I’m going to try something a little different today and present an argument in three parts. The goal of this first part is really to lay the groundwork for the other two. All I want to do here is convince you of something that you probably already know. 2020 polling was really, really bad. And from there we’ll move on to examining what that means for the media that relied on these polls and what it meant for the race itself.
There are lots of ways we could do this but I’m going to keep it simple and just present a list of examples, starting with one I wrote about Wednesday:
Maine – Senate
The most striking thing about the race is that Susan Collins never led in a single poll prior to winning the election. Here’s the full list from RCP. I realize it may be a bit hard to read but you don’t really need to see all the details, just look at the column of uninterrupted blue numbers on the right, they range from +1 to +12 for the Democrat. The final poll, completed 3 days before the election, showed Gideon at +6.
Susan Collins won the race by 7.5 points.
Florida – President
There were a lot of polls right up to the Sunday before the election in Florida. The final average of those polls showed Biden with a +0.9 lead, but if you expand to include polls that closed up to 9 days before the election, you once again see polls mostly favoring Biden. There are 13 polls favoring Biden compared to 6 favoring Trump (two of those from the same pollster, Trafalgar).
Trump wound up winning Florida by 3.4 points. Only 1 poll in that list came within a point of the final result.
Wisconsin – President
Here again we have an uninterrupted string of polls showing a solid Biden victory ranging in the final week from +1 to +11. The final RCP average was Biden +6.7. And it’s worth nothing that if you look at all the Wisconsin polling you’d have to go back all the way to mid-August to find a poll showing Trump leading (Trafalgar, +1).
Biden did win Wisconsin but just barely. RCP has the final result as +0.7 but the NY Times shows +0.6 with Biden winning by just over 20,000 votes. Only one poll on that list above (Trafalgar) was within a point of the actual outcome.
Ohio – President
Moving on to another must win state for President Trump, most of the polling in the final month showed Biden with a slight lead, though a few polls did show a Trump lead. RCP’s final average of polls in the final week was Trump +1, though it’s worth nothing that the only reason Trump wins the average is a Trafalgar poll that had him at +5. Take out that one poll and you would basically have a dead heat.
Trump won Ohio by 8.2 points. No poll, not even Trafalgar, came within 3 points of reality.
Texas – President
Are we having fun yet? You remember all the excitement among Democrats about their chances of turning Texas blue this year? That didn’t happen but as you can see, the polling in October probably gave them some hope it was close. The final RCP average was Trump +1.3, which is a squeaker. Notice there’s even one poll in there which was Biden +3.
Trump won Texas by 6.8 points or 654,000 votes. One poll on that list above (Rasmussen) was within a point.
Michigan – President
Below are all the polls that were completed in October. You’ll notice immediately there are only 3 polls that show a Trump lead. All are from the same pollster (Trafalgar). All of the other polls range from Biden +7 to Biden +12 with the exception of one Insider Advantage poll that had Biden at +2 in the final week. The RCP average was Biden +4.2, but again that was much lower that what most pollsters seemed to showing. If you averaged the entire month it would probably be around +9 for Biden.
Biden won Michigan by 2.7 points (the NY Times shows 2.6). The only poll that came within a point of that result in the entire month of August was the Insider Advantage poll. Everyone else was off by 4-7 points.
Michigan – Senate
The same was true for the Michigan Senate race. Here’s all the polling for October. The final RCP average for the race was Gary Peters over John James +5.4. It seemed clear, especially with Trump also trailing by 4-5 all month, that James didn’t have much of a shot.
Peters won by 1.7. This was a relatively close race. Some of the polls (CNN +12 in the final week) were off by more than 10 points. Only a single poll in that list (NY Times/Siena) came within a point.
Iowa – Senate
Here are all of the polls for the Iowa Senate race completed in October. As you can see, many pollsters showed Democrat Theresa Greenfield leading. by 4-6 points. The final polls were more favorable for Joni Ernst and RCP’s final average was Ernst +1.4.
Ernst won the election by 6.6. Only one poll in October was within a point of the actual result, the one by Insider Advantage.
Kentucky – Senate
Polling never showed Mitch McConnell was in any real danger, despite Democrats spending around $100 million to unseat him. Nevertheless, the polling that was done showed a race as close as 5 points. That poll was completed in August, but as Joe Biden might say ‘C’mon, man.’ The only October poll shoed McConnell at +9.
McConnell won by 20.4. Only one poll came within 10 points.
South Carolina – Senate
There wasn’t a lot of polling here so I’m going to show you polls going back to September. There’s no average but you can see Quinnipiac had a couple polls showing a tie. There were plenty of stories about this being the race of Lindsey Graham’s life. The final polls from October had Graham up 3 and 6.
Graham won the race by 10.3.
Georgia – Senate (special election)
The special election in Georgia between Democrat Raphael Warnock and Republican Kelly Loeffler had more than a dozen candidates and will be headed for a runoff. Polling showed Warnock winning by double digits. As you can see, most polls in October showed Warnock winning by 11-19. One poll (Emerson) showed a tie and one showed Warnock at +4. But the final RCP average was Warnock +15.7.
Warnock did win, though again this is going to a runoff because neither candidate his 50%. However, the actual gap was +6.6. So the average of polls concluding the week before the election was off by more than 9 points. Special credit to PPP and Monmouth that had Warnock at +19 in their final polls.
Admittedly I’m handpicking the worst here. There are polls that were close to what looks like the final results. For instance, the RCP average in North Carolina was Trump +0.2. Right now Trump is up 0.4 with thousands of ballots still to be counted. So some of the polling may turn out to be close.
There were also some polls that underestimated Biden. For instance, in Minnesota the final RCP average was Biden +4.3. The actual result was Biden +7.2. Though if you look at all of the polls for October (not just the final ones) the average would have been a point higher at 5.2. Still, at least some polling did underestimate Biden by a bit.
But it remains pretty remarkable that a lot of polling in key battleground states, Florida, Ohio, Michigan, Wisconsin (and Texas which plays big in Democrats’ fantasy life) were pretty far off and mostly in the same direction. Key races that would determine control of the Senate also tended to strongly overestimate Democrats’ chances.
Needless to say, this isn’t the first time we’ve seen this happen. Trump’s surprise win in 2016 was also put down to polling that largely and consistently underestimated his chances.
It’s a hell of a coincidence that polling in places Democrats most hoped to win this time out seemed to tell them what they wanted to hear. In part 2 of this series we’ll talk about the crisis of credibility that creates for the media.
Update: 2020 Polling Part 2: Incompetence Or Partisanship? Pollsters Need To Explain Their Failure…Again