there’s no evidence for “the polls are unreliable” in any historically special way either. even if you do the simon rosenberg thing and only count the longstanding reputable ones. you can theorize on why This Time Is Different and might even be right. but you’re just vibing, sorry
it’s a weird thing to see folks hyper-focused on any shift (or no shift) in polls and take them as gospel when the vibe was “the polls are unreliable” up until 5 days ago tbh
“what if they’re off 4 points?” great so you only lost wisconsin by 1 or you lost nevada (the actual tipping point i think) and it went to the house. but off by 4 is well into the norm
I mean you can quibble about magnitudes but all reports are that pollsters are trying to come up with ways to combat nonresponse bias and there is literally no credible way to do that that doesn't involve getting people to answer the phone
NYT's method literally just adds statistical bias, like... definitionally. If their weights weren't trimmed it would be pointless, and any effect that can be attributed *directly* to weight trimming can only be attributed to statistical bias
bsky.app/profile/john...
The more I think about the NYT/Siena decision to include partial respondents, the more silly it seems.
For those who aren't familiar: in the NYT/Siena poll, people who hang up partway through the call are now included in the final results.
the national polls were basically dead-on in 16 and what they missed was largely explainable by non-college in the midwest
in 2020 they mostly fixed the latter and just underestimated trump on the overall weights, so the national polls were off by about 3
"Underestimated Trump on the overall weights" is a pretty... nonsensical combination of words.
The conventional wisdom is that people showed up who don't answer the phone for pollsters. That's not a problem you can fix with weights.
Or at least, you can't fix it in a way that isn't equivalent to just giving Trump a 3 point boost. And that's what they're doing.
Maybe it's the right call because those same people still don't answer the phone and still plan to vote, but this is unknowable and not something you can validate
Which is why I think "polls are less reliable this year" is a perfectly accurate statement. That doesn't mean they're totally useless, it just means that their toplines rest on shakier assumptions than usual.
Right, and the least charitable way of describing what pollsters are doing is taking that exact idea and just giving him three more points (in a complicated way). This is at odds with how pollsters usually make adjustments between cycles. But it's their only choice, non-response bias is unfixable.