"As of this morning, county boards of elections had received 1,193,857 mail-in ballot requests! That's nearly 9X the number of requests received at this time in 2016."
North Carolina State Board of Elections, in a tweet on Monday.
From here on out I think it is going to be difficult to avoid mentioning the new polling every day. With 28 days until the election (exactly four weeks from today), we’ll likely be seeing more polling each day. So, I expect that most, if not all, posts from here on out will include some discussion of the most recent polling.
First of all, let’s get something straight. Polls are never perfect. Probability doesn’t work that way. Polls, if done scientifically, can give us a range of likely outcomes 95% of the time. If a poll has a margin of error of 3 points and Candidate A has 51% while Candidate B has 49%, then we can presume that 19 times out of 20 we will find that Candidate A has between 48% and 54% and Candidate B has between 46% and 51%. A single poll, however, could be an outlier (the 20th out of 20). So that’s why we pay attention to multiple polls over time. It’s also why polling averages are better indicators of voter preference than single polls are.
Do not give any weight to an argument that polls do not matter because they do not get the results exactly correct. Trumpsters are pushing that out over social media right now. I don’t know why. Are they trying to whip up turnout? Do they think this will convince undecideds to vote for Trump? Do they just need to believe their own bullshit? Maybe it’s all three. In any case, it is a silly argument. Polls do not predict future results. They are snapshots of public opinion taken at the time the surveys are in the field. We can make some assumptions about what the future electoral result may be based on polling, but that is a different thing that claiming that polls should perfectly predict outcomes or else they are meaningless.
There has been a lot of discussion this year about the 2016 polling from all political and ideological corners. Most of it has been unhelpful. Not only is 2020 not 2016 for dozens of reasons, but the polling error in 2016 was not that unusual. (The degree of error was not that unusual; pollsters did err in not weighting for education, which has mostly been corrected.) It made a difference because the race was so close in so many states that had a good chunk of electoral votes. In fact, the 2012 polls were even further off. No one remembers that because the polling error worked in Romney’s favor. And he lost. Few people talk about 2018 as well, in which the polls were largely spot-on.
Okay, so polling is not exact and that imprecision can allow observers to make wrong calls in a close election. But, not in a blow-out. Will 2020 be a blow-out? Some Republicans seem to think so. The polling is heading in that direction. There are still problems with turnout models due to the multiple crises we are facing this year, so likely voter modeling is tricky and that could miss something when we finally see the full count. That said, Nate Cohn and the good folks at the New York Times’s The Upshot have come to the rescue. Want to know what will happen if today’s polls are as off as they were in 2016? How about the same question for 2012? Now, we know.
Consider this table from The Upshot:
If the polls have the same error rate as 2016, Biden will win the election with 319 electoral votes. Thanks to the interactive tool at ElectoralVoteMap.com, this is what the map would look like in this scenario:
If they have the same error rate as 2012, Biden will win the election with 375 electoral votes. The map for this scenario:
The current polling has Biden winning 375 electoral votes too, but the popular vote margin would be nine points where using 2012 error rates it would be 12 points. When someone tells you that the polls were wrong in 2016, you can counter with “if they are just as wrong this year, then Biden still wins.”
On Saturday, the polling average margin was Biden +7.4. Two and a half weeks ago it was Biden +6.6. This morning it is Biden +8.8. This is the highest it has been for Biden since August 29th, when it was +9.1. On the face of it, the polls look like there has been movement from undecideds and others to Biden. Trump has lost some support, but it is starting to look like Biden is getting all the people who refused to choose one of the two in prior polls. Thus, CNN/SSRS's 57-41 result yesterday.
Most polls until now have had Biden up something like 51-44. The ones with bigger margins have been more like 52-42. The race was not exactly close to begin with, but Biden’s margin is getting bigger. Meanwhile, the Trump campaign is dark in Ohio for the third straight week and is pulling ads from Iowa.
In state polling today, we have some interesting news from Florida and Pennsylvania. The Florida results - all from quality pollsters - are all over the place. It’s too bad this state has so many electoral votes (29) because it always seems like such a mess. This year, Florida is important for two related reasons. One, it is almost certain that Trump cannot win without Florida. Two, if Biden wins Florida the election is over. Trump has no plausible route to victory without Florida, while Biden has several. Considering how small the margin of victory typically seems to be in the Sunshine State and how incompetent the elections officials seem to be, there’s a good chance we won’t know the outcome on election night. However, if the state goes for Biden early it’s safe to conclude that Trump has lost the election. If the state goes for Trump early, then it just keeps him in the game.
Suffolk (A) finds the race in Florida even at 45-45. University of North Florida (A/B) has Biden up by up by six points (51-45). SurveyUSA (A) has Biden up by ten (53-43). All surveyed likely voters, but SurveyUSA and UNF used much bigger sample sizes with lower MOEs. In Pennsylvania, Monmouth (A+) has Biden up by between eight (53-45) and eleven (54-43) points among likely voters depending on turnout model. Change Research has it closer with Biden up by four (50-46).
Finally, there is one really interesting bit of news out of a new SurveyUSA national poll that finds Biden up by ten points (53-43). According to Politicalwire.com: “In interviews completed after Trump had been helicoptered to Walter Reed Medical Center, Biden appears to have consolidated support, leading by 16 points among likely voters interviewed most recently, 56% to 40%.” It is getting very late for Trump to change perceptions voters have of him right now. And it’s not just close to the election, over three million votes have already been cast in 26 states - two million of them in the ten most competitive states according to CNN. And voting has only just begun.