
This is a “hot take” so don’t hold me to it. But I’ve had this idea ever since the results started coming in Tuesday evening and it became clear that the Blue Wave was not happening. Instead, we got the kind of election that most of us were assuming back in March. What happened? Why were the polls so bad?
I haven’t seen any analysis and it will doubtless turn out that there were a number of reasons that the polls to not do a good job predicting the election results. But I want to suggest that a big one may be feedback.
The idea is a common one in physics. The act of measuring an attribute of an object changes that object. For example: you can’t see what something looks like in the dark. You have to shine a light at it.
Under most circumstances, this doesn’t matter. It’s normally only at the quantum level that the effect is big enough to be concerned about. As a result, this is normally an issue that is only discussed along with quantum mechanics.
Polling Feedbacks
Could it be that the polling results affected voting?
During this election, a large percentage of the news coverage talked about the polls. Remembering 2016, everyone was clear that the polls could be wrong. But maybe they weren’t.
Maybe instead, the polls got people to change their behavior. In Arizona, most of the late-arriving mail-in ballots (Saturday through Election Day) favored Trump by a wide margin. This could just be votes from people who normally mail-in ballots. Elderly people often use mail-in voting.
But it could also indicate that people who might have sat out the race were motivated to vote because they didn’t like what they were seeing.
The other side is also possible: people on the left didn’t think they needed to vote because there was obviously a big lead for Biden. But this seems unlikely because (1) most liberals were suspicious of the polls and (2) the turn-out was high.

Just a Thought
I don’t know if this was a big effect or not. I do, however, feel certain that public polling and election models are bad for our society.
They feed “horse race” coverage. People watch them the way they watch a sporting event. And by “people” I mean “everyone I know including me.”
The media is really good at giving us what our basest instincts crave. This is the political equivalent of gorging ourselves on refined sugar. On the other hand, we are already well sorted in terms of policy preferences. And we know that the kind of policy detail we might hear on The Weeds will never come to pass. So it’s understandable that we would pour those political white crystals directly from a C&H bag into our mouths.
But we need to find a way forward without polls. Because for whatever reason, they aren’t very good at predicting modern American election results.
Blue Wave 2020 image by AnnaliseArt under Pixabay License. C&H Factory by Cullen328 under CC BY-SA 3.0.
In baseball writing, any writer as wrong as Nate Silver has been about predicting future results would be jeered off the website. You can get some predictions wrong (baseball has a lot of random outcomes), but if you claim to have a perfect predictive model and it’s not All That, readers won’t like it. That’s why I never make predictions. (I only cite Vegas odds.)
Unrelated, but worth reading as always, Thomas Frank: https://www.theguardian.com/commentisfree/2020/nov/07/trump-defeat-election
I’m more positive toward Silver than anyone I know. His election model is driven by the polls and the state polls seem to be shit. I think he is better at sports. I recommend checking out the 538 podcasts. They are very good. Of course, they are a bunch of number nerds like me. But they discuss uncertainty a lot. One thing they discussed in the lead-up to the election is that they only have a dozen or so cases to model presidential elections; they have many thousands of cases to model sports. Note: that was said before the election.
Great! I’ll check out that Thomas Frank article as soon as I’m done with more chores around here!
The Minneapolis Star-Tribune was basically numb on state polling for weeks before the election. The Saint Paul Pioneer Press is thrashed, it’s now run out of some dreary office complex, and they’ve converted the old newspaper building into high-end condos.
(Old realtor joke: what’s the difference between a condo and a cold? You can get rid of a cold.)
Not to pick on Silver, but the state data is totally crap, and he shouldn’t present it as more.
I just saw polling numbers for the two Georgia races. Both are basically tied. So I messaged Will, “Democrats losing badly in Georgia.” I’m just going to assume any state-level polls have about a 6-point Democratic lean.
This may be related to the “Bystander Effect”, where a person is less likely to help if they think others are already helping
https://en.wikipedia.org/wiki/Bystander_effect
Yeah, that would be the same thing. Although I’m now thinking that it’s just that at the state level, they can’t get a decent sample population. I’m very interested to see what we learn when all the data is available. If public polling went away, I wouldn’t be sad.