Democratic and Republican pollsters alike are determined to get their predictions for the 2016 elections right in the wake of a series of high-profile missed calls.
Partisan claims of skewed results have also escalated, with Trump and his unconventional claim asserting that polls aren't capturing the Republican presidential nominee's true support.
Trump claims his campaign will turn out millions of new or irregular voters in November, some of whom will be voting for the first time.
Some pollsters acknowledge the race presents some new challenges.
"We know some people who are traditionally seen as unlikely voters are going to vote," said Nick Gourevitch, a partner at Global Strategy Group, which polls for Democratic candidates. "You need to take those people into account, and if you just lop those people off, you're going to miss something."
Modern public opinion polling is as much art as science.
The science comes in measuring the attitudes of the American electorate, and key demographic groups, in a statistically valid way. The art comes in defining just what that electorate will look like, and how much of a percentage of the electorate key demographic groups will make up.
Variations in those models can change the outcome of a poll.
In 2012, Republican polling firms underestimated the number of minority voters who would show up to vote in November. That left a number of GOP pollsters anticipating a win for their presidential nominee, Mitt Romney. Democrats and most non-partisan pollsters predicted a different model, one in which minorities would make up a greater share of the electorate. Those pollsters were closer to the electorate's actual makeup, and President Obama won reelection.
"At some level, [voter models are] based on a hunch about the future and how it will resemble the past," said Ann Selzer, a nonpartisan pollster who conducts surveys for Bloomberg, the Des Moines Register and the Indianapolis Star. "The best predictor of future behavior is past behavior, until there is change."
Pollsters typically ask a battery of screening questions designed to test whether a respondent is likely to vote in November. Those who are likely to vote make it into the poll's sample; those who say they are unlikely to vote get a polite thank you and goodbye. The pollster's goal is to survey a sample that most accurately reflects what the general electorate will look like.
After Romney's defeat, a postmortem conducted by the Republican National Committee found many GOP pollsters were concerned that the rising use of cellphones and the decreasing willingness of Americans to participate in half-hour long surveys were harming their ability to predict accurate turnout models.
"The problem with our previous assumptions were that there were just too many of them," said Gene Ulm, a partner at Public Opinion Strategies, the largest Republican polling firm in the country. "We're casting a much wider net nowadays."
This year, both parties are preparing for the unknown, scrambling to find pockets of voters who rarely turn out but who plan to show up this time around. Those voters may be disaffected white voters fired up by Trump and his promise to blow up the political system — or they might be African-Americans, Hispanics and younger voters motivated to cast a ballot against Trump and the party he heads.
In states like North Carolina, where Republicans such as Gov. Pat McCrory and Sen. Richard Burr are running for reelection, the GOP is experimenting with new turnout models to plan for both the best- and worst-case scenarios. They are also testing whether the casual voter who favors Trump will stick around to vote for candidates like McCrory and Burr, or whether they will vote for Trump and leave the rest of their ballots blank.
"We're modeling in different ways so that we're not shocked by any of this," said one senior Republican involved in the 2016 campaign, who spoke on condition of anonymity to discuss internal party strategy.
On the Democratic side, Hillary Clinton's campaign is relying on Elan Kriegel, one of the party's top analytic experts, to monitor the shape of the electorate. The campaign takes pains to survey irregular voters, even those who say they are unlikely to vote, sources inside the campaign said.
But pollsters — left, right and nonpartisan — say it is unlikely that a wellspring of overlooked voters will show up in November, for several reasons.
First, interest in the 2016 presidential campaign is at a record high. A Pew Research Center survey conducted in July showed 80 percent of registered voters have thought about the election quite a lot, a level higher than Pew has ever recorded. That suggests the vast majority of voters, even those who rarely cast votes, are making it through screens designed to weed out unlikely voters.
Second, volumes of political science research suggests that voters who tell a pollster they are unlikely to vote are telling the truth. If they don't plan to show up on Election Day, they almost certainly won't.
In fact, both Democrats and Republicans are coming to concrete conclusions about the types of irregular or rare voters who might be attracted to the polls to vote for Trump. Trump is likely to outperform Romney among blue-collar whites in suburban and exurban areas, especially in Midwestern and Rust Belt states hit hard by the decline in manufacturing.
Pollsters are including those voters in their samples, strategists on both sides said.
"We're not having trouble interviewing blue-collar people," Republican Ulm said. "I don't think there's a ton of mystery here."
At the same time, polls show a number of well-educated, economically upscale whites who typically vote Republican have yet to coalesce behind Trump — leaving some in the GOP worried that those voters, critical to down-ballot candidates like McCrory and Burr, might stay home in November.
In recent surveys conducted by polling firms like Monmouth University and Quinnipiac University, Clinton performs 8 to 10 points better among Democratic voters than Trump does among Republicans.
Still, pollsters in both parties concede a variety of factors — Trump's unpopularity, Clinton's own high negatives, a high number of voters who say they will cast a ballot for the Libertarian ticket — that are likely to change polls in the 10 weeks before Election Day.
"There's wrinkles, there's new dynamics, there's third parties, there's potential turnout issues," Gourevitch said. "All those things together create more uncertainty that there has been in recent years."
Reader Comments
to our Newsletter