clock menu more-arrow no yes

Political scientists have found the weird reason polls bounce around wildly during conventions

If you buy something from a Vox link, Vox Media may earn a commission. See our ethics statement.

Hillary Clinton speaking at a bus tour in Columbus, Ohio, on Monday.
Hillary Clinton speaking at a bus tour in Columbus, Ohio, on Monday.
Justin Sullivan/Getty Images

Two weeks ago, Hillary Clinton was consistently beating Donald Trump by around 5 points in the polls. Last week, after the Republican National Convention, Trump took the lead, and the LA Times put him up by 7 points.

But after the Democratic National Convention concluded Thursday, the pendulum swung back. Clinton is now leading by 6 points, according to an LA Times poll published Monday. The swing in poll averages is less dramatic than in the Times’ polling, but still very noticeable:

Up and down they go — why are Trump and Clinton trading places in the polls? (Courtesy of RealClearPolitics)

What’s going on here? Does the Times’ 13-point polling swing suggest that 13 percent of eligible voters in America — about 26 million people — are wildly switching back and forth between the two candidates? Are millions of undecided voters just changing their minds on the fly?

Well, no. What’s really happening, according to several political scientists, is that the conventions are raising both candidates’ poll numbers by temporarily increasing their voters’ response rates to pollsters. By getting lukewarm supporters of the parties excited for the election, the conventions have made each side’s voters more eager to say who they’re voting for — and that, in practice, makes the polls swing wildly.

Why polling swings mostly aren’t about people really changing their minds

If you look at the polling from the 2012 general election, you could be forgiven for thinking that Mitt Romney would have won if Election Day had only been held in early October instead:

Romney’s polling bump came after the debate on October 3, 2012, which he was widely perceived to have won. According to some media accounts of the election, Barack Obama wrestled back his polling lead by beating Romney in the next two debates and convincing America to swing his way.

That’s probably not what really happened. In a study of the 2012 election called "The Mythical Swing Voter," Columbia University professor Andrew Gelman found that Romney’s bump was a result of his voters responding more to pollsters after the debate — not a result of swing voters switching back and forth. When you bore down into who the polls’ respondents were, it just wasn’t true that Romney had momentarily won independent fence-sitters to his side who then swung back to Obama, according to Gelman.

This tied in with Gelman’s broader point: that despite the coverage in the press, ups and downs in the polls in the home stretch of an election don’t really appear to reflect "independent" voters changing their minds.

Apparent swings in vote intention represent mostly changes in sample composition, not actual swings. These are phantom swings arising from sample selection bias in survey participation. Previous studies have tended to assume that campaign events cause changes in vote intentions, while ignoring the possibility that they may cause changes in survey participation.

We will show that in 2012, campaign events more strongly correlated with changes in survey participation than vote intentions.

This is certainly at least what’s partially going on now with post-convention polling, says Alan Abramowitz, a political scientist at Emory. "These fluctuations are kind of artificial — they’re difference of response rates to pollsters rather than real shifts in how people are thinking," Abramowitz says.

When do I start paying attention again?

The conventions make party voters suddenly proud to declare their support for the nominee. But that spike in pride dissipates, and we’re soon left with response rates not temporarily affected up by the excitement of the conventions.

So polls a little bit after the conventions, not in the midst of them, are where to look to see what'll happen in November.

Political scientists Robert Erikson and Christopher Wlezien, authors of the book The Timeline of Presidential Elections: How Campaigns Do (and Do Not) Matter, are widely regarded as the top authorities on general election polling.

And one thing their data from 1952 to 2008 clearly shows, according to Princeton election expert Sam Wang, is that head-to-head polls after the conventions are actually great predictors of who will win the race. In fact, the candidate ahead in polls two weeks after the conventions has gone on to win the popular vote in all 16 elections the political scientists studied, according to Vox’s Andrew Prokop.

"The polling averages just don’t move much more than 2 or 3 percentage points after the conventions," says Georgetown political scientist Jonathan Ladd says. "Especially since the conventions are in July, the person in the lead in August will have a really good chance of winning."

It’s fun to imagine the election as a horserace in which the players can fall behind and surge ahead only to catch up again. The underlying reality, however, is probably much more mundane.


The bad map we see every presidential election