ol tonyo: Fwiw this probably at least allows for the possibility that the olds (who were 55 in '16) now have cell phones and habits consistent with younger folks (i.e., they might not pick up unless they're politically motivated to do so) though this is prolly bordering on conspiracy now, i dunno

Blocked.

imo its been a bit disingenous of elex forecasters to say that lay ppl shouldnt dig into crosstabs when they regularly mine them to write "Biden in trouble w XYZ group" articles. Presently, toplines are built on crosstabs that are telling us some wild stuff that we arent seeing in actual elections.

Like the thing is, if black voters were 88-12 Biden, and they're 10% - 15% of any given poll, you have a wide leeway for screwing up your black sample, maybe even 70-30 Biden doesn't blow up the topline, but 65-35 Biden is probably too far!

Right. Nitpicking some crosstab weirdness around the margins isnt helpful, but what we're experiencing rn in polling isnt "the numbers are a little odd but shake out in the end," they are pitching us a radically different electorate than what is actually showing up to vote. Seems like a big issue!

I've said it a million times - you either believe the polls and there's a generational realignment in nonwhite voters, or you believe there's a sampling fuckup. One of these has never happened, and one happens with some regularity, so you know, choose your own adventure.

here’s the thing: it is easy to believe a sampling fuckup with hard to reach voters (PoC, youth) it is much harder to believe a sampling fuckup with Olds

It isn’t if it’s the same olds who never picked up the phone to tell us they supported Trump in 16 or 20!

but they’re probably not!

Fwiw this probably at least allows for the possibility that the olds (who were 55 in '16) now have cell phones and habits consistent with younger folks (i.e., they might not pick up unless they're politically motivated to do so) though this is prolly bordering on conspiracy now, i dunno

I mean the options are: 1) there is a seismic shift with old voters 2) the polling is fucked, pretty consistently across different pollsters, with old voters

if you think there’s an enthusiasm gap with old voters *in a blue direction* that would explain what we’re seeing but it sounds pretty implausible to me just on the face of it

Right (and also with younger voters and POC). Which is ... it's just a little tough to fully buy into. And 2 is consistent with now about a decade of falling response rates etc.

not to insert myself here but we’ve had scores of 100% reliable real “polls” the past 4 years in the form of elections and none of those results show any indications whatsoever of dovetailing with the weird findings described here

Yeah, the argument from the GOP that these special elections don't mean anything because "low turnout benefits Democrats now" doesn't seem based on much other than the fact that it's necessary to explain the gulf between this and the polls.

Special election results and overperformance, by Nathaniel Rakich of 538docs.google.com 2024 Date,State,Office,D %,R %,Special Margin,Weighted Partisan Lean,Swing,Biden %,Trump %,Clinton %,Trump %,Notes 1/3/2023,Georgia,HD-07,N/A,100.0%,N/A,No Democrat ran; jungle primary, so GOP result...

Yeah I understand how overperformance in low turnout elex could be consistent with a changing electorate. But then what are we saying, that souls to the polls is going for republicans now? that college campuses are going to turnout for trump? i dunno man. I'll eat crow if i have to but seems weird!

It does seem weird! The best-case for Trump is that Black voters and young voters are unenthusiastic about Biden and that Republicans win on turnout. And while that's completely doable, it's just not what the polls are saying. So even if he wins that way, the polling error stands, IMO.

I can buy the idea that now that Dems have more highly educated voters in their collation they’ve now got the more certain voters, but what bothers me is that back when Republicans were the ones certain to vote every election, that was understood as a strength.

but the people who produce the wacky polls told me those aren’t representative of real life.

One of the few things I find useful about Cohn's stuff is that he gets into details of the actual process of conducting a poll & its pretty clear that its just increasingly hard to actually get ppl to take these surveys. They go thru something like 50k contacts in order to get an N of 500. Its wild.

It's totally nuts and I think he's been broadly really good at descdribing the tensions in the NYT polls and how they might not be giving us a full picture etc.

which is all the more funny to me that he articulates this stuff and then is like "but obvi all my polling is correct and should be received as such" lol

All the more reason to think the samples are just getting less representative, unfortunately. I think weighting can fix a lot but not everything.

yeah, I saw the unweighted data for a poll recently and they had like 20 respondents under 35 but something like 250 over 65, just this insane inbalance between age cohorts. Weighting can only do so much before you are just warping the picture.

This is the reason why pollsters are turning to AI.

All the right-leaning olds refused to wear masks

Post