Advice regarding polling in the final week of a presidential race has long since calcified into one universal dictum – ignore the polls.
Well, thanks for nothin’.
If I wasn’t online half the day for work, I might have a 10% chance of ignoring the polls, but as it is, I can’t click three times without seeing national numbers for Donald Trump and Joe Biden, numbers from up to eight battleground states, numbers for undecided voters, independent voters, voters who voted for Barack Obama in 2012 but went for Trump in 2016, voters who disliked Trump in 2016 but disliked Hillary Clinton even more, and the always enthralling voters who aren’t really voters at all but delight in screwing up the data.
The election is in seven days, and it’s oddly comforting somehow that while just about everything you can think of is at stake, that would include polling itself.
“If Donald Trump surprises people, if Joe Biden’s got a five or six point lead and then Donald Trump wins, my profession is done; it’s finished,” Republican pollster Frank Lutz said on Fox News the other day. “Right now the biggest issue is a trust deficit, and pollsters did not do a good job in 2016. You can make a mistake once; you cannot make it again.”
Lutz dearly wishes he could take the same approach to polling as the president, who explained his approach to interpreting the data to a North Carolina rally over the weekend.
“That’s when I like polls, you know, when I get them in my favor, I like them. Right?” Trump said. “When they come in my favor, I repeat them all the time. I say they’re great. When they’re not in my favor? I don’t. I don’t discuss them.”
Right. Like a 6-year-old. We get it. Things you don’t like, you ignore, like COVID-19.
The question is, how are the adults who can’t avoid them supposed to interpret the polls that failed us in 2016? Damned if I know, but the climate does feel different this time around.
Not to put too fine a point on it, but the failure Lutz describes was narrow rather than broad. National polls on the eve of the 2016 election had Clinton leading by three percentage points. She won the national vote by two percentage points. The failure was in state polling, particularly in the Upper Midwest, where Clinton’s narrow leads were swamped by undercounted (some say silent) Trump voters, enabling Trump to enter the White House as a graduate of the Electoral College.
That would be one of the key differences this time. It’s doubtful there are any silent Trump supporters. They’re the loudest contingent on the electoral map, mostly because Trump is more interested in convening them to hear their cheers than he is with exposing them to coronavirus. Another big difference this time is that Biden’s lead, seven to 10 points nationally in data that was fresh Monday, hasn’t been fading in the way Hillary’s did. Clinton was up by seven with three weeks to go, but it was down to four eight days out. To this point in 2020, Trump has not closed the gap.
NBC’s national polling averages have Biden ahead 51-44, which duplicate the numbers in potentially make-or-break Pennsylvania. The same averages had Biden ahead 50-48 in Florida, ahead 51-43 in Wisconsin, ahead 51-42 in Michigan. The Guardian’s average had Biden up 52-42 nationally, 49-47 in Florida, 50-45 in Pennsylvania, 50-43 in Michigan, 49-47 in North Carolina, 49-45 in Arizona, and 51-44 in Wisconsin. Biden and Trump were tied in Iowa and Trump led 48-47 in Ohio. A New York Times/Siena poll late Monday put Trump ahead in Texas 47-43. If Trump can’t win in Texas, he can’t win.
Pollsters are so skittish after 2016 that even what is represented overwhelmingly in the preceding paragraph has to carry everything short of a Surgeon General’s Warning in 2020. I feel badly for them. When did the margin of error go from 3% to 93%?
“We must caution that the polls – particularly some swing state polls – severely undercounted Trump supporters in 2016,” said the editors of the Guardian. “We are not certain, despite assurances, that they have corrected this. Additionally, they may be over-counting Democratic support (more people may say they will vote for Biden than actually turn out.)”
You don’t need any evidence that polling is the best available example of an inexact science, but one reason for that isn’t so evident, and it’s perhaps best explained by Bruce Anderson, chairman of Abacus Data and Summa Communications, who was quoted in Maclean’s magazine:
“In 38 years polling public opinion, you don’t only learn what people think, you learn a lot about what they don’t know. Knowledge of the world seems to be deteriorating in America, abetted by a president – ignorant of the world himself – whose formula for political success depends on more people becoming less informed.”
Yeah, you wouldn’t believe, Anderson seems to be saying, who is out there in 21st century America just waiting to be polled.
People who think you could drop a nuclear bomb into a hurricane to break it apart. People who think the noise from windmills causes cancer. People who think you can take coal out of the ground and clean it. People who think it would be interesting to check out whether you can inject disinfectant into your body to cure COVID-19.
These wingnuts can screw up your polling.
Gene Collier: gcollier@post-gazette.com and Twitter @genecollier.








Comments
Welcome to Post-Gazette subscriber-only commenting community. Please review our community guidelines.