All of the Swings in This Year’s Polls May Have Been Meaningless

By
Photo: Justin Sullivan/Getty Images

Over the past two months, America has been transfixed by the violent mood swings of its least-decisive citizens, watching in slack-jawed wonder as millions of Ken Bones switched their allegiance with every passing news event.

At times, the fickleness of these voters seemed literally unbelievable: Why would anyone switch their vote from Clinton to Trump just because the former came down with pneumonia? And what kind of person decides they can’t support Trump after hearing him boast about sexual assault one week — and then changes his or her mind after reading the words “Clinton” and “emails” on a front page three weeks later? How could any of these stories suddenly make millions of voters prefer Clinton, the technocratic, politically correct, pro-choice, pro-immigration, tax-and-spend Washington insider, over Trump, the populist, overtly racist, pro-life, anti-immigration, tax-cut-loving outsider, or vice versa?

One political scientist offers a simple answer: They didn’t.

In a new column for Vox, Columbia University’s Andrew Gelman argues that the 2016 race is far more stable than it looks. Gelman contends that in our polarized era, very few voters change their opinions over the course of a general-election campaign. This conclusion is supported by his own research on surveys of the 2012 campaign, along with Alan Abramowitz’s analysis of this year’s polls, which demonstrated a near-perfect correlation between Clinton or Trump support in a given survey and the proportion of Democrats or Republicans in its sample.

In Gelman’s view, the wild swings in this year’s polls aren’t the product of changes in voter preferences, but merely changes in voters’ willingness to speak with pollsters.

The electorate hasn’t just become more polarized over the past three decades — it’s also become far less interested in participating in surveys. Where quality polls once boasted response rates above 50 percent, they’re now lucky to get an answer from 10 percent of the voters they call.

When answering a poll becomes this optional, the mood of each candidate’s supporters in a given moment can have a profound influence over its findings. Which is to say: Trump may not have lost a significant number of votes following the “grab ’em by the pussy” tape; rather, his drop in the polls may have simply reflected his unhappy supporters’ relative disinterest in talking to a stranger about the election. And the same would apply to Clinton’s numbers after James Comey’s letter: The news made Clinton supporters taciturn, while Trump backers were eager to tell pollsters that they want to make America great again.

By itself, this may not upset most pundits’ understanding of the 2016 horse race. It’s long been common wisdom that this election would be decided more by turnout than persuasion. If there were a strong correlation between participation in surveys and turnout at the ballot box, the poll swings would still be significant. Comey’s letter didn’t need to change anyone’s mind to have an influence — it could merely energize Trump’s base, while depressing Clinton’s.

But Gelman pours some cold water on that notion:

Voter turnout rates in the general election for a US president is about 60 percent. Survey response rates are below 10 percent. Survey response is a much more optional thing, hence it makes sense to see much bigger swings in differential survey responses than in differential turnout. So, yes, differential turnout in voting is a thing, it’s just not as big as differential nonresponse in surveys.

Gelman’s argument makes sense. It’s easier to imagine news events producing wild swings in differential nonresponse than wild swings in voter preference during a closely watched race between radically different candidates.

Still, it’s worth noting that his reasoning does not negate the possibility of a loose correlation between a voter’s interest in taking a pollster’s call and his or her interest in turning out on Election Day. It only suggests that voters’ interest in the former is likely to vary much more wildly than their interest in the latter over the course of a campaign.

What’s more, the Democratic Party’s internal data showed the revival of Clinton’s email saga did genuine damage to her candidacy. It’s possible that even the party’s internal polls, which are more rigorous than the average public survey, are nonetheless distorted by differential nonresponse. But a post-Comey-letter drop-off in enthusiasm was also detected in the party’s focus groups. Per the Washington Post’s Greg Sargent:

Democratic pollster Celinda Lake told me that the impact of the Comey letter was to sour millennials on Clinton and on the political process — potentially to the detriment of Dem Senate candidates in states like New Hampshire, Nevada, North Carolina, and Wisconsin. “Many millennials who were already discouraged about politics,” Lake said, saw increased “doubts whether voting for any politician makes any difference.”

By way of illustration, Lake recounted that in one focus group in Nevada, one millennial said she was down to a choice between two options. When the moderator asked her if that meant Trump and Clinton, Lake recounts, this young voter replied, “No, between Clinton and not voting.”

There’s also reason to think that Trump’s recent resurgence was related to actual shifts in voter preference: The GOP nominee’s share of the vote has steadily increased over the past month, as Libertarian candidate Gary Johnson’s has steadily decreased. That trend seems better explained by right-leaning voters deciding to come home for their party’s nominee than by differential nonresponse to public polling.

Finally, even in a polarized electorate, it’s possible to imagine how a voter could rationally vacillate between the two candidates. If you’re a nonwhite immigrant who believes that abortion is genocide — or a racist who believes in common-sense gun control — neither candidate represents your views perfectly, so it makes sense that your loyalty might waver with each new bit of information.

All that said, Gelman’s core contention — that this race has been more stable than polling changes have made it look — is convincing. And to the extent that a less volatile race is a less unpredictable one, that’s good news for the front-runner.