Get ready for a wild polling ride

Image result for wild polling

‘There’s every reason to believe that there is huge instability in the current electorate,’ says one prominent pollster.

The nine-week, post-Labor Day sprint to Election Day began Tuesday with a muddle of new poll results that confirm only one thing: Hillary Clinton no longer owns the commanding lead she held a month ago.

A CNN/ORC International poll generated the most headlines Tuesday. The survey found Donald Trump ahead by 2 points nationally among likely voters — the first live-interview poll in six weeks to show Trump in the lead.

Clinton fared better in other new polls. A Franklin Pierce University/Boston Herald poll, conducted over virtually the same time period, showed Clinton leading by 3 points. The most recent weekly wave of the NBC News/SurveyMonkey tracking poll, conducted over the internet, showed Clinton’s lead over Trump holding steady at 6 points. And state-level SurveyMonkey data compiled for The Washington Post gives Clinton the edge in the Electoral College.

While volatility at this stage of the campaign isn’t unexpected, the latest polling data point to an stunningly unstable election environment, unlike any in recent decades.

There’s the sheer number of voters who aren’t lining up with Clinton or Trump. In the Franklin Pierce/Herald poll, 15 percent of likely voters were undecided or chose third-party candidates. One-in-10 voters in the NBC/SurveyMonkey poll offered no response when asked to choose just between Clinton and Trump — a number that, if anything, has been rising closer to Election Day. There were fewer undecided voters in the CNN/ORC poll, but that’s because it pushes voters to choose one of the candidates more vigorously than other pollsters.

Pollsters in both parties stressed Tuesday that the public polling is likely to be a roller-coaster ride from now until November 8.

“There’s every reason to believe that there is huge instability in the current electorate,” said Neil Newhouse, a GOP pollster who led the survey operation for Mitt Romney’s campaign in 2012. “Four years ago, there were probably 6, 7, 8 percent of voters up for grabs. Now you have 17, 18 percent up for grabs.”

Underscoring that volatility: Even after the release of the CNN/ORC poll showing Trump in the lead, Clinton still has a small-but-resilient advantage in the national polling averages. She’s ahead by 3.3 points in the RealClearPolitics average, and 5.4 points in the HuffPost Pollster model.

Private campaign pollsters in both parties aren’t shy about critiquing those public surveys, insisting that they say bounce around too much. Their own methodology is different: Private pollsters make assumptions about the composition of the electorate that control for such things like party identification. Public pollsters, on the other hand, typically don’t control for party ID, which they describe as an attitude, not a demographic. Instead, they weight for variables like age, or race, or geographic distribution to ensure they have the right mix of voters.

The subject was a hot topic in 2012, when Republicans dismissed polls showing more Democrats than Republicans in the electorate.

On Tuesday, Democrats noted that self-identified Republicans (32 percent) outnumbered Democrats (28 percent) in the CNN/ORC poll — a reversal from the previous poll, which showed Democrats with a 4-point advantage in party ID. John Anzalone, a pollster with the Clinton campaign, tweeted critically that CNN should weight by party ID.

“What we’ve generally seen — not to pick on CNN or anything — I think their polling has shown a lot of volatility,” added Democratic pollster Nick Gourevitch. “This was a hot topic last election: Public polls, lots of volatility. Private polls, not as volatile.”

Indeed, the CNN/ORC poll has bounced around quite a bit — at least in previous surveys of registered voters. Clinton led by seven points going into the conventions, 49 percent to 42 percent. But Trump pulled ahead following the GOP convention, 48 percent to 45 percent.

A week later, after Democrats’ concluded their convention, the CNN/ORC poll swung all the way back, with Clinton posting a 9-point lead, 52 percent to 43 percent.

The new poll was CNN’s first to employ a likely-voter model. Screening out voters less likely to cast ballots typically boosts the Republican candidate, though early likely-voter polls have been less consistently pro-Trump this year.

This time, the likely-voter screen helped Trump, turning a three-point Clinton advantage with all registered voters (44 percent to 41 percent) into a two-point Trump edge with likely voters.

Newhouse, the former Romney pollster, questioned CNN’s likely-voter model, pointing to unusual trends among the 100 respondents who qualified as registered voters but were deemed unlikely to cast ballots.

“When you actually do the reverse calculation on that, 52 were Clinton voters, 9 were Trump voters, and about 25 were Johnson voters,” Newhouse said, explaining why the likely-voter results moved toward Trump.

Newhouse said, when examining the public polls, he might look more at surveys of all registered voters, recounting a lesson from four years ago, when Romney voters were more passionate than Obama voters: “An unenthusiastic vote counts as much as an enthusiastic one.”

There’s another question about some of the polls out Tuesday: when they were conducted. A number of pollsters noted how unusual it is to conduct polling over Labor Day weekend, the end-of-summer marker when many Americans vacation away from home.

“It’s generally industry-standard not to call over Labor Day or similar holidays,” said Gourevitch. “A lot of people are on vacation. A lot of people are tuning out.”

The issue isn’t whether it’s harder to reach voters over Labor Day. Rather, it’s the danger that the voters you can’t reach are different from those you can, thereby biasing the results.

There isn’t much in the way of evidence on holiday-weekend polling. Four years ago, CNN/ORC was the only pollster to go into the field over Labor Day weekend, showing Obama and Romney deadlocked at 48 percent. (Labor Day weekend also marked the time in between the Republican and Democratic conventions that year.)

Newhouse noted that the increase in calls to cell phones means the old polling maxim of not calling over holiday weekends doesn’t necessarily apply anymore.

“I just don’t think it’s that big a deal,” he said. “Some of the old rules kind of fall by the wayside. I didn’t do any polls over Labor Day weekend, but it certainly doesn’t bother me.”

But more than focusing on any one poll, pollsters said they will be monitoring the next wave of major national and state-level surveys — many of which will be going into the field now that Labor Day is passed.

“I’m not convinced that [the CNN result] is a Labor Day thing,” Gourevitch said. “It could be a polling volatility thing. I don’t think we’ll ever really know the answer to that question. But what we’re looking for is what everyone else shows, not what one pollster shows over a potentially problematic time period.”

 

 

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: