Even if you’re the poll-watching type, there’s a pretty good chance you’ve never heard of the Democracy Institute.
The Anglo-American think-tank, which is associated with the libertarian Cato Institute, was one of the only pollsters to correctly predict two seismic political events. As Jim Rossi at Forbes pointed out, “[t]he Democracy Institute is a relative newcomer to political polling – but it correctly forecast Brexit and Trump’s historic 2016 upset.”
Yet, the group’s 2020 presidential polls aren’t included in either of the two major polling average indices, RealClearPolitics or FiveThirtyEight.
There’s no denying the Democracy Institute’s polling represents an outlier — but there are plenty of outliers within both indices, outliers which can’t lay claim to predicting either Brexit or the 2016 election. The major difference seems to be that its polling has consistently shown President Donald Trump ahead of Democratic nominee Joe Biden, both nationally and in major battleground states.
In Forbes’ words, it’s the “only” poll that “shows Trump leading.”
So how is the organization coming up with numbers so markedly different than those from almost any other pollster? In an interview with Forbes published Tuesday, Democracy Institute polling director Patrick Basham said “there is a misreading of the electorate, based on many assumptions that are not likely accurate.”
“Some polls may just be off, some are advancing an agenda, and there is an intertwining of the two,” Basham told Rossi.
“Polling is supposed to be incredibly scientific, and the science is more advanced than ever. But polling has always been a synthesis of science and art — and polls are more art than science in 2020. One of the major challenges is figuring out how the electorate will look.”
For the uninitiated, pollsters just don’t sample the first random thousand people that answer the phone and call it a day. Polls have to be weighted based on assumptions about the makeup of the electorate.
It’s not just how it’ll look demographically — how many minority voters, how many men versus women, how many voters with college degrees, those sorts of things — but also how motivated those voters will be to show up on Election Day. (Or to mail in their ballot, as the case may be this year.)
Basham thinks the rest of the polling world has that last part spectacularly wrong.
“With most of the mainstream polls — New York Times, CNN, the university polls – a fundamental assumption is that the electorate — aka turnout — will be much, much larger than 2016,” he told Forbes, saying those outlets estimate “10-30 million more [voters], that’s 25-30 percent larger. That makes it essential that polls capture many, many more Democrat voters.”
Instead, Basham said the Democracy Institute’s polls weight for lower turnout and try to “estimat[e] shy Trump voters or secret Trump voters.”
The outfit’s most recent poll, released earlier this month, showed Trump still leading Biden nationally, 46 percent to 45 percent. This was within the poll’s margin of error of +/- 2.5 percentage points; it was conducted among 1,500 likely voters between Sept. 30 and Oct. 2.
1) BREAKING: DEMOCRACY INSTITUTE POLL SHOWS TRUMP LEADING NATIONALLY, IN KEY BATTLEGROUND STATES, AND ELECTORAL COLLEGE
NATIONAL (Trump +1)
6 BATTLEGROUNDS of FL, IA, MI, MN, PA, WI (Trump +4)
TRUMP’S APPROVAL: 50%https://t.co/T9JxFp2mXI
— PollWatch (@PollWatch2020) October 3, 2020
The survey actually had Trump trending downward from August poll, where he held a 48 percent to 45 percent lead nationally over Biden; that poll was taken in the final two days and immediate aftermath of the Republican National Convention.
However, the group’s latest poll showed the president ahead in several major battleground states, including Florida, Iowa, Michigan, Minnesota, Pennsylvania and Wisconsin.
2) BATTLEGROUND STATES:
— PollWatch (@PollWatch2020) October 3, 2020
This, again, is a major outlier. The RealClearPolitics polling average, as of Saturday, has Biden up nationally 51.2 percent to 42.3 percent.
Only two national polls taken this year out out of the many included on the RealClearPolitics index have shown Trump ahead of Biden: one by Rasmussen in September which showed him up by 1 percentage point and another by Emerson in February which showed him up by 4 percentage points. The latter poll, too, can basically be thrown out; it was taken in the immediate aftermath of the New Hampshire primary, a period when Biden’s campaign was all but declared dead after a stunningly pathetic fifth-place finish.
However, Basham told Forbes the homogenous nature of the polls had a bit to do with the media outlets that commissioned them offering a “[c]ertain reward for a certain outcome.”
He offered several examples of the phenomenon.
“ORC — Opinion Research Corporation — got fired by CNN after the 2016 election,” Basham said. “Their main transgression was that they were pretty accurate,” showing Trump ahead of Clinton in key swing states.
“The USC poll — University of Southern California — repeatedly asks the same respondents 2 or 3 questions daily … and it showed Trump competitive in 2016. In late September 2020, it was showing Biden’s lead evaporate. They claimed ‘technical problems’ and changed their methodology. They are underpolling GOP voters.”
Basham pointed out the Democracy Institute’s modeling included shy Brexit and Trump voters in 2016 and got fairly accurate results. They see an even greater number of shy Trump voters this time around — having found some of them through their polling methods.
“Now we come to 2020, and a lot of people are looking. What we found is they do exist and in greater numbers,” Basham said.
“Our questions go like this: ‘If you were a Trump voter, would you tell anybody? Would you tell family? A friend? A coworker? Would you put a sign on your lawn or your car?’”
The answer to many of those questions, of course, is no. That’s why the Democracy Institute is weighting things differently. Basham sees three types of shy Trump voters — blue-collar male Midwesterners, suburban white women and people of color.
On that last count, consider this: “The #1 predictor of voting Republican is owning a handgun,” Basham told Forbes. “Black female gun ownership has skyrocketed. Forty percent of new handgun owners are female; 60 percent are African-American.”
According to the Democracy Institute’s model, Trump will end up winning 320 electoral votes this November. That kind of talk would get Democracy Institute pollsters laughed out of most rooms.
Perhaps “laughed” isn’t the right word; Basham said that “when the October poll came out, Trump was in the hospital with COVID-19. Our poll was the only thing he tweeted about that day. We got hate mail and personal attacks on a scale that was hard to encounter.
“How we produced the numbers was not challenged, just that we produced numbers others don’t like. We’ve gone from having the most media-ignored poll to the most infamous poll.”
Whether or not they remain either infamous or ignored, of course, will depend on whether they’ve gotten it right again.
Truth and Accuracy
We are committed to truth and accuracy in all of our journalism. Read our editorial standards.