That's similar to the shy Trump voter hypothesis. However, we never found strong evidences to support it, either in 2016, 2020 or 2024.

Yup, same here!

What an exciting future! 😭

Yeah, I'm not at all surprised. They always pull the same kind of BS in every other country they have a bad performance. Either that, or they just never talk about it, like in the primary of the gubernatorial election in Puerto Rico or the presidential election in Mexico last year.

Based on their performance in São Paulo mayor election last year, they will come up with the lamest excuse ever and then never mention it again lol

How something like ~ 1.96*se(estimate) is incorporating bias from sampling and non-sampling error?

That's the issue with bias, in general, you don't know what its magnitude, so there really isn't a lot you can do about it!

Oh sure, the media there surely give them much more credit than they deserve. But that's just because Atlas is very good and aggressive with their marketing strategy. Methodologically, they are really bad, and in Brazil worse than in the US.
Eu também sou brasileiro ;)

Not sure what you are talking about. They managed to be one of the few pollsters to stay outside the margin of error in the 2nd round of the 2022 presidential election in Brazil and did terrible in the city elections last year, especially in the 1st round.

Looks like this is a poll with n=500 RVs. Assuming it was a gen pop poll, a back-of-the-envelope calculation is getting me that this estimate of 765,000 people (the "nearly a million New Yorkers") would have a margin of error of about 212,000 people. So, yeah, there is a lot of error in there!

She did. She doesn't fully acknowledge, but had she used a better weighting strategy, it wouldn't have been so bad: www.documentcloud.org/documents/25...

Also, even if that's how it worked, if your subgroups don't have enough sample size to support estimates with a reasonable level of precision, you don't have to publish them. That's pretty standard in the industry.