Top pollster: 'I have never seen anything like' what happened in Michigan
He's "never seen anything like" what happened there Tuesday night.
"No. No. No," Steve Mitchell, who conducts the Fox 2/Mitchell polls, told Business Insider in an interview Wednesday.
He was referring to the result in the state that stunned the collective political world Tuesday night.
Sen. Bernie Sanders, down in public polls by more than 20 points to Hillary Clinton in the state, stormed to an upset win over the former secretary of state.
"I've only been wrong like this twice in my life. And, that's a career that spans since 1986, 30 years," he said. "I've done this for a long time. I've polled every statewide election in Michigan going back 30 years ... I don't like to get it wrong, but I'm not alone."
He added: "I have never seen anything like this."
Mitchell's poll had Clinton winning by 27 points over Sen. Bernie Sanders of Vermont. Other polls were also significantly off the mark.
The past 15 polls listed for Michigan on RealClearPolitics each had Clinton winning the state by at least 10 points. The three most recent found her leading by an average of more than 20 points. Another recent poll from Mitchell had her winning by a whopping 37 points.
Sanders won Tuesday night's Michigan primary, 49.8% to 48.3%, over Clinton.
At the same time, many of the same polls that were so far off about the Democratic race were spot on for the GOP results. Mitchell's most recent poll for the GOP had party frontrunner Donald Trump beating Ohio Gov. John Kasich by a 41% to 23% spread. Trump ended up finishing with 36.5%, with Texas Sen. Ted Cruz barely edging out a second-place finish over Kasich.
"As I told my client, I'm in some pretty good company," Mitchell said before listing other major polling institutions that missed the mark with the Democratic primary. "Everybody else who polled it, nobody had it right, nobody had it within double digits. We're pleased with our Republican numbers."
He said his team is still looking into what exactly they missed. But he and other pollsters said it's likely a combination of small factors that led to the shocking final result.
First, he said his reliance on landlines, rather than cell phones, may have played a role in the distorted numbers. But added that it doesn't tell the full story.
"Because we thought 82% or 86% of voters were going to be over the age of 50, we thought we could use just landlines," he said. "But some of the polling [conducted by others] was also done with cell phones. There were a lot of companies that were doing both and came up with the same bad results. Something happened in terms of the voter turnout that we did not anticipate or build into our model."
Tom Jensen, the director of Public Policy Polling, told Business Insider in an email that the landline aspect did play a big part in the inaccuracy.
"There's a huge divide in the Dem race between people who have landlines and people who don't, and your polls are going to massively overstate Clinton's support if you don't contact the people without them," he said. PPP had found Clinton up 10 points in a mid-February poll.
Young voters turned out in much higher numbers than anticipated. Exit polls from NBC News showed that 37% of voters were between the age of 18 and 39. And 21% overall were between 18 and 29. Mitchell said he had only projected 8.7% of the total vote would come from those between 18 and 39. His survey also had Sanders winning that age group by a 2-to-1 split. Exit polls showed he won the age group by about 4-to-1.
Mitchell said the numbers regarding Clinton's African-American support were expected to be far stronger, as well. He projected a 40- to 50-point margin in the results, but she won instead by about a 2-to-1 gap, far less than in other states.
"We're going to figure out what went wrong and we're going to figure out how to fix it," he said.
Another reason floated for the polling discrepancy: Michigan's status as an open primary, in which registered Democrats and Republicans can vote in the other party's primary. Democrats could theoretically have voted for a Republican in hopes of upsetting Trump or ensuring his win, rather than vote for Clinton if they thought it was a done deal.
Reporters found some anecdotal evidence of this occurring:
Exit polls also provided this theory some weight. According to those exit surveys, 7% of Michigan Republican primary voters identified as Democrats. Just 3% of Democratic primary voters identified themselves as Republicans.
"You know, I think a lot of inside baseball people do think like that. How large the number is, I don't know," Mitchell said. "But obviously if it was only 20,000 to 30,000 that made the difference between Hillary winning and losing, it could've played a factor."
He added that he did not see much crossover in his polling.
Jensen agreed that the crossover factor may have played a role in the disparity between the polls and the results. He said that, because of the "overwhelming media narrative based on the polls that Michigan was a blowout for Clinton," Democrats found it worthwhile to vote Republican.
He cited the Ann Arbor precinct where he grew up as an example. That precinct provided Mitt Romney with 175 total votes in the 2012 general election. On Tuesday night, 250 people cast ballots in the GOP primary, with about two-thirds of the vote going to Kasich.
"That's obviously a very micro level observation but kind of remarkable to have 40% more people vote in a GOP primary than voted GOP in the last general," he said. "And I think it was much more likely to be complacent Hillary voters who did that kind of thing."
Upon finding out that the results would be far closer than anyone anticipated, Mitchell said he still believed Clinton would pull out a close win once all of the votes in places like Detroit and Grand Rapids were tabulated.
Once he knew that wouldn't be true, he thought to himself: "Everybody has gotten this thing wrong."
Brett LoGiurato contributed reporting.