Report: Polling Numbers Should Be Apples-to-Apples, NYTimes!

Polling analysis

Democrats are polling strong ahead of the 2022 Pie Fight, er, midterms. But… Nate Cohn at the NYTimes tells us that some areas where Dems are polling well are the exact same spots that polls got wrong in the 2016/2020 Goat Rodeos, and since then, polling methods haven’t changed much.

Our pals at Electoral-Vote are taking Nate Cohn to task on this, and it is a pretty glorious thing to see (emphasis mine):

And now that Cohn has rained on the Democrats’ parade, let us rain on his parade a little bit. He, of course, has to produce copy a certain number of times per week, and he’s gotta come up with provocative things to say. But comparing midterm numbers to presidential cycle numbers is more than a bit problematic, keeping in mind that 2018 and 2014 were both actually pretty good polling years while 2016 and 2020 were not. For example, if you look at our final 2018 U.S. Senate map, we had the Democrats finishing with 46 seats, the Republicans finishing with 51, and 3 seats too close to call. The final tally was 47-52, with one seat left vacant (that seat eventually went to the GOP, making it 47-53).

And there is, of course, a very plausible explanation for why 2016 and 2020 were out of whack, while the midterms were not. That explanation is: Trump voters. There is clearly some segment of the electorate that’s fanatical about him, and that makes sure to get to their polling places when he’s on the ballot, while also lying about their intentions to pollsters (or simply not taking the pollsters’ calls). We can’t prove that this dynamic is what separates 2016/2020 on one hand, and 2014/2018 on the other. And we can’t be certain that, even if that is the pattern, it will continue into 2022. However, again, it’s plausible.

Beyond that, it’s a little sloppy to suggest that because pollsters didn’t reinvent the wheel, they probably can’t do better. They don’t have to radically reimagine polling if they can just get their models of the electorate correct. And that’s where the changes appear to be this year. Emerson and several others appear to be using very tight likely voter models, such that nearly all the swing-state races (except Pennsylvania) appear to be very close. Much closer than we would expect, in many cases.

I’m still firmly of the belief that polling is a dead art/science since the rise of the Mobile Phone, and I would not be surprised in the slightest if we stick another nail in the coffin this cycle.

The new voter registration numbers (Google your state to see, it’s amazing) have surged since Republicans overplayed their hands; the Dobbs decision to force women to have their rapists babies are driving it. The Yutes of Today are registering in YUGE numbers, and women are outpacing men. Let’s just say that they are motivated.

Anyway, there is no conventional wisdom any longer, and we don’t know how this will play out, but we don’t think Nate Cohn knows either.

This entry was posted in 2022 Pie Fight, Get Out The Vote, The NYTimes, voting. Bookmark the permalink.

2 Responses to Report: Polling Numbers Should Be Apples-to-Apples, NYTimes!

  1. MDavis says:

    We have a land line. It’s how I learned that pollsters don’t have a Do Not Call list.
    Whatever. I don’t have the time to answer their stupid poll questions, so let them waste their resources on calling, only to get no response.

    Liked by 1 person

    • Martin Pollard says:

      I got rid of my land line over 20 years ago. The number belonged to my late grandmother and everyone in the family had it; I only kept it in case they wanted/needed to get in touch or in the event of a power outage (it came in handy during the northeast blackout of 2003). Nobody did (we were kinda the black sheep of the family anyway), and I finally concluded that since I had a smartphone, I was paying $50/month for virtually no benefit whatsoever, so bye-bye.

      Liked by 1 person

Comments are closed.