Carl Allen
@realcarlallen.bsky.social
Analyst, Scientist, Misinformation Fighter, Debate Winner, Author
📖 Book: *The Polls Weren't Wrong* out now 👉 https://www.routledge.com/The-Polls-Werent-Wrong/Allen/p/book/9781032483023
📖 Book: *The Polls Weren't Wrong* out now 👉 https://www.routledge.com/The-Polls-Werent-Wrong/Allen/p/book/9781032483023
You know he's in a desperate spot because it's a full year until midterms, so when this doesn't happen, making more desperate promises closer to midterms will make him (somehow) look even weaker.
Remember when he said he would cut electricity prices in half by January?
How we looking?
Remember when he said he would cut electricity prices in half by January?
How we looking?
November 10, 2025 at 7:15 PM
You know he's in a desperate spot because it's a full year until midterms, so when this doesn't happen, making more desperate promises closer to midterms will make him (somehow) look even weaker.
Remember when he said he would cut electricity prices in half by January?
How we looking?
Remember when he said he would cut electricity prices in half by January?
How we looking?
Consensus of analysts around the world:
Polls were wrong about NYC Mayoral Election.
Is that right?
Or do you all detect another possible explanation?
Polls were wrong about NYC Mayoral Election.
Is that right?
Or do you all detect another possible explanation?
November 6, 2025 at 6:23 PM
Consensus of analysts around the world:
Polls were wrong about NYC Mayoral Election.
Is that right?
Or do you all detect another possible explanation?
Polls were wrong about NYC Mayoral Election.
Is that right?
Or do you all detect another possible explanation?
Nate Silver: prediction markets are smart
Carl Allen: Mamdani has a 25% chance of winning by a margin of 20-30 and also a 25% chance of winning by a margin of 24+
👌
Carl Allen: Mamdani has a 25% chance of winning by a margin of 20-30 and also a 25% chance of winning by a margin of 24+
👌
November 4, 2025 at 10:06 PM
Nate Silver: prediction markets are smart
Carl Allen: Mamdani has a 25% chance of winning by a margin of 20-30 and also a 25% chance of winning by a margin of 24+
👌
Carl Allen: Mamdani has a 25% chance of winning by a margin of 20-30 and also a 25% chance of winning by a margin of 24+
👌
An actual, published article that people read to try and understand things, in two sentences:
Wrongly informs people that the "most accurate" poll can be determined by a sample size of one election
Says Silver's "56%" probability for Trump was accurate...(???)
This is just garbage.
Wrongly informs people that the "most accurate" poll can be determined by a sample size of one election
Says Silver's "56%" probability for Trump was accurate...(???)
This is just garbage.
November 4, 2025 at 2:27 PM
An actual, published article that people read to try and understand things, in two sentences:
Wrongly informs people that the "most accurate" poll can be determined by a sample size of one election
Says Silver's "56%" probability for Trump was accurate...(???)
This is just garbage.
Wrongly informs people that the "most accurate" poll can be determined by a sample size of one election
Says Silver's "56%" probability for Trump was accurate...(???)
This is just garbage.
Either the method embodies assumptions that hold only in particular contexts (low undecided shares, two-parties), or it is truly general. If it’s the former, the correct course is to specify the domain of applicability and the assumptions clearly, not to proclaim universal adequacy as they do.
October 31, 2025 at 2:02 PM
Either the method embodies assumptions that hold only in particular contexts (low undecided shares, two-parties), or it is truly general. If it’s the former, the correct course is to specify the domain of applicability and the assumptions clearly, not to proclaim universal adequacy as they do.
My debates with analysts in the field take the form of
"But it works fine for Australia and Africa!"
"It's close enough for United States!"
The proper response is... So?
"But it works fine for Australia and Africa!"
"It's close enough for United States!"
The proper response is... So?
October 31, 2025 at 2:00 PM
My debates with analysts in the field take the form of
"But it works fine for Australia and Africa!"
"It's close enough for United States!"
The proper response is... So?
"But it works fine for Australia and Africa!"
"It's close enough for United States!"
The proper response is... So?
This is almost perfectly analogous to what we're dealing with in poll data.
Two and three-dimensional poll data is simplified to this one dimensional line segment: Candidate A, Candidate B.
Undecideds? Project them.
Third party voters? Ignore them.
The "poll data" maps that look like this.
Two and three-dimensional poll data is simplified to this one dimensional line segment: Candidate A, Candidate B.
Undecideds? Project them.
Third party voters? Ignore them.
The "poll data" maps that look like this.
October 31, 2025 at 1:49 PM
This is almost perfectly analogous to what we're dealing with in poll data.
Two and three-dimensional poll data is simplified to this one dimensional line segment: Candidate A, Candidate B.
Undecideds? Project them.
Third party voters? Ignore them.
The "poll data" maps that look like this.
Two and three-dimensional poll data is simplified to this one dimensional line segment: Candidate A, Candidate B.
Undecideds? Project them.
Third party voters? Ignore them.
The "poll data" maps that look like this.
When you reduce dimensionality for the sake of simplification, you subject your data to predictable distortions.
There's nothing inherently wrong with using the Mercator projection, but if that projection leads analysts to conclude "Greenland is bigger than Australia because look at the map"
...
There's nothing inherently wrong with using the Mercator projection, but if that projection leads analysts to conclude "Greenland is bigger than Australia because look at the map"
...
October 31, 2025 at 1:46 PM
When you reduce dimensionality for the sake of simplification, you subject your data to predictable distortions.
There's nothing inherently wrong with using the Mercator projection, but if that projection leads analysts to conclude "Greenland is bigger than Australia because look at the map"
...
There's nothing inherently wrong with using the Mercator projection, but if that projection leads analysts to conclude "Greenland is bigger than Australia because look at the map"
...
If Borat can get this level of compromising footage, which he cut off before it went further, imagine what a skilled operative has on Trump.
He thought the reporter was 15.
No AI needed to make MAGA look bad.
He thought the reporter was 15.
No AI needed to make MAGA look bad.
October 21, 2025 at 1:04 PM
If Borat can get this level of compromising footage, which he cut off before it went further, imagine what a skilled operative has on Trump.
He thought the reporter was 15.
No AI needed to make MAGA look bad.
He thought the reporter was 15.
No AI needed to make MAGA look bad.
Pictured: @vp.govpeeps.us sticking his neck out for Trump
October 19, 2025 at 9:51 PM
Pictured: @vp.govpeeps.us sticking his neck out for Trump
You all are so gullible it's precious 😂
October 19, 2025 at 3:15 PM
You all are so gullible it's precious 😂
If you understand these facts -
One, it proves that any poll with "zero error" is luck, by definition
Because the people who work in the field ascribe undue meaning (they wrongly believe polls are predictions) they think pollsters whose poll spread equal the election spread are objectively better.
One, it proves that any poll with "zero error" is luck, by definition
Because the people who work in the field ascribe undue meaning (they wrongly believe polls are predictions) they think pollsters whose poll spread equal the election spread are objectively better.
October 18, 2025 at 4:16 PM
If you understand these facts -
One, it proves that any poll with "zero error" is luck, by definition
Because the people who work in the field ascribe undue meaning (they wrongly believe polls are predictions) they think pollsters whose poll spread equal the election spread are objectively better.
One, it proves that any poll with "zero error" is luck, by definition
Because the people who work in the field ascribe undue meaning (they wrongly believe polls are predictions) they think pollsters whose poll spread equal the election spread are objectively better.
In political data, this scientific and statistical fact is betrayed
The Pre-Election population - between the poll and the election - can change their mind
Also, the Pre-Election population can be undecided - the election population can't be.
This problem is unaddressed and unsolved until now
The Pre-Election population - between the poll and the election - can change their mind
Also, the Pre-Election population can be undecided - the election population can't be.
This problem is unaddressed and unsolved until now
October 18, 2025 at 4:14 PM
In political data, this scientific and statistical fact is betrayed
The Pre-Election population - between the poll and the election - can change their mind
Also, the Pre-Election population can be undecided - the election population can't be.
This problem is unaddressed and unsolved until now
The Pre-Election population - between the poll and the election - can change their mind
Also, the Pre-Election population can be undecided - the election population can't be.
This problem is unaddressed and unsolved until now
Which leads to the second part.
Polls take random samples from a larger population.
The "true value" is the proportion of that entire population, that's what poll data helps us infer.
BUT
Polls take random samples from a larger population.
The "true value" is the proportion of that entire population, that's what poll data helps us infer.
BUT
October 18, 2025 at 4:13 PM
Which leads to the second part.
Polls take random samples from a larger population.
The "true value" is the proportion of that entire population, that's what poll data helps us infer.
BUT
Polls take random samples from a larger population.
The "true value" is the proportion of that entire population, that's what poll data helps us infer.
BUT
A random sample of 600 gets you to +/- 4% of a population of any size!
Below is a simulation of what ideal polls taken on the same population could produce
By definition, all polls have a margin of error. But defining "ideal poll", we have a benchmark against which we can measure all other polls
Below is a simulation of what ideal polls taken on the same population could produce
By definition, all polls have a margin of error. But defining "ideal poll", we have a benchmark against which we can measure all other polls
October 18, 2025 at 4:12 PM
A random sample of 600 gets you to +/- 4% of a population of any size!
Below is a simulation of what ideal polls taken on the same population could produce
By definition, all polls have a margin of error. But defining "ideal poll", we have a benchmark against which we can measure all other polls
Below is a simulation of what ideal polls taken on the same population could produce
By definition, all polls have a margin of error. But defining "ideal poll", we have a benchmark against which we can measure all other polls
Trying to explain the Ideal Poll concept as a teaching tool for learners, to help them understand the Simultaneous Census as the appropriate true value for any poll, to a bunch of people who think they're experts - but believe that polls are predictions of election outcomes
A short thread
👇
A short thread
👇
October 18, 2025 at 4:10 PM
Trying to explain the Ideal Poll concept as a teaching tool for learners, to help them understand the Simultaneous Census as the appropriate true value for any poll, to a bunch of people who think they're experts - but believe that polls are predictions of election outcomes
A short thread
👇
A short thread
👇
Oh yeah, it's all coming together.
An actual response I got from a peer-reviewer on my poll data work
He wrote about one page of comments, a few useful notes, some requests to lighten my tone as it was "unnecessarily contemptuous"
Then closed with this quote:
An actual response I got from a peer-reviewer on my poll data work
He wrote about one page of comments, a few useful notes, some requests to lighten my tone as it was "unnecessarily contemptuous"
Then closed with this quote:
October 18, 2025 at 2:05 AM
Oh yeah, it's all coming together.
An actual response I got from a peer-reviewer on my poll data work
He wrote about one page of comments, a few useful notes, some requests to lighten my tone as it was "unnecessarily contemptuous"
Then closed with this quote:
An actual response I got from a peer-reviewer on my poll data work
He wrote about one page of comments, a few useful notes, some requests to lighten my tone as it was "unnecessarily contemptuous"
Then closed with this quote:
I'm getting ready
October 17, 2025 at 2:47 PM
I'm getting ready
I'm not seeing any Republicans giving "young men" a pass when it comes to offensive Charlie Kirk jokes
Only when it comes to offensive jokes about black people...Jewish people...supporting the H*locaust and H*tler...
Why do you think that is?
Only when it comes to offensive jokes about black people...Jewish people...supporting the H*locaust and H*tler...
Why do you think that is?
October 17, 2025 at 2:46 PM
I'm not seeing any Republicans giving "young men" a pass when it comes to offensive Charlie Kirk jokes
Only when it comes to offensive jokes about black people...Jewish people...supporting the H*locaust and H*tler...
Why do you think that is?
Only when it comes to offensive jokes about black people...Jewish people...supporting the H*locaust and H*tler...
Why do you think that is?
Many students (and teachers) believe that the margin of error is an abstract concept, true "in theory"
This is not the case.
It is a testable and provable one.
Statistics are not always intuitive, but when we can, educators must make these lessons tangible
I love doing these kinds of experiments
This is not the case.
It is a testable and provable one.
Statistics are not always intuitive, but when we can, educators must make these lessons tangible
I love doing these kinds of experiments
October 15, 2025 at 3:39 PM
Many students (and teachers) believe that the margin of error is an abstract concept, true "in theory"
This is not the case.
It is a testable and provable one.
Statistics are not always intuitive, but when we can, educators must make these lessons tangible
I love doing these kinds of experiments
This is not the case.
It is a testable and provable one.
Statistics are not always intuitive, but when we can, educators must make these lessons tangible
I love doing these kinds of experiments
Remember when war plans were texted to a reporter?
In any other administration, this scandal would still be news
In any other administration, this scandal would still be news
October 12, 2025 at 11:17 AM
Remember when war plans were texted to a reporter?
In any other administration, this scandal would still be news
In any other administration, this scandal would still be news
Based on actual events
The worldwide consensus of analysts on "poll error" are very wrong. And none of them can successfully argue otherwise
Coming soon to a journal near you. Happy to chat
@jrssa.bsky.social @oxfordacademic.bsky.social @aaas.org @royalstatsoc.bsky.social @warwickuni.bsky.social
The worldwide consensus of analysts on "poll error" are very wrong. And none of them can successfully argue otherwise
Coming soon to a journal near you. Happy to chat
@jrssa.bsky.social @oxfordacademic.bsky.social @aaas.org @royalstatsoc.bsky.social @warwickuni.bsky.social
October 11, 2025 at 3:15 PM
Based on actual events
The worldwide consensus of analysts on "poll error" are very wrong. And none of them can successfully argue otherwise
Coming soon to a journal near you. Happy to chat
@jrssa.bsky.social @oxfordacademic.bsky.social @aaas.org @royalstatsoc.bsky.social @warwickuni.bsky.social
The worldwide consensus of analysts on "poll error" are very wrong. And none of them can successfully argue otherwise
Coming soon to a journal near you. Happy to chat
@jrssa.bsky.social @oxfordacademic.bsky.social @aaas.org @royalstatsoc.bsky.social @warwickuni.bsky.social
Do you like science?
Fun fact:
The consensus of experts worldwide in are wrong about the answer to this basic question.
I think the fact that we're on the verge of a revolution in how polls are understood might be understated.
Want to chat about it?
@nprpatrick.bsky.social @jrssa.bsky.social
Fun fact:
The consensus of experts worldwide in are wrong about the answer to this basic question.
I think the fact that we're on the verge of a revolution in how polls are understood might be understated.
Want to chat about it?
@nprpatrick.bsky.social @jrssa.bsky.social
October 9, 2025 at 7:10 PM
Do you like science?
Fun fact:
The consensus of experts worldwide in are wrong about the answer to this basic question.
I think the fact that we're on the verge of a revolution in how polls are understood might be understated.
Want to chat about it?
@nprpatrick.bsky.social @jrssa.bsky.social
Fun fact:
The consensus of experts worldwide in are wrong about the answer to this basic question.
I think the fact that we're on the verge of a revolution in how polls are understood might be understated.
Want to chat about it?
@nprpatrick.bsky.social @jrssa.bsky.social
My wife says:
"Please buy Carl's book so I don't have to hear more about undecided voters and how stupid the f**king spread is"
"Please buy Carl's book so I don't have to hear more about undecided voters and how stupid the f**king spread is"
October 9, 2025 at 12:01 PM
My wife says:
"Please buy Carl's book so I don't have to hear more about undecided voters and how stupid the f**king spread is"
"Please buy Carl's book so I don't have to hear more about undecided voters and how stupid the f**king spread is"