One of the big takeaways from the 2020 election cycle was that the polling industry had no idea what it was doing and was gaslighting the American people with lies about a massive groundswell for Democrats at all levels. And few in the polling industry has the ears of more liberal prognosticators that Nate Silver, the founder of analytical firm FiveThirtyEight. In his Sunday appearance on ABC’s This Week, Silver placed the blame for grossly inaccurate polling on COVID, the Midwest, and Republicans.
Leading into the segment, chief anchor George Stephanopoulos played a soundbite of Silver from November 1 noting that there was a possibility “you could have a polling error of the magnitude of 2016.”
“[A]nd that polling error possibility he raised did pop up again this year. The margin in most battleground states is farther narrower than we saw in pre-election polls, underestimating Trump's support again,” Stephanopoulos conceded. This gave way to a prerecorded video of Silver delivering his “Do You Buy That?” segment.
Of course, Silver started by trying to deflect and argue that the grossly inaccurate polling shouldn’t be an issue discussed right now and that they got the larger picture correct:
With all that went on in this election, including President Trump's refusal to concede so far, I'm not sure the performance of the polling is the most important story, and polls did predict the right winner in all but two states in the presidential race.
Though, he would admit “the margins were pretty far off in a lot of places,” he still insisted “it wasn't a total disaster.” Despite how some pollsters were predicting a decisive flip of red Senate seats to blue, he boasted: “Polls did call every state but Florida and North Carolina correctly in the presidential race, and everywhere but North Carolina and Maine correctly in the Senate.”
But he would go on to admit: “Still, overall, the polls were mediocre at best with numbers off by three or four points in the presidential race, and by more like five points in races for Congress.” “The problems were often biggest in the Midwest,” he griped as he noted some of the largest errors came from Iowa and Wisconsin.
When it came to placing blame, he said COVID played a role but mostly because Democrats stayed home, thus they were easier to reach, while Republicans there were more willing to go out (Click “expand”):
One reason for these issues might be COVID. If people are changing their living patterns around the pandemic, that might affect how they respond to polls too.
Democrats have been more likely than Republican voters to embrace social distancing. So, if you’re home more often, you’re easier to reach by phone. In fact, research has shown poll response rates for Democratic voters shot up once the pandemic hit in March increasing from 12 percent to 16 or 17 percent. That's enough to potentially skew the numbers.
Silver went on to argue that Republican voters often worked blue-collar jobs that didn’t require “knowledge,” thus they weren’t working from home where they could be reached.
“And remember that only about 37 percent of jobs can be performed at home. A lot of those are white-collar, knowledge-sector jobs held by college-educated professionals, a group that mostly votes for Democrats these days,” he said.
As he was wrapping up, Silver continued to blame COVID for pollsters “underestimating Republicans” and vaguely tagged “other issues” for that discrepancy. In reality, we know that pollsters oversample Democrats, to begin with.
The transcript is below, click "expand" to read:
ABC’s This Week
November 22, 2020
9:37:43 a.m. EasternNATE SILVER (November 1): So, we have trump with a 10 percent shot, and Biden with a 90 percent shot. So, 10 percent things happen fairly often. At the same time, you could have a polling error of the magnitude of 2016, and instead of losing all these states by a point, and Biden would win Pennsylvania by a point or two, Michigan by two or three points, Arizona by a point. There are, like, lots of upside cases for Biden, and there are also cases where he wins in a squeaker.
GEORGE STEPHANOPOULOS: There you see Nate Silver just before Election Day, and that polling error possibility he raised did pop up again this year. The margin in most battleground states is farther narrower than we saw in pre-election polls, underestimating Trump's support again. Here's Nate’s take on what that means.
[Cuts to video]
SILVER: With all that went on in this election, including President Trump's refusal to concede so far, I'm not sure the performance of the polling is the most important story, and polls did predict the right winner in all but two states in the presidential race.
Still, the margins were pretty far off in a lot of places, and as the founder of FiveThirtyEight, I certainly do have some thoughts.
Again, it wasn't a total disaster. Polls did call every state but Florida and North Carolina correctly in the presidential race, and everywhere but North Carolina and Maine correctly in the Senate. Still, overall, the polls were mediocre at best with numbers off by three or four points in the presidential race, and by more like five points in races for Congress.
The problems were often biggest in the Midwest. That includes states like Iowa or especially Wisconsin where Joe Biden ended up winning by less than one percent, a far cry from polls that had him winning eight points in the final days of the campaign.
One reason for these issues might be COVID. If people are changing their living patterns around the pandemic, that might affect how they respond to polls too.
Democrats have been more likely than Republican voters to embrace social distancing. So, if you’re home more often, you’re easier to reach by phone. In fact, research has shown poll response rates for Democratic voters shot up once the pandemic hit in March increasing from 12 percent to 16 or 17 percent. That's enough to potentially skew the numbers.
And remember that only about 37 percent of jobs can be performed at home. A lot of those are white-collar, knowledge-sector jobs held by college-educated professionals, a group that mostly votes for Democrats these days.
So, I buy that COVID was a factor in polls underestimating Republicans. The only factor probably not. I think there were other issues too. But still knock on wood, there will not be another global pandemic in 2024, so that will be one thing that pollsters don't have to worry about.
[Cuts back to live]
STEPHANOPOULOS: Let's hope you're right about that.
