Ha, I'm honestly impressed with the astroturfing and Pravda-style coordination. I don't even think it's all that coordinated; I think the comms powers that be are simply great at following a single storyline.
It's fairly harrowing to see, but also kind of satisfying to watch Americans once again fall for cheap tricks. God forbid the people actually unite against crony capitalism and the rich.
You always get the government you deserve. The accelerationists among us should take solace in the fact that Harris' coronation will move things along.
"September 12, three weeks after the Democratic National Convention, at which point the race will finally enter some sort of a steady state following all of these landmark events."
I like the confidence that wild things will stop happening.
Thought this model talk was great. In a future one it would be interesting to get your (or Eli’s) perspective of some of the nuts and bolts aspects:
* How did you get Eli up to speed on the code and the model?
* What language / libraries do you use? Is everything set up in a single path or are there a mix of technologies (maybe too technical of a question)
* What are the most common data cleansing challenges?
* How do you sense check between runs? Or is writing up and thinking through the changes in factors sufficient given the code and data stability at this point?
Thanks for creating Silver Bulletin and Risky Business have been learning a lot from both.
Moreover...I would love to know your (Nate's) interview protocol. I've already gone on the record stating Eli was a bad assignment hire. 500 words in 20 minutes, any topic? How many tennis balls can fit in a 737max? Empirical evidence suggests Eli was a great choice. I wanna know how that choice was made.
It is time to put an end to Allan Lichtman’s ridiculous 13 Keys to the White House. I will not be able to withstand four more years of Lichtman pontificating about the brilliance of his keys. For the good of the country, we must vote Trump.
What stops it from being a model is the subjective nature of some of the keys. I think Allan Lichtman just has a better than average gut-instinct and he works backward from there with the keys
That is one of the problems with Lichtman’s two economy keys. They are directly related to real economic factors. They don’t account for the people’s perception of the economy, so they ignore any polling where the people say the economy is bad.
I think the current issue is that you have a "selective recession", where the bottom half of the economy is in a recession while high earners are making bank. Modeling that is going to take more effort than devising "13 keys".
Hello Nate (or anyone else reading this). I've been looking back at the 2016 and 2020 election results and had forgotten just how profoundly the polls missed in WI and MI both times. National polls seem to mostly be doing fine, but those both missed by 2-8 points. Is there any explanation as to why pollsters seem to find these two states, specifically, so difficult to poll accurately, and is there any suggestion that they've fixed whatever the problem was?
I’m interested in the “house effects” you mentioned - are these based on who they poll, the nature of the questions, or ??. I took one poll years ago (“it will only take a few minutes…”) about guns in America and the questions were so biased it was infuriating. You could be pro/anti guns and it would have been exasperating - a lot of “have you stopped beating your wife “ type questions. Anyway, after 15 minutes, I just hung up.
Could be wrong, but it is based on pollsters and their historical bias. For example, if a pollster consistently produces results that are shifted right by ~3 points, it reflects that. (Based on historical data from many races)
Sorry to harp, but why aren't there more polls of NE CD2? It must be an order of magnitude cheaper to do than a whole state. 538 has one Biden +5 and another Trump +3, so effectively nothing useful. Nate says Harris probability is 58% but where the heck did that come from? As I predicted, Trump is beginning to struggle in the blue wall but is holding the South and West for now. That would mean Omaha picks our next president. I am a bit curious how it would go, why am I the only one?
How do we deal with several polls seeming to over sample democrats vs republicans? I’m seeing D+6 polls regularly, but Gallup shows there are more people who identify as republicans than democrats. Couldn’t this be an indicator of response bias, which would make sense given new found enthusiasm among democrats?
Has there been any consideration of which third-party candidates will be eligible in which states (or at least the swing states)? For example, the Libertarian Party has pretty-much guaranteed POTUS ballot access in GA, while independents have more hoops to jump through--although RFK has qualified (along with two other independents), and Stein and the Green Party have taken a different approach legalized just this year (parties on the ballots of at least 20 other states are in).
Slight irrelevant note about the Tetragrammaton podcast (honestly the content is enough to justify the 2:25 hour time, it’s damn thorough and wide ranging)… I still get surprised with how normal tobacco advertising is - albeit, chewing tobacco- from the podcast. Like seeing a cigarette ad in a 2013? Pop Sci mag.
This is coming from a country with “plain packaging” laws and duties/sin taxes so high on cigarettes it’s likely causing a spate of arsons and a black market for ciggies
If we're switching to sports topics for a bit, I have an idea that I would like to throw out there. With Michigan football being formally investigated for sign stealing going back several years, how would you go about answering the question of how much of an advantage sign stealing provides?
Question about the cutoff date switch (which makes sense): what would the model have predicted Biden’s odds to be if you used a June 28th cutoff date for the last run of the Biden model? I’m sure the polls before that add some predictive value, but just curious what an apples-to-apples switch like that would yield
The conventional wisdom has always been that 75% of undecided voters will vote Democrat, but with Trump vs Biden, I expected that old orthodoxy would be blown up. Now that it is Trump vs Harris, I suspect that Harris will be the greater beneficiary of the late deciders.
> at which point the race will finally enter some sort of a steady state following all of these landmark events.
DAMNIT, NATE JINXED IT
Yeah I was going to add “…just in time for the space alien invasion”
Lol, I had the same thought, I hope there was some deceased xylem in close proximity he was able smite with gusto while typing that.
We have to factor in how extremely popular and funny and not weird Harris is. The media said so.
Weird Trump/Vance vs Cringe Harris.
Ha, I'm honestly impressed with the astroturfing and Pravda-style coordination. I don't even think it's all that coordinated; I think the comms powers that be are simply great at following a single storyline.
It's fairly harrowing to see, but also kind of satisfying to watch Americans once again fall for cheap tricks. God forbid the people actually unite against crony capitalism and the rich.
You always get the government you deserve. The accelerationists among us should take solace in the fact that Harris' coronation will move things along.
"September 12, three weeks after the Democratic National Convention, at which point the race will finally enter some sort of a steady state following all of these landmark events."
I like the confidence that wild things will stop happening.
Thought this model talk was great. In a future one it would be interesting to get your (or Eli’s) perspective of some of the nuts and bolts aspects:
* How did you get Eli up to speed on the code and the model?
* What language / libraries do you use? Is everything set up in a single path or are there a mix of technologies (maybe too technical of a question)
* What are the most common data cleansing challenges?
* How do you sense check between runs? Or is writing up and thinking through the changes in factors sufficient given the code and data stability at this point?
Thanks for creating Silver Bulletin and Risky Business have been learning a lot from both.
Moreover...I would love to know your (Nate's) interview protocol. I've already gone on the record stating Eli was a bad assignment hire. 500 words in 20 minutes, any topic? How many tennis balls can fit in a 737max? Empirical evidence suggests Eli was a great choice. I wanna know how that choice was made.
*bad ass hire...apologies for autocorrect
It is time to put an end to Allan Lichtman’s ridiculous 13 Keys to the White House. I will not be able to withstand four more years of Lichtman pontificating about the brilliance of his keys. For the good of the country, we must vote Trump.
It's just a simplistic "fundamentals" model with a good PR person pushing it.
What stops it from being a model is the subjective nature of some of the keys. I think Allan Lichtman just has a better than average gut-instinct and he works backward from there with the keys
Look at the current economy. It's very much an "in the eyes of the beholder" question as to whether it's great or terrible.
That is one of the problems with Lichtman’s two economy keys. They are directly related to real economic factors. They don’t account for the people’s perception of the economy, so they ignore any polling where the people say the economy is bad.
I think the current issue is that you have a "selective recession", where the bottom half of the economy is in a recession while high earners are making bank. Modeling that is going to take more effort than devising "13 keys".
This is gonna be a long three months.
Hello Nate (or anyone else reading this). I've been looking back at the 2016 and 2020 election results and had forgotten just how profoundly the polls missed in WI and MI both times. National polls seem to mostly be doing fine, but those both missed by 2-8 points. Is there any explanation as to why pollsters seem to find these two states, specifically, so difficult to poll accurately, and is there any suggestion that they've fixed whatever the problem was?
I’m interested in the “house effects” you mentioned - are these based on who they poll, the nature of the questions, or ??. I took one poll years ago (“it will only take a few minutes…”) about guns in America and the questions were so biased it was infuriating. You could be pro/anti guns and it would have been exasperating - a lot of “have you stopped beating your wife “ type questions. Anyway, after 15 minutes, I just hung up.
Could be wrong, but it is based on pollsters and their historical bias. For example, if a pollster consistently produces results that are shifted right by ~3 points, it reflects that. (Based on historical data from many races)
Sorry to harp, but why aren't there more polls of NE CD2? It must be an order of magnitude cheaper to do than a whole state. 538 has one Biden +5 and another Trump +3, so effectively nothing useful. Nate says Harris probability is 58% but where the heck did that come from? As I predicted, Trump is beginning to struggle in the blue wall but is holding the South and West for now. That would mean Omaha picks our next president. I am a bit curious how it would go, why am I the only one?
How do we deal with several polls seeming to over sample democrats vs republicans? I’m seeing D+6 polls regularly, but Gallup shows there are more people who identify as republicans than democrats. Couldn’t this be an indicator of response bias, which would make sense given new found enthusiasm among democrats?
Has there been any consideration of which third-party candidates will be eligible in which states (or at least the swing states)? For example, the Libertarian Party has pretty-much guaranteed POTUS ballot access in GA, while independents have more hoops to jump through--although RFK has qualified (along with two other independents), and Stein and the Green Party have taken a different approach legalized just this year (parties on the ballots of at least 20 other states are in).
RFK is on the ballot in 41 states.
Slight irrelevant note about the Tetragrammaton podcast (honestly the content is enough to justify the 2:25 hour time, it’s damn thorough and wide ranging)… I still get surprised with how normal tobacco advertising is - albeit, chewing tobacco- from the podcast. Like seeing a cigarette ad in a 2013? Pop Sci mag.
This is coming from a country with “plain packaging” laws and duties/sin taxes so high on cigarettes it’s likely causing a spate of arsons and a black market for ciggies
If we're switching to sports topics for a bit, I have an idea that I would like to throw out there. With Michigan football being formally investigated for sign stealing going back several years, how would you go about answering the question of how much of an advantage sign stealing provides?
How can the model account for the pure exhaustion in the current political climate?
LV polling
Question about the cutoff date switch (which makes sense): what would the model have predicted Biden’s odds to be if you used a June 28th cutoff date for the last run of the Biden model? I’m sure the polls before that add some predictive value, but just curious what an apples-to-apples switch like that would yield
The conventional wisdom has always been that 75% of undecided voters will vote Democrat, but with Trump vs Biden, I expected that old orthodoxy would be blown up. Now that it is Trump vs Harris, I suspect that Harris will be the greater beneficiary of the late deciders.