Elections

Why Pollsters Are Now Claiming They Got Everything Right

SORRY, NOT SORRY

Shocked by Trump’s crushing victory? Silly you for not realizing the expression “a coin toss” could mean 51/47 on the popular vote and 312 electoral college votes.

Polling figures
NurPhoto/Jaap Arriens/Getty Images

How did the pollsters get it so wrong?

If that’s what you are thinking this morning, well, the polling gurus have news for you. They didn’t screw it up at all. You just heard them wrong.

And they have the receipts to back it up: Nate Cohn in The New York Times, for example, summarized his most recent analysis by saying: “The polls tend to err as a group in one direction or the other, so a fairly decisive win by either candidate is still possible.”

ADVERTISEMENT

Then again, he did also say in the same piece: “If the poll averages are exactly right down to the decimal (they will not be), Ms. Harris would barely need to outperform the polls to prevail.”

Cynics will note the word “if” doing a lot of work there.

The other Nate, Nate Silver, eloquently made his defense by re-posting a piece written before the election citing 24 reasons Trump could win (although, less confidently, he pulled his own prediction model at 10:30 p.m.)

Sir John Curtice, a famed British political scientist, polling expert and senior research fellow at the National Centre for Social Research, chastised the Daily Beast when we asked him to put the Trump ‘landslide’ in context, responding, via email: “It’s not a landslide (4 point lead in the popular vote).”

In a follow up phone call, Curtice said: “It was an old fashioned election in which people voted against the incumbent government because they thought the incumbent government had messed up.

“Every poll aggregator said ‘It’s 50/50, it’s close,’ and guess what, it was.”

Curtice pointed out that Trump’s leads in the swing states were all within the margin of error: In Georgia and North Carolina, Trump was +3 against a forecast of +1, he was forecast to tie in Pennsylvania but will likely end up +3, in Arizona he was forecast to be +2 and looks set for +5. Harris was forecast to win Michigan and Wisconsin with margins of +1, in fact Trump had won them by +2 and + 1 respectively. The numbers may be pushing the envelope but they are clearly within it.

Curtice added that no pollsters had said anything other than that Trump could indeed win all seven swing states.

Asked to explain more generally if polling showed why Trump had won, Curtice said: “Exit polls said 45 percent of people felt worse off now than they did under Trump. All this stuff about the gender divide, Black men swinging towards Trump, none of it matters beside the central fact; income. The Democrats have lost ground with the less well off. It’s the economy, stupid.”

Constantine Boussalis, Head of Department of Political Science at Trinity University in Dublin, told the Daily Beast: “Essentially polling has a long history of massive failure over time, from Truman holding up the ‘Dewey Defeats Truman,’ newspaper [the inaccurate early call was based on a poll] onwards.

“Every pollster is trying to find, in any given poll, the closest match to the voting population and each has their own secret sauce to try and accomplish it, but ultimately, it’s a very difficult job.”

With Trump now forecast to collect 312 electoral college seats, having crushed Harris in all the competitive states from Pennsylvania to Nevada, and even coming close to threatening the true blue stronghold of New Jersey, Democrats are unlikely to be moved to compassion for the pollsters who they could convincingly argue gave the clear impression the race would be down to the wire, with talk of coin flips and counting continuing for days, if not weeks.

However Zachary Greene, an Associate Professor of Politics at Strathclyde University suggests that those who were shocked by the polls, may have only been hearing what they wanted to hear: “There is a selection bias in what we look at. In fact the results are pretty close to getting what the polls were showing in many cases.”

Robert Mattes, Professor of Government and Public Policy at the University of Strathclyde, and co-founder of the African polling tool Afro Barometer said: “It is very hard to poll for the electoral college. But almost all polls do underestimate Trump. That suggests there is a systematic error somewhere in the polling—which is probably that Trump voters are harder to reach and access—and that is important because the effect is disproportionate under winner takes all rules.”

Probably the most explicit and high-profile example of poll failure this cycle will be the Ann Selzer/Des Moines Register poll of Iowa, which found Harris leading 47/44 in Iowa. On the night, Trump pulverized Harris 56/42.

After the result came in, Selzer issued an almost-contrite statement saying she was “thinking about how we got where we are,” saying she would “be reviewing data from multiple sources with hopes of learning why that happened. And, I welcome what that process might teach me.”

For anyone looking for that rarest of things, an apology from a pollster, this mealy mouthed excuse for the hardest word might just be the most fulsome it’s gonna get.

The Daily Beast Podcast episodes are released every Thursday. Like and download on Spotify, Apple Podcasts, YouTube, or your favorite podcast app. And click here for email updates as each new episode drops.