For a special Labor Day edition of the Sunday roundup, I’ve written some personal thoughts on journalism, polling, forecasting, and the value of staying off algorithmic social media in 2026.
Nate Silver lost me when his book went from quoting Paul Meehl (giant of mid-20th century psychology) and other advocates of using base rate information to advocating using something akin to prediction markets to produce public policy. He seemed to get lazier and lazier and more willing to give poorly thought through opinions.
Having had to aggregate datasets in my early days, I can confirm that it easily takes a lot of time and labor.
You will always have a niche because journalists all seem to love talking about polls while admitting they're not good at math.
"We ask a lot of interesting questions here that other pollsters don’t ask, since in the relationship sense I am “closer” to the audience than, say, the pollster at a major network news organization (and I would know, I used to do that kind of work!)."
That's the answer.
I think I'm not the only American who is highly aware that our country is falling apart, and that gaining political power back from the fascist felon and his gangsters is the only (faint) hope we have of ever building it back.
Polls that ask real questions and dissect the answers can provide powerful information for future political campaigns and popular government policies. If We The People want to keep our republic, we have to participate, and understand the real issues and problems so we can do better at solving them.
Many thanks for this "open house" on your work and approach to it. We very much share both politics and aversions, so just about everything you conveyed resonates and makes great sense to me.
Only one thing gives me pause. In Section 4, you discuss Cook's conservative approach to forecasting (this has led to a current forecast in which the GOP can hold the House by winning fewer than half the tossup-rated contests, which just doesn't make any historical or vibratory sense, the Trump redistricting +10 notwithstanding).
But, casting back to 2020, Cook (and those, like the NYT, who relied on Cook) were traumatized when, of 27 House contests they rated tossup, the GOP won EVERY LAST ONE OF THEM (and that was in a Democratic year [Biden +7M]; the GOP even picked off six D-Lean, while losing none of their own R-Lean).
I'm not sure what Cook made of that colossal miss (I know for a fact they didn't bother considering the possible impact of either overt or covert red thumbs on the electoral scales), but I'd assume that one thing they did was tweak their sampling algorithm to the right going forward. Who wouldn't?! In fact, every red shift (there are very few instances of blue shift) in this century has challenged pollsters who stay in business by getting it right (no pun intended) to nudge their methodology accordingly.
I'm not saying that our post-HAVA elections have been rigged (though the vulnerabilities are demonstrable), and certainly can't prove it -- after a quarter century working at statistical election forensics. What I am suggesting is that IF such things were going on, the reaction of the polling industry would be to adapt to it by "correcting" their methodological "mistakes." Even if they had not made such mistakes. And in the process corrupting just about the only numerical baseline we might have to assess the honesty and accuracy of the vote recording/counting processes.
And it could be a progressive, multi-step, rolling process -- ultimately leading to a grossly distorted picture of the public will in both vote counts and polls.
I know all too well how uncomfortable this is (for me too), but would greatly value any thoughts you might have about, the Cook debacle specifically, and more generally about the assumptions about electoral legitimacy and the sanctity of vote counts in the computerized voting era.
Thanks, Elliott, for this explanation and for all you do. I’m happy to support this project. Your subscribers get so much from your hard work and dedication!
Thank you, Elliott for your hard (and productive) work every day. I know when I read your newsletter that I am getting an honest, well-thought-out and thorough take on the subject of the day. Thanks for your commitment to your readers and the care taken to cut through the bullshit that many pass off as journalism.
Nate Silver lost me when his book went from quoting Paul Meehl (giant of mid-20th century psychology) and other advocates of using base rate information to advocating using something akin to prediction markets to produce public policy. He seemed to get lazier and lazier and more willing to give poorly thought through opinions.
Having had to aggregate datasets in my early days, I can confirm that it easily takes a lot of time and labor.
You will always have a niche because journalists all seem to love talking about polls while admitting they're not good at math.
"We ask a lot of interesting questions here that other pollsters don’t ask, since in the relationship sense I am “closer” to the audience than, say, the pollster at a major network news organization (and I would know, I used to do that kind of work!)."
That's the answer.
I think I'm not the only American who is highly aware that our country is falling apart, and that gaining political power back from the fascist felon and his gangsters is the only (faint) hope we have of ever building it back.
Polls that ask real questions and dissect the answers can provide powerful information for future political campaigns and popular government policies. If We The People want to keep our republic, we have to participate, and understand the real issues and problems so we can do better at solving them.
Many thanks for this "open house" on your work and approach to it. We very much share both politics and aversions, so just about everything you conveyed resonates and makes great sense to me.
Only one thing gives me pause. In Section 4, you discuss Cook's conservative approach to forecasting (this has led to a current forecast in which the GOP can hold the House by winning fewer than half the tossup-rated contests, which just doesn't make any historical or vibratory sense, the Trump redistricting +10 notwithstanding).
But, casting back to 2020, Cook (and those, like the NYT, who relied on Cook) were traumatized when, of 27 House contests they rated tossup, the GOP won EVERY LAST ONE OF THEM (and that was in a Democratic year [Biden +7M]; the GOP even picked off six D-Lean, while losing none of their own R-Lean).
I'm not sure what Cook made of that colossal miss (I know for a fact they didn't bother considering the possible impact of either overt or covert red thumbs on the electoral scales), but I'd assume that one thing they did was tweak their sampling algorithm to the right going forward. Who wouldn't?! In fact, every red shift (there are very few instances of blue shift) in this century has challenged pollsters who stay in business by getting it right (no pun intended) to nudge their methodology accordingly.
I'm not saying that our post-HAVA elections have been rigged (though the vulnerabilities are demonstrable), and certainly can't prove it -- after a quarter century working at statistical election forensics. What I am suggesting is that IF such things were going on, the reaction of the polling industry would be to adapt to it by "correcting" their methodological "mistakes." Even if they had not made such mistakes. And in the process corrupting just about the only numerical baseline we might have to assess the honesty and accuracy of the vote recording/counting processes.
And it could be a progressive, multi-step, rolling process -- ultimately leading to a grossly distorted picture of the public will in both vote counts and polls.
I know all too well how uncomfortable this is (for me too), but would greatly value any thoughts you might have about, the Cook debacle specifically, and more generally about the assumptions about electoral legitimacy and the sanctity of vote counts in the computerized voting era.
Many thanks!
Thanks, Elliott, for this explanation and for all you do. I’m happy to support this project. Your subscribers get so much from your hard work and dedication!
Well said Elliot. Your making the space and commitment have resulted in a trusted and insightful read each and every time.
Thank you, Elliott for your hard (and productive) work every day. I know when I read your newsletter that I am getting an honest, well-thought-out and thorough take on the subject of the day. Thanks for your commitment to your readers and the care taken to cut through the bullshit that many pass off as journalism.
I knew that Cook is extremely conservative in his forecasts. But I had never seen how conservative. thank you for the data.