2. Don't cringe at advertising your commentary. It's what hooked my on 538 and why I followed you when you left. The model is great, but it's your immunity to the 'buzz' that makes up so much commentary that is well worth the subscription.
I will note that I’m disappointed about the congressional and gubernatorial model news. Particularly with both the house and the senate very closely divided currently, the stakes (while lower than the presidency) are quite high. I often find myself spending more time looking at the congressional forecasts than the presidential one. I understand that you are limited to the same 24 hours within the day as the rest of us, but I hope you’ll reconsider this choice. Perhaps there is a way you can provide fewer updates and even less commentary on this model, but still run it occasionally for those interested in perusing it.
Just wanted to voice this piece in case there are others who feel the same as me!
The other factor, if we're just being honest, is that the Congressional model requires a *lot* more data than the presidential model does: fundraising data, expert ratings data, biographical data on ~1000 candidates, demographic data on Congressional districts, etc. So it requires a much more streamlined data pipeline than I have now, plus somebody who's devoted ~full time on the Congressional model from now through November.
I can imagine it working if I find the right candidate who's basically like "just give me the code and I'll take this off your hands". But that feels like a challenging hire to make, one where I'd have to get very lucky and/or devote a lot of time to finding the right candidate. Maybe it will feel less challenging one the presidential model is up and running, but I just felt a lot more comfortable once I recognized that I probably had to punt something, and resurrecting the Congressional model (and the non-NCAA sports models) is probably more of a project for after the election.
I think we can guess that the House will follow the presidency, especially if Republicans win the latter. If there's a significant chance of that not being true, it's something I would expect Nate to write a post on, whether or not he's running a proper forecast on it.
It will be very hard for the Democrats to hold 50 Senate seats, but if Tester and Brown can somehow both win it's very likely they have the other 48. And if Democrats hold the Senate they'll almost certainly win the House (and obviously the presidency). This is also something I'd expect to get a post on even if not a proper forecast.
I suspect Nate is likely to end up doing more than he promises (and have misgivings about it) than do less :)
Not all the prediction market sites have predictions for the House yet, but Polymarket does and it gives Biden a 45% chance of being President while a 56% chance of Democrats taking the House, so it's not perfectly aligned.
I will echo too that your commentary is why I’m a subscriber, and I definitely don’t always agree with you - but I value it. I commented in the past about this, but I pursued and earned a Statistics PhD in no small part due to how you hooked me as a teenager when I was reading analysis from Poblano on daily kos haha.
All that is to say, congrats and looking forward to it all!
Does anyone else remember the website “Is Nate Silver a Witch,” which was just a soothing gray background with the single word “yes”?
I used to pull up that website and smile at it goofily before jumping into poll analysis at the old 538 and while I am already happily paying $80/year here it would totally be worth the $95 to continue reading all the nerdiest deep dives.
Nate
1. Woohoo on the news!
2. Don't cringe at advertising your commentary. It's what hooked my on 538 and why I followed you when you left. The model is great, but it's your immunity to the 'buzz' that makes up so much commentary that is well worth the subscription.
This is exciting news!
I will note that I’m disappointed about the congressional and gubernatorial model news. Particularly with both the house and the senate very closely divided currently, the stakes (while lower than the presidency) are quite high. I often find myself spending more time looking at the congressional forecasts than the presidential one. I understand that you are limited to the same 24 hours within the day as the rest of us, but I hope you’ll reconsider this choice. Perhaps there is a way you can provide fewer updates and even less commentary on this model, but still run it occasionally for those interested in perusing it.
Just wanted to voice this piece in case there are others who feel the same as me!
I appreciate that, Kevin.
The other factor, if we're just being honest, is that the Congressional model requires a *lot* more data than the presidential model does: fundraising data, expert ratings data, biographical data on ~1000 candidates, demographic data on Congressional districts, etc. So it requires a much more streamlined data pipeline than I have now, plus somebody who's devoted ~full time on the Congressional model from now through November.
I can imagine it working if I find the right candidate who's basically like "just give me the code and I'll take this off your hands". But that feels like a challenging hire to make, one where I'd have to get very lucky and/or devote a lot of time to finding the right candidate. Maybe it will feel less challenging one the presidential model is up and running, but I just felt a lot more comfortable once I recognized that I probably had to punt something, and resurrecting the Congressional model (and the non-NCAA sports models) is probably more of a project for after the election.
I think we can guess that the House will follow the presidency, especially if Republicans win the latter. If there's a significant chance of that not being true, it's something I would expect Nate to write a post on, whether or not he's running a proper forecast on it.
It will be very hard for the Democrats to hold 50 Senate seats, but if Tester and Brown can somehow both win it's very likely they have the other 48. And if Democrats hold the Senate they'll almost certainly win the House (and obviously the presidency). This is also something I'd expect to get a post on even if not a proper forecast.
I suspect Nate is likely to end up doing more than he promises (and have misgivings about it) than do less :)
Not all the prediction market sites have predictions for the House yet, but Polymarket does and it gives Biden a 45% chance of being President while a 56% chance of Democrats taking the House, so it's not perfectly aligned.
I will echo too that your commentary is why I’m a subscriber, and I definitely don’t always agree with you - but I value it. I commented in the past about this, but I pursued and earned a Statistics PhD in no small part due to how you hooked me as a teenager when I was reading analysis from Poblano on daily kos haha.
All that is to say, congrats and looking forward to it all!
Your commentary is worth the price of admission.
Any plans or ideas to talk on a podcast? Miss hearing you talk with Galen and stupid jokes, scoffs, and data-driven dismissal of stupid takes.
Anything coming on the NBA playoffs! Playoffs
Hooray! It all sounds awesome. I hope to hear a certain whiz kid on your pod, at least as a guest. 😉
Your transparency and obvious rigor are commendable and appreciated! Thank you for the Silver Bulletin!
Exciting news
i eat cats
Thank you for not letting me click refresh on the model over and over this year, even though I will want to.
Does anyone else remember the website “Is Nate Silver a Witch,” which was just a soothing gray background with the single word “yes”?
I used to pull up that website and smile at it goofily before jumping into poll analysis at the old 538 and while I am already happily paying $80/year here it would totally be worth the $95 to continue reading all the nerdiest deep dives.
$80 well spent. I'm already in.
Yet the question nags, how do you ensure the integrity and of your raw data? Bot or not on a level even you cannot detect any more.
"And so it begins"
Has the election model been published at this point, and if so, where can we access it? Thank you.