As they investigate the forces behind the party’s stunning losses in November, Democrats are coming to a troubling conclusion. The party didn’t just lose among rural white voters on Election Day, it may have failed to capture them in its pre-election polling as well.
Many pollsters and strategists believe that rural white voters, particularly those without college degrees, eluded the party’s polling altogether — and their absence from poll results may have been both a cause and a symptom of Donald Trump’s upset victory over Hillary Clinton in several states.
Determining what exactly happened is one of the most pressing problems facing the out-of-power party. In order to win those voters back — or figure out a future path to victory without them — party strategists say they first need to measure the size of that rural and working class cohort.
John Hagner, a partner at Clarity Campaign Labs, a D.C.-based Democratic analytics firm, said 2016 taught the party a hard lesson about polling in the Trump Era.
“The folks who would talk to a stranger about politics just aren’t representative of people who wouldn’t,” he said.
The first evidence of the party’s polling blind spot surfaced in a governor’s race: the 2015 contest in Kentucky. Both public and private polls going into the election showed Democrat Jack Conway and Republican Matt Bevin running neck-and-neck — Conway had a 3-point lead in the final RealClearPolitics average — but Bevin won by a comfortable, 9-point margin.
Like some of the more Democratic states where Trump upset Clinton last year, Kentucky has a large rural and a large working-class-white population (often there is considerable overlap in the groups). Whites make up 88 percent of Kentucky’s population, and fewer than a quarter of Kentucky residents over age 25 have a college degree.
Demographic trends confirm that these voters have been moving toward Republicans, but they don’t provide an easy answer for why pollsters have struggled to capture them in surveys.
Hagner sees some similarities between Bevin and Trump — both businessmen who initially positioned themselves as insurgent candidates within the GOP. In both cases, there were signs of what’s known as ‘social-desirability bias:’ the idea that voters won’t admit for whom they intend to vote because they think others will look unfavorably on their choice.
“With both Bevin and Trump, every newspaper endorsed against them,” Hagner said. “The right answer, in air quotes, was, ‘I’m not going to vote for them.’ … There’s a small group of people who knew that, at some level, they didn’t want their support for Trump to be scrutinized.”
Pollsters are still analyzing whether a “shy Trump voter” effect may have been decisive in some states. Like the public polls, Democrats struggled to measure the presidential race in private polls in a number of Upper Midwest states with large numbers of working-class white voters.
Clinton’s campaign mostly ignored Michigan and Wisconsin — where public and private surveys showed Clinton consistently ahead — until the final days of the race and was edged narrowly on Election Day by Trump. And the campaign invested heavily in Iowa and Ohio — two traditional battlegrounds where she trailed – only to lose both by larger margins than expected.
“We projected Clinton to lose Ohio by 200,000 votes,” said Hagner, “and she lost by 450,000.”
Democrats’ polling problems might not only be voters hiding their intentions from pollsters — some voters may have been hiding altogether.
That bias against responding covers a number of different elements, including geography. One top Democratic strategist who requested anonymity to discuss candidly what went wrong with the 2016 polls pointed to difficulty in reaching voters in more rural districts because of spotty cell phone service.
The same strategist added that many of these voters also may choose not to participate in polls “because they don’t like the establishment and they don’t want to take a survey.”
The yawning education gap among white voters’ preferences — Trump clobbered Clinton among white voters without a college degree, while the two ran neck-and-neck among those with a degree — means that non-response bias may have been determinative, said Democratic pollster Nick Gourevitch, a partner at Global Strategy Group. And it may have been going on for some time.
“I think it’s very plausible that for years pollsters have been over-representing educated voters, and that it only came back to bite us recently because it was a key driver in vote preferences this time,” Gourevitch said.
It’s too early to say for sure that this explains Democrats’ struggles over the past two election cycles — or that these issues will still be relevant in 2017 and 2018. Most Democrats — along with Republicans and nonpartisan analysts — are waiting for more states to collect and publish data of which voters did and did not cast ballots, a process expected to conclude later this spring.
Democrats aren’t ready to prescribe remedies yet, but officials at the national party committees are sending strong signals that they plan to hold pollsters to a higher standard in the upcoming midterm elections. Rep. Ben Ray Luján (D-N.M.), who is chairing the Democratic Congressional Campaign Committee for the second consecutive election cycle, ruffled feathers last month when he suggested that “unreliable pollsters will not be invited back to the DCCC.”
A committee spokeswoman, Meredith Kelly, clarified last month that pollsters’ reliability isn’t just going to be determined by their 2016 results, but also by their willingness to participate in a DCCC-driven effort to test various polling methods.
“It’s more about unreliable data combined with an unwillingness to do better and to learn from that,” said Kelly, the DCCC’s communications director. “That’s when we’ll stop working with people.”
To that end, the DCCC plans to use this year’s races for other offices to test its pollsters – and different methods to reach the voters who caused problems in recent elections. That could include using its own automated survey infrastructure.
“We’re going to use the 2017 elections to basically ask multiple pollsters to test rural and exurban areas that have overlaps with some of our [target] districts,” Kelly said. “It’ll be an ongoing thing, so we’ll have a way to test whose approaches worked and were most predictive.”
Elisabeth Pearson, the executive director of the Democratic Governors Association, said her organization conducted a review after the 2015 Kentucky governor’s race and intends to use it as a model for how to proceed headed into the next two years, when gubernatorial elections will be held in 38 of 50 states.
“I’ve seen a ton of openness from pollsters. We’ve done a couple of these meetings where we’ve brought all these pollsters that we worked with and had a great conversation about best practices, deep dives into things like sampling,” said Pearson. “I think they all understand that it’s in their best interests.”
Powered by WPeMatico