it not for the chorus of polls -- by Rasmussen Reports, the Star Tribune, St. Cloud State University, MPR/Humphrey, SurveyUSA, and Public Policy Polling -- showing Dayton with varying leads.
But it appears to be on the minds of some Republicans.
"There is a time to have that conversation, and it's after the recount," Brodkorb said. "But when these late polls come out, it's right when we're gearing up our get-out-the-vote efforts, and they have the effect of deflating enthusiasm in the activists who we rely on to knock on doors."
On the other hand, a bad-news poll can shock supporters out of complacency. Dayton supporters, using polls as their basis, said they knew the race was much closer than the media reported.
The morning after the election, Dayton said he knew it would be close. "I predicted between 1 percent and 0.1 percent," he told reporters.
The Alliance for a Better Minnesota, a group that ran ads attacking Emmer, had internal polls showing the race "very close," said executive director Denise Cardinal, who is now part of Dayton's recount team. She wouldn't say how close, but she said it was enough that her organization pushed the information out to their organizers in an attempt to express urgency in getting Democratic voters out.
"We shared our polls with labor leaders and donors, the kinds of people we need to work for us," she said. "You could make the argument it helped us. I don't necessarily disagree with (Republicans) that reporting of polls can affect the outcome."
Piling on the pollsters is something of a tradition in politics, dating back long before current polling controversies over whether a poll that doesn't use cell phones can be accurate.
And just because a poll several weeks before an election doesn't yield the same numbers as the votes on Election Day doesn't necessarily make the poll wrong.
"Some people call it a 'snapshot in time,' but I prefer to call it a fluid opinion," said Steve Frank, co-director of the St. Cloud State University Survey, which in late October said Dayton led by some 10 percentage points.
He stands by the survey, noting that the key question invariably begins with "If the election were held today..."
"I really do believe that our survey was right at that time," he said. Like many -- but not all -- surveys these days, St. Cloud's uses a combination of cell phone numbers and land lines and doesn't rely solely on automated calling systems that can't detect if, say, an 8-year-old answers the phone and responds to the questions.
Late shifts in attitudes of the electorate can be the Achilles' heel of a poll done even a few days before an election, pollsters say, and while a last-minute scandal can be the most obvious cause of such a shift, other more-subtle factors can make it happen.
"All the polls missed it in the same direction, and it calls into question whether there's something sort of generic about the whole situation there in Minnesota," said Scott Keeter, director of research at the Pew Research Center for the People and the Press. "Some of the pollsters there worked elsewhere and did just fine. I haven't studied it closely, but these are my impressions. It makes me think there was something idiosyncratic about the Minnesota situation."
Frank thinks he knows what it was: Independence Party candidate Tom Horner, a former Republican.
"I'm convinced people who might have voted for Horner moved to Emmer, assuming Horner didn't have a chance," he said. "And the Republicans were able to get their people out to vote."
That last bit about getting out the vote is what John Schadl of U.S. Rep. Jim Oberstar's campaign said he believes explains Oberstar's loss to Republican Chip Cravaack, described almost universally as an upset.
Oberstar's campaign polled a month before the election. Up 15 points, Schadl said. They polled again a week before. Up 9 points. "So we figured we would be about 10 points ahead on Election Night," he said.
Cravaack won by about 1.5 points.
Cravaack's people saw the race tightening, as did a SurveyUSA poll, which gave Oberstar the slightest of edges -- within the margin of error -- in the final week leading up to the election. So what happened?
It might be all about "likely voters," said Neil Newhouse, founding partner of Public Opinion Strategies. Cravaack's campaign, like dozens of congressional campaigns nationwide, hired the firm to do their own polling. A month before the election, it showed Cravaack trailing by only 3 percentage points.
"There was a huge enthusiasm gap across the country," Newhouse said, explaining that Republicans were much more likely to say they were going to vote than Democrats.
There also was a massive swing from 2008, when support for Barack Obama turned legions of previously unlikely voters into voters.
"If Democrats used the 2008 models to determine likely voters, then they might very well be looking at bad data, especially for a midterm election year," Newhouse said.
Schadl said that could have affected their numbers. For example, he said Oberstar's campaign predicted a turnout of roughly 100,000 voters in DFL-heavy St. Louis County. The actual turnout was closer to 85,000 -- a gap that easily could have closed the 4,400-vote margin by which Oberstar lost.
Rarely do parties or candidates release full details -- questions and methodologies -- of their own polls, which is why the media often give them little credibility.
Cravaack's campaign took the unusual step of making much of the Public Opinion Strategies Poll public and availing its researchers to the media.
But don't expect that to happen too often. The same day the MPR/Humphrey poll was released, Emmer's campaign released numbers saying the race was a "dead heat," but it released only numbers.
Officials with the Republican and DFL parties said that because their polls often reveal tests of talking points and geographic targeting of resources, they don't support the idea of releasing their polls to the media.
Larry Jacobs, director of the Humphrey Institute's Center for the Study of Politics and Government, said transparency will be key to ensuring his organization's polls remain credible.
"We are committed to a transparent review of our polling methodology because we value the importance of continuous improvement in our efforts," Jacobs said in announcing the independent review. "If a shortcoming is identified, we will fix it. If not, we will have third-party verification that our methods are sound."
The most predictive poll in Minnesota, SurveyUSA, turned out to be among those conducted closest to the election. St. Cloud State's Frank said high costs -- polls can run tens of thousands of dollars -- prohibit him and many other pollsters from conducting last-minute polls.
But he said the most important feature of all polls is sound methodology and transparency.
Meanwhile, those licking their wounds from the election are reflecting on how much emphasis to place on polls.
Schadl said he doubts polls will go away, but they must be taken in context.
"You can do a lot of analysis and polling and everything, but in the end, the ultimate poll is the tally on Election Night, and you're talking about actual voters, not likely voters."
Distributed by McClatchy-Tribune Information Services.