@jackmiller.bsky.social and i went to #AAPOR in may and wrote up some of our findings and others' interesting research for the @bluelabs.bsky.social blog! disengaged voters - how to reach them & why they matter: bluelabs.com/dynamic-elec... AI fraud in online panels - bluelabs.com/when-ai-come...

Em pleno 2026 e tem gente que ainda leva AtlasIntel a sério... Se você realmente acha que não tem nada de problemático com a metodologia deles, eu tenho um princípe nigeriano aqui para te apresentar... 😅

C'mon folks, what are we even doing here? Pollster self-proclaims themselves as the most accurate pollster in the world, according to analyses conducted by the pollster themselves using metrics defined by the very same pollster. Shameless self-promotion with a pinch of conflict of intererest 😅

Glad to see that the NYT (Nate Cohn) has finally decided to join the past vote weighting team! :) Their "synthetic past vote" approach is quite similar to what I've been advocating: using the voter file to reduce turnout overreporting. www.nytimes.com/2026/05/18/u...

É frustante ver gente tomando isso pelo valor de face, sem o menor senso crítico, e depois sai por aí repetindo que esse foi o instituto que mais "acertou" as eleições. Por conta disso, coloco nesse fio alguns dos problemas e limitações dessa comparação: 🧵+

The methodology of publicly published pre-election polls shouldn't be a black box!

Bolsonaristas surtados com a AtlasIntel no Brasil, enquanto nos EUA eles veem sofrendo hate dos Trumpistas há meses, e eu só assim...

Our @um-src.bsky.social super-star, no introductions needed, @bradytwest.bsky.social delivering his presentation on how to improve survey data integration between probability and non-probability samples at #AAPOR #AAPOR26 #AAPOR2026

Back to LA for #AAPOR! Many places to visit, lots of people to see and not enough meals for all the restaurants we need to go! Oh, and of course, brand new things to share and to learn in survey research!

If you think we had already talked enough about weighting on "recalled vote", think again! This time, we will be presenting some interesting results of the effects of this approach on measures of association under a missing not at random mechanism.

Celebrating 80 years of the world-renowned Survey Research Center of the University of Michigan! 🥳🎉 What a huge honor to be part of this history!

The contrast of me giving an interview and @joywilke.bsky.social giving an interview, despite both of us being under the same roof!

👀👀👀👀

This is an interesting illustration of potential bias due to non-ignorable selection: respondents from a political poll recruited by ads on Instagram reporting that their main source of information about politics is Instagram.

Partisan nonresponse bias in a nutshell

A couple of examples of what could happen to the betting odds on these types of platforms if there were no election polls:

Às vezes me perguntam quais considero os melhores institutos de pesquisa no Brasil. Como metodologista e estatístico, prefiro me concentrar nas melhores práticas metodológicas. Então fica aqui uma colinha para o que considero as melhores prática metodológicas em pesquisas de opinião no Brasil: 🧵+

Me at looking at the logos of Michigan, Oklahoma and Eastern during the gymnastics meet: "Margin Of Error" 😜

Love surveys? Work with the best!

Love surveys? We're hiring! We're looking for new members of the polling team at @bluelabs.bsky.social. Two roles are focused on execution of day-to-day polling operations (Polling Manager, Lead Polling Manager) while the Survey Scientist is an R&D role! Share & apply! bluelabs.com/careers/

Huge loss to the public opinion field!

One of my many pet peeves is people who denote survey sample sizes by N. In surveys statistics, because we are dealing with finite population inferences, we reserve the notation N for population size and use n for sample size.

The Shades of ISR Pantone ™️ palette 💙 💛

I always thought this rule that the total error is 2x reported MoE nonsensical, because bias in surveys doesn’t typically scale with sample size. In fact, we usually see the opposite: smaller, more carefully designed prob samples tend to present smaller total error than large nonprob samples.

The Survey Research Center at the Institute for Social Research of the University of Michigan turns 80 years old this year. Very proud to be part of this history! youtu.be/giSsP7o4PZg?...

Resultados da pesquisa Meio Ideia, primeira pesquisa pré-eleitoral nacional de 2026 registrada no TSE (BR-06731/2026). Pesquisa telefônica com 2 mil entrevistas realizadas entre 08/01 e 12/01. Perguntas espotânea e estimuladas com cenários de 1° turno com Tarcísio e Flávio): +

Merry Christmas! 🎅🎄

I remember that when I would urge media to talk more about winning probabilities derived by forecasters/aggregators in their polling coverage, people would tell me the general public would not understand it. I guess the media decided that this is not a problem anymore? www.cnbc.com/2025/12/04/c...

What would once be the dream of any subjectivity Bayesian -- the popularization of using wagering odds as probabilities, is now becoming a real dystopia... 😫

This is some sort of measurement error, just don't ask me which 😅 www.nbcnews.com/news/amp/rcn...

This is an incredible opportunity if you want to work in polling and surveys with some of the best in the field! 😊

For reference, it took Selzer 12 days after the election last year to produce and release a full post-mortem of her infamous poll. Given their background on methodological transparency, I'd be very surprise if we see anything remotely similar from this so-called "America's most accurate pollster".

Did you say "representative sample"?

A year ago today, AtlasIntel said they would deposit at the Roper Center the raw data from their final polls. It's still not there. Something tells me they will not be depositing that data file, neither the ones from this week's polls...

I was just saying this in the other site: Mamdani's share should probably have a heavier weight in the post-election polling analysis than Cuomo's share, or at least it should be put together with Sliwa's, given the likely late shift. And looks like somebody already did it.

Believe or not, the New Jersey poll wasn't the lowest point in US elections for the (former?) "American most accurate pollster", AtlasIntel Take a look at their final poll in the PNP Primaries in Puerto Rico last year: Poll: Pierluisi +14 Result: González +9 That's a 23-points difference!

l guess NJ is a small demographic subgroup for AtlasIntel standards?

Mood da noite

Also, can we all put to bed now the myth of "the most accurate pollster in the US"? I know I've been saying this since after the 2020 election, but I think we can finally agree on that, right?! RIGHT?!?

Wow, turns out weighting the hell out of your sample recruited through river sampling out of ads in social media doesn't work after all, who knew?! 🤭

I've been calling AtlasIntel's poor (to say the least) methodology out since after the 2020 elections, when it was not a popular thing to say. I'm just glad you all caught up and we are all on the same page now!

ANES Data Release! electionstudies.org/data-center/... The 3-wave ANES panel is now available. It merges data from 3 election studies (2016-2020-2024), the first time the ANES has collected interviews of the same respondents across 3 presidential elections.

They literally just published a NYC poll two days ago, but sure, why not?

LMAO That's not how margins of error work. They reflect random sampling error (variance), not systematic errors, which clearly seems to be the case here, when looking across polls. Also, saying with a straight face that what's shady is if the crosstab looks okay. C'mon, you are not even trying! 😅

🚨It's finally here!🚨 AAPOR's Taskforce on 2024 Pre-Election Polling report is out! Full report: /https://aapor.org/wp-content/uploads/2025/10/AAPOR-Task-Force-on-2024-Pre-Election-Polling_Report.pdf Executive summary: aapor.org/wp-content/u...

I don't know who else still needs to hear this, but raking is not post-stratification. Could you all just stop using "post-stratification" as a synonym for "raking", or more generally, "calibration"?

I'm a sampling statistician and a lot of people ask me about "representative samples". Since there is quite some confusion about it, I'm putting here an entire thread with everything you need to know about "representative samples": 🧵 [1/n]

Happy World Statistics Day! 🎉📊🤓 Statistics plays a fundamental role in helping us describe and better understand our world through high-quality data. #WorldStatisticsDay

new polls website just dropped fiftyplusone.news

Oh, I see some new weighting variables here. Cool! And yes, the methods report is the very first thing I look at a new poll 🤓