Congratulations to AtlasIntel and Trafalgar for another set of outstanding polls. 
If you want to follow the Wall Street Journal Live Blog, that’s a free link.
Once again, AtlasIntel and Trafalgar were the two most accurate pollsters based on margins.
If the current trends hold, with some states uncalled, AtlasIntel will have called every battleground state correctly except Georgia which it rated as even. But even Georgia was within the AtlasIntel stated margin of error.
Trafalgar has Wisconsin and Nevada wrong. I only missed on Michigan.
An average of Trafalgar and AtlatIntel has every state correct but Wisconsin, and that was rated as Even.
I picked Wisconsin as the tipping point based largely on the average of Trafalgar and AtlatIntel, but I have Michigan in the wrong column. I got the national vote wrong.
You-Gov, allegedly high-quality was anything but. The same can be said for New York Times/Siena with Democrats winning North Carolina and Nevada.
Both NC and NV will be hugely out of the NYT/Siena stated margins of error.
Accuracy Assessment
Calling the winning state accurately is not the best measure. Rather, the best measure is how close a call is vs the final result.
AtlasIntel was closest on 4. Traflagar was closest on 3.
I seriously considered adopting the average of AtlasIntel and Trafalgar as my model. It is why I included that line in the first place.
But I didn’t. Instead I averaged in Nate Silver and Real Clear Politics. A big mistake.
Silver can whine about being within margins of error because +-3 he was within margins of error in most states. And did manage to beat “High Quality” NYT/Siena in Wisconsin, Pennsylvania, Nevada, and Georgia.
Congrats?
Two Questions for Nate Silver
- What is the likelihood that Silver and Real Clear Politics (RCP), would not come in first, second, or even third in any state in my lead table, published in advance?
- How do you explain losing head-to-head against RCP in 7 of 8 categories with one tie?
The only acceptable answer to those questions is a self- admission by Nate Silver that his model is broken vs true high-quality pollsters, broken vs RCP, and even broken vs my crude aggregation seat-of-the pants method published in advance.
Seat-of-the-Pants vs Silver
My seat-of-the-pants method had one advantage over Silver.
From February on, I repeatedly stated Trump would dramatically outperform on the youth vote and black vote vs 2020.
He did, by amazing margins.
And that is why I put all the battleground states in Trump’s column except for Michigan which I got wrong as did Trafalgar.
On a win-loss basis, I beat Trafalgar but Trafalgar posted more accurate numbers on average, and that’s what’s really important.
I will do a deeper dive on the deciding factors coming up.
Meanwhile, please consider my sarcastic synopsis Glass Ceiling Breaks, Injuring Kamala Harris
“Mish, you seem to have your thumb on this sort of thing and you appear to enjoy it. What’s stopping you from opening a competing site to Silver’s, RCP, 538, etc? I think it would be right up your alley.”
Thanks Woodsie,
I typed out a long response but deleted it. I will make the Q&A part of my nest post
I like it because I like markets
If you like to do something and can provide something of money-making or entertainment value, people will buy, listen and/or stick around. And you’ll make more money yourself or get more attention. If not, you don’t. We all find out through markets where we stand relative to others’ opinions.
Nate Silver has been around for awhile. So enough people value his opinion or statistical analysis (who knows which?) that he’s able to profitably (?) put out more work
Nate Silver, along with all pollsters, need to find a new job.
I posted on FB the day before the election that polls were almost certainly incorrect by an even wider margin than normal for one simple reason.
Poll after poll after poll after poll consistently showed that support for Kamala among virtually every demographic that typically strongly supports democrats was down to lows not seen since Reagan if not longer.
I knew that pollsters were not adjusting for these dramatic drops in support for Kamala/democrats among those demographics.
As exit polls indicated, that was exactly the case.
The election results proved that in spades as well.
As a result, this was an landslide election not seen since Reagan.
Personally, I’m stoked that we aren’t burdened by four more years of what has been.
I think the difference between you and Silver is you want to be statistically honest and accurate. Silver has an undercurrent of political pressure. Generally polling has been pretty bad of late because of the attempt to manipulate electorate. I always tell people to lie if they are polled and bust up the math
Michigan, Pennsylvania and Wisconsin have voted together every election going back to 1988. I thought it was possible for them to break this year, but I thought if Trump was performing anywhere like what polls seemed to suggest that he would win all three. Maybe I’m getting better at this as I do it longer.
For a bette(o)r viewpoint of polls, try this-
https://triblive.com/news/betting-odds-more-accurate-than-polls-in-forecasting-trumps-win/
So the U.S. unwillingness to allow betting on election results drives the pollsters (or was that hucksters?) in their business efforts. At least they provide a lot of entertaining, though error-filled, evenings.
Excellent and useful analysis. Now I will know who is producing garbage.
Also, can you please fix the BLS?
Nate Silver is a one hit wonder who is still riding the momentum of his great call years ago.
The key to a good pollster, is what you have Mish. A desire to post “Accurate & Timely” information. It makes it actually usable, because of that. More importantly, You have no Agenda, other than “Accurate & Timely”information. That is what destroys Nate, and his ability to turn out “Accurate & Timely” polls, His Agenda! Trying to look good, be first, get clicks, be favored, Etc. ALL leads to garbage information in the end, as it must, being done in that manner!
The media polls are generally suspect because they have an incentive to craft their polls to fit an agenda. That agenda can be a political point of view – like ABC/Disney – or even just to grab eyeballs with a clickbait headline.
The media and university polls usually have to revert to relative accuracy – which means correcting to the right as the election nears. Atlas Intel, Trafalgar and Rasmussen (another accurate one) didn’t have to change much from early-mid October until their final poll – because they didn’t have to scrub their polls of bias. There’s a good argument that pollsters should be judged on the accuracy of their polling 2-3 weeks before the election.
On another site I saw something interesting that said those who had access to the real data knew what was about to happen and quickly sided with Trump.
Musk (changed from Democrat to one of Trumps biggest cheer leaders), Bezos (who refused to allow his newspapers like the Washington post to endorse Kamala) and Zuckerberg (stopped talking about election plans, influence after cheer leading Biden 4 years ago). Those 3 guys run some of the absolute biggest social media and commerce sites. In other words they had access to the data that’s kept about every American. It’s likely they had a really good idea the tide had turned for Trump by looking at comments/searches etc and wanted to be sure they were on the winning side.
I suspect a big reason for Musk’s change is that he “lost a son to the woke hive mind virus”.
Sure because it’s all about money not morals and ethics.
There’s a rumour out there that some polls are deliberately contrived to try and steer the result, and by doing so they influence other polls, knowing that there’s a tendency to herd, and by herding, we really mean wishful thinking, as the establishment clearly has.
There is often an attempt to create a bandwagon effect. I think that is what the poll that had Harris up four points out of the blue in Iowa was trying to do.
I’m a scientist. As far as data goes: garbage in, garbage out. You do the best with what you have. Nate has no control over polls, and as he makes no secret he doesn’t believe every poll out there.
I think it comes down to, and to me is obvious that alot of people didn’t want to admit they were going to vote for Trump, simple as that. An interesting poll for 2028 would be ‘how many people who answer a poll are likely to be lying?’
Nate knows what he’s doing and when he doesn’t know, he says he doesn’t know. He was not the only person to say the race was too close to call.
Read “The Signal and the Noise” by Nate. Easy reading, very well written. Good insights into the 2008 crash (what, not everything finance was looking at was uncorrelated. But but, if a recession comes alot of people are going to default on mortgages. Events like that are highly correlated (linked by cause and effect) which is something the finance modelers somehow overlooked. That would get you an F in any high school probabilty class).
Even better is the discussion of how he used statistical analysis to predict which minor league players would do well in the majors, even though they didn’t seem to be exceptional players at the time. I get the impression, and Nate hints at this, that the ball players thought what he was doing with stats was b*llshit. Which is just as well, no one ever hit a home run thinking about their slugging percentage.
One thing that Nate didn’t go into much is that election results are correlated (no one else did either, as far as I can tell). As in, if polls showed Trump up by 3 in seven states, then there’s a good chance he’d carry 5-6 of them. The data is correlated: if people in one state are voting based on inflation, then probably the other six are going to have pretty much the same view. The effects of inflation, for example, are correlated (almost for sure) from one state to another. To treat it as independent from one state to another is nonsense.
Mish, take comfort in the fact that you’ve been, “unburdened by what has been.” 🙂
All to say that polls are worthless, practically speaking.
…or they are not real …why would that Selzer woman risk her rep by issuing a duff poll? Perhaps she was caught up in placating the angst of those of a particular political persuasion that she shares?
Or maybe the Harris campaign gave her a fat envelope of money to conduct this “slanted” poll to gin up enthusiasm for election day.
Why wouldn’t she? She’s nearing retirement.
To be fair to Silver, a few days ago his model did suggest the most likely outcome was a clean sweep of the battleground states. At the time that didn’t seem realistic but in fact was correct. The problem was he didn’t believe his own model.
The dems ousted Biden and blame Biden for their defeat.
I watched some professors blaming white supremacy, misogyny and the return of the Confederacy (Democracy Now). It’s a pity that they cannot engage in some reflection on the issues which motivated voters. Parading celebrities and purported experts is not going to hack it.
Shoehorning in their hideous and divisive agenda is the prime directive!
They seem to labour under a strange notion that voters are like simplistic oblivious untermensch who can’t see what they are doing and who can’t possibly autogenerate a sentiment of outrage and exasperation with the pompous elite who want to blame everything on a set list of talking points, rather than themselves – they are basically fighting a war but rubbish at camouflage, but refuse to believe it, even when they are dying of bullet wounds inflicted by their inferior enemy (the voters).
That’s what “I’d like to thank all the little people” means. Yeah, untermenschen.
I’m waiting for Martin Sheen and Hot Lips Houlihan to make a commercial begging electors to go rogue like in 2016.
They lost because they had a crappy candidate. To be fair, Biden was also a horrible candidate. Since they won with Biden they thought they could fool America again.
There seems to have been a very large and suspicious decline in (D) voters from 2020….might I suggest the reasons for this get some journalistic scrutiny and actual data analytics?
Could it be that many thought they had been offered up an unworthy candidate selected for them by the big funders?
Sure, but it could also be that drop boxes and mass mailing of ballots to the public lead to more voter fraud
Ya think? 😅
It could, and presumably the same sentiment was present in 2016 and 2012 and 2008, when the vote level was similar to 2024 as well, which might explain why they lost in all but one those years… oh but wait, they didn’t.
Take the top 5 polls from the previous cycle. The rest can be tossed out until they manage to break into the top 5 and should not be taken seriously until they do. This to me is the most logical course of action. It creates tough competition and punishes fraudsters which there are many.
Years ago when NPR still had some respect, they had a well known pollster who explained the industry and how the fraud is perpetrated on the public. It opened my eyes to the machine behind the curtain which will say whatever you want it to say….for the right amount of coin of course.
Polls remain the perfect propaganda snake oil for media companies pretending they are serious journalistic enterprises.
1) They can easily engineer a Poll to produce the narrative these propaganda outlets want to peddle, so the appearance of shoehorning in the agenda isn’t as obvious.
2) Polls are lazy journalism in an era where newsrooms no longer exist (somebody else is now doing the work journalists used to do).
3) Polls are the perfect scapegoat for the blame game and accountability when the narrative and the polling inevitably proves to be incorrect; allowing the “news” to deflect all blame.
4) Polls have just enough mathematical credibility that substitutes aren’t adopted, and like the weatherman repeatedly getting forecasts wrong, they just keep coming right back.
5) Polls are relatively simple and can effortlessly produce the sound bites and glittering infographics that saturate all forms of media and market the propaganda narrative.
Exactly.
Nate is a poll reader and a poor one. He changes his weightings based on the news cycle. Hes terrible at this just like he was as a sports handicapper
Little Nate. Ran 80k simulations and it’s a massive blowout. Some here called this. Others sleep with the fishes
It’s reminiscent of the EU, who like to keep having votes until they get the right answer.
Mish, you seem to have your thumb on this sort of thing and you appear to enjoy it. What’s stopping you from opening a competing site to Silver’s, RCP, 538, etc? I think it would be right up your alley.