Donald Trump and his supporters are clinging to a new — and false — theory that polls are somehow being falsified to downplay Trump’s backing.
The idea behind this theory: Independent pollsters are oversampling Democrats in a nefarious plot — coordinated with Hillary Clinton’s campaign — to convince Trump supporters that voting is a lost cause and rig the election against him.
The origin of the claims is a hacked 2008 email to Clinton campaign chairman John Podesta that was released over the weekend by WikiLeaks. In it, a pollster asks Podesta for “oversamples for our polling.”
Trump himself fueled the conspiracy theory in Florida on Monday, saying: “WikiLeaks also shows how John Podesta rigged the polls by oversampling Democrats, a voter suppression technique, and that’s happening to me all the time.”
The problem: What Trump is alleging is flatly inaccurate and the argument is built on a misunderstanding of what “oversampling” means in polling.
“Pollsters just saw this and rolled our eyes,” said Dan Judy, a Republican pollster for North Star Opinion Research.
“This is the classic case of people using an intentional or unintentional misunderstanding of polling to pretend results they don’t like are invalid,” Judy said. “Most voters aren’t that sophisticated when it comes to ins and outs of sampling and statistics and polling. But there are a lot of people spreading this around who know better — or should know better.”
The polling conspiracy theory starts with the 2008 email to Podesta from Tom Matzzie, who was conducting private polling for an outside political group at the time.
His email doesn’t reveal much of anything. It mentions “media” — but that’s likely a reference to TV ads, or “paid media” in campaign lingo, not to news outfits’ polling.
He asks Podesta for “oversamples for our polling.” But he makes no reference, as Trump claims, to oversampling Democrats — and because his polls were private, for the group’s use in determining when and where to air TV ads, Matzzie had no incentive to falsify his results.
What is ‘oversampling’?
Here’s the reality about “oversampling.” Pollsters often dive deeper into certain subgroups (such as Latinos or African-Americans) to reduce their margins of error for those groups. Then they weight those groups to their actual proportion of the population.
Judy laid out an example.
If he were polling 600 likely voters in a state with a 13% Hispanic population, that would mean 78 of the voters surveyed were Hispanic. “The margin of error of that is extremely high — it’s over 10 points — and you can’t at all break that down. You can’t say, ‘What do Hispanic men or Hispanic women think?’ You couldn’t do that with any degree of mathematical certainty,” he said.
So, instead, Judy said he’d call 300 Hispanic voters — enough to look at “men and women, Republicans and Dems, age breakdowns, regional breakdowns, and in a state like Florida some ethnic breakdowns — Cubans, Puerto Ricans, South and Central Americans. And when you run your survey numbers, you weight that 300 back down to 78.”
The nonpartisan Pew Research Center also explains oversampling on its website: “A survey that includes an oversample weights the results so that members in the oversampled group are weighted to their actual proportion in the population; this allows for the overall survey results to represent both the national population and the oversampled subgroup.”
Matzzie, himself, mocked on Twitter the notion that his 2008 email could suggest 2016 polling was rigged.
“Option 1: 8 years ago we presciently rigged 2016 polling. Option 2: the people commenting on this are being misled by a demagogue,” he wrote.
Too many Democrats?
Another element of the poll critique Trump and his supporters are pushing is that national surveys are including too many Democrats and too few Republicans.
An ABC News poll released Sunday, for example, found Clinton ahead by 12 percentage points — its sample featured 36% Democrats, 27% Republicans and 31% independents.
A CNN/ORC poll out Monday that found Clinton leading Trump by 5 points among likely voters included 37% Democrats, 30% Republicans and 33% independents.
Even the Investors Business Daily/TIPP poll, which has been much more favorable to Trump than others throughout the 2016 cycle, is weighted to use 35% Democrats and 29% Republicans.
Trump blasted the ABC News poll Monday in Florida, calling it “a totally phony poll.”
“These are, they call them dark polls. Phony polls put out by phony media, and I’ll tell you what: All of us are affected by this. They’re trying to suppress the vote so people don’t come out and vote. But I’ll tell you what: We’re winning this race,” Trump said.
Trump also appeared to cite those breakdowns in a tweet Monday, writing: “Major story that the Dems are making up phony polls in order to suppress the the Trump . We are going to WIN!”
But criticizing these samples ignores two important realities about party identification.
One is that, unlike age, gender and ethnicity data, which the US Census collects to map out the makeup of each state, there are no solid national figures on party ID. Identification with a political party is more of an attitude, so pollsters typically ask respondents which party they identify with but don’t weight their results to match a specific party breakdown.
The other is that in presidential elections — unlike some midterms — there simply are more self-identified Democratic voters than Republicans.
“In a presidential electorate nationwide, there have been mid-single digits more Democrats than Republicans going back as many elections as there’s been polling,” said Judy, the Republican pollster.
The oddest part of all this: It’s happened before.
In 2012, Mitt Romney’s supporters argued polls were “skewed” against him. Dean Chambers launched the now-defunct site UnskewedPolls.com, based on the claim that the electorate’s demographics would look more like the 2010 Republican wave than that of the 2008 presidential race. It turned out to be wrong.
Dana Perino, a former George W. Bush White House press secretary, wrote on Medium: “By believing that the polls were wrong, I had let both myself and our viewers down. I had done them a disservice. … We shouldn’t have to learn this lesson again.”
Yet other influential conservative voices insist on pushing the conspiracy theory.
Conservative talk radio host Rush Limbaugh rehashed those 2012 missteps on his show Monday, noting that he thought polls were skewed against Romney, expected a clear Romney victory, and was wrong.
Still, Limbaugh concluded of the 2016 race: “There’s clear evidence here that the Democrats have seen to it that the polls in this presidential cycle are oversampling Democrats, and there’s clear evidence the pollsters have done that as well.”
It’s been a frustrating episode for Republicans who work in politics.
Ex-National Republican Senatorial Committee staffer Liam Donovan tweeted: “You can’t fix what’s wrong with the party so long as its most influential voices persist in lying to the base.”