I've made this bet with some of my friends for this election cycle: the polls are wrong, I just don't know in which direction yet.
I'll save this topic at least. In just a few short weeks we will see which theories were correct. But its important to see these predictions _BEFORE_ the election results are announced.
There will be a lot of people after the election pretending to be correct and on the "I Told You So" despite their poll numbers being bullshit. So now is the time to collect arguments and theories.
If your friends took that bet, and if "wrong" doesn't mean "more than 5 points off", then they are suckers, since the historical average polling error is 2-4 points, so betting that "the polls are wrong" without picking a direction and magnitude is a gimme.
There is a small chance of a tie, because of how two states split their electoral votes. If they were both all-in like the other states I don't think there'd be a way to split the swing states to make a tie.
It would result in a contingent election, meaning the House of Representatives votes for President and the Senate votes for the Vice President. It is entirely possible, if extremely unlikely, for both Trump and Harris to be in office - one as as President and the other as Vice President.
No one expects a blowout. If you had expected a blowout, you wouldn’t have changed your infant into nice clean clothes and taken them to the zoo.
Who are these people who predict an electoral landslide for the dems? It feels like they’re trying to set themselves up to be the fivethirtyeight of 2026 and 2028 if they happen to be proven right.
I mean what if it is a toss up? Were we talking about a literal, actual coin flipping, calling it as 50-50 is a better insight than calling it as either heads or tails.
To state the obvious---if nobody was doing the math to get the result that (by certain methodologies) it's a toss-up, we wouldn't know it's a toss-up. Should forecasters stop publishing results as soon as the needle hits 50/50?
Also, FYI, 538 is no longer using Silver's model. He took it with him when he left.
To the point of the article, nobody is polling and reporting on polls out of the goodness of their hearts.
"Everybody loves a horse race" is the maxim which news sites adhere to - news sites are never going to predict a blowout or landslide, because people will feel secure in the outcome and stop reading the news. Every single news site has an incentive to skew more towards calling the race as evenly matched, because that drives page views.
It's a similar talking point to the one about Democratic pollsters not wanting to release their more accurate numbers, because they could lose voter momentum. (Democratic voters love being able to stay home, it seems)
I'm not sure what is the relevance of any of this to my comment. Are you suggesting that the mainstream forecasters (Nate Silver, 538, Economist, etc) are intentionally biasing their models toward a toss-up in order to drive page views?
I find that very unlikely, not least because from their perspective that will probably actually reduce their page views.
Nate Silver and his team are not at 538 any more, they were kicked out by ABC. He's at Silver Bulletin. You're correct he's saying it's a tossup but he seems to be favoring Trump the last few days. And he's also recently writing about fishiness he's seeing when comparing state vs national measures.
> he seems to be favoring Trump the last few days.
"But I don’t think you should put any value whatsoever on anyone’s gut — including mine. Instead, you should resign yourself to the fact that a 50-50 forecast really does mean 50-50. And you should be open to the possibility that those forecasts are wrong, and that could be the case equally in the direction of Mr. Trump or Ms. Harris." - Nate Silver, NYT, last week
Kinda funny that they’ll get clout if it pans out but not because they’ve got a better model or are better at statistics or anything, just that they made an insane prediction and it worked out.
a lot of debate here but one thing is not a debate: both parties want to portray the other one as a devil, in truth the actually both are going to be OK. We had Trump for 4 years, we basically had Harris for 4 years. We are not China or Venezuela. Everyone relax and figure out how to focus on similarities rather than difference. One Love!
More like, they're both equally shitty for the average person. We should focus on how little they've done collectively for the people. If people would just stop registering as one or the other they might be a little more objective about how shit they both are.
Who didn't have much of an answer when I asked him who it was that enacted Medicaid, which paid for the major surgery his son needed when he was born that the parents could not possibly have paid for. He also didn't have much of an answer when I asked him who constantly wants to nibble away at Medicaid, as a testimony to their fundamental opposition to such a "safety net".
The Democrats have lots and lots of problems, including deference to the rich and corporations. But for all their failings, they have consistently attempted to make the lives of regular people a little bit better and sometimes even succeeded.
The problem is we are not holding both parties accountable for not doing their jobs. They have stopped working together. That is a total failure. They all should be fired by all objective measure of doing your fucking job. Those of us who share some ideals with both parties have nowhere to turn. Having said that I can’t in good conscience vote for republicans this cycle. The democrats have finally seen the light and stopped highlighting some of their more polarizing agenda items, while the republicans just continue to double down on the extremism.
Parties don't have jobs in government. Elected representatives do. I don't believe that working together is necessarily a part of the job, though it can be productive and even nice when they do.
If you have nowhere to turn, start a new party, instead of imagining that the role of the existing parties is to cater to your particular political outlook. And yes, that won't work, so first work to change the electoral system.
Herein lies the problem, I want single payer healthcare, I don’t think guns are a problem on their own. I think we should have safety nets, but they shouldn’t be long term handouts. Both parties spend like drunken sailors whether it’s on social programs or the military. Our two party system is fucked.
Pretty low chance of that happening haha. The one similarity I see between both parties is they both have a ton of money in the stock market and so do the corporations that lobby them. So my political stance is to keep following their investments because they have both shown a willingness to do backflips to keep everything moving up.
Many of the reasons political polling is tricky are unique to politics, especially the poll-result-as-propaganda thing. I expect the program of only weird people talking to pollsters is universal, though.
It's obvious for several reasons that political polling is at best nonsense and worst outright fraudulent. Reason 1: people don't change their voting intentions much. So how can polls swing one way and the other so much? Answer must be that they are measuring noise. Reason 2: headline numbers aggregate several polling results, but apparently don't discard known fraudulent data from partisan polsters. So the results can be trivially changed by simply spending money. Reason 3: nobody sane responds to a polster. So they are measuring data from provably insane people. Reason 4: polsters often exclude samples that should be included, for example by requiring respondents to have voted in two previous elections. So the samples is skewed. And on and on. Add to this the incentives in the news industry to make things click worthy and you have the result we observe. Nobody wants to say the emperor has no clothes. There's also a bias in some media to report a close race for fear of being accused of influencing the outcome.
> There's also a bias in some media to report a close race for fear of being accused of influencing the outcome.
This is certainly true. the response I see nowadays if an NYT sponsored poll shows Trump is ahead is… not pretty. People are treating it like a sports game nowadays, wish they would stick to football.
Re your reason 4: the hardest part of political polling is working out who will actually vote. If your polls are going to get judged on their similarity to the final result, you want to filter out the people who won’t vote.
How do you know that people don't change their voting intentions much? The only way you could know what people intend to do is by... umm... polling.
Anyway, the polls haven't actually been swinging much this year, I don't know where you're getting that from. The only significant movement was when Biden withdrew from the race, which you would expect to make a big difference.
I like the dashboarding more than the analysis lol. I think FL is just a red state now and positing it as anything else smells a little fishy to me. Also, I think there's a thing now where all sorts of analysts know they can get a name out of making one big correct prediction that everyone else got wrong. That's lead to a lot of predictions being made with the intention of being contrarian.
Never heard of this company before. LinkedIn shows five employees, most of whom joined in 2024 and a company founded ding in 2023. Seems pretty unlikely that this group has identified a blowout that all the bigger and far more experienced players refuse to point to - could be wrong, of course, and maybe they've got a brand new model that is best in class, but seems unlikely.
All polls have numerous corrections, as simple random sampling is basically impossible.
The way polls actually work is that they build up the "typical voter" (a graph of race, ethnicity, age, past behaviors, etc. etc.), and then sample those groups, and then add up the numbers at the end. This is the dirty truth about polling, its impossible to get clean data so these "cleanup" steps have substantial room for errors.
Worse: pollsters talk to each other, meaning a lot of them share methodologies. So we're likely to see every pollster make the same mistake in the same direction. Its just the nature of how communication works.
---------
This blogpost has numerous claims:
1. Senate Race predicts the Presidential race -- It looks like everyone's senate race numbers match up. But the typical media have
2. Independent voters -- I dunno about this one. I can believe its true, but I'm not seeing how they picked out (or the methodology) behind independent voters. I've also witnessed a lot of behavior where my friends tell me they're independent but then suddenly spout off extreme political viewpoints. I don't understand why people like to pretend they're independent, but... if they do that kind of make-believe or pretend to a pollster, the independent vote number will be wrong.
So the argument almost entirely lies on the Senate vs Presidential race numbers. So there-in lies the question. Are we about to see unprescedented levels of split ticket votes, and are they all going to be for Trump for President / Democrat for Senate?
That.... seems unlikely to me. The Senate Race correlation with the Presidential race is a very strong argument to me. At least within my social circle, I cannot imagine anyone voting for a Democrat in the Senate but Trump for President.
-------
So the blogpost's argument is that we use this new metric (ie: Senate polls) and try to calculate out the correlation to the Presidential race. Its... a new theory but one that I can largely get behind.
I have to imagine that the people are reasonably consistent between Senate Races and Presidential Races with regards to party affiliation.
You're probably right (and I do use the term "probably" in a rigorous way to mean, statistically, when there are 10 pollsters and 1 reports different data, they are statistically less likely to be correct). However: Many would have said the same thing about Nate Silver in 2008 or the minority of pollsters which predicted a Trump win in 2016. The political landscape is in increasingly uncharted waters right now.
That's not true, the most respected analyst says a blowout is reasonably likely. Nate Silver says that there's a 40% chance of a blowout even though the odds are 50/50.
There are 7 swing states, and Nate Silver says that the odds of all 7 states going to the same candidate is 40%. 25% chance Trump takes all 7, and a 15% chance Harris takes all 7.
Essentially, a broad 2% polling error in either direction means a blowout. What are the chances of a 2% polling erorr? Pretty darn high.
For those handwaving that "it's a dead heat", I'm genuinely curious: how do people explain that national polls are 48% to 48% but everyone expects Trump to lose the popular vote? It's estimated he may lose it by as much as 10 million votes.
52 comments
[ 0.24 ms ] story [ 117 ms ] threadI'll save this topic at least. In just a few short weeks we will see which theories were correct. But its important to see these predictions _BEFORE_ the election results are announced.
There will be a lot of people after the election pretending to be correct and on the "I Told You So" despite their poll numbers being bullshit. So now is the time to collect arguments and theories.
I don't know that anyone who is actually paying attention to the data will be surprised by any outcome that isn't more extreme than that.
https://en.wikipedia.org/wiki/Contingent_election
Who are these people who predict an electoral landslide for the dems? It feels like they’re trying to set themselves up to be the fivethirtyeight of 2026 and 2028 if they happen to be proven right.
Also, FYI, 538 is no longer using Silver's model. He took it with him when he left.
To the point of the article, nobody is polling and reporting on polls out of the goodness of their hearts.
"Everybody loves a horse race" is the maxim which news sites adhere to - news sites are never going to predict a blowout or landslide, because people will feel secure in the outcome and stop reading the news. Every single news site has an incentive to skew more towards calling the race as evenly matched, because that drives page views.
It's a similar talking point to the one about Democratic pollsters not wanting to release their more accurate numbers, because they could lose voter momentum. (Democratic voters love being able to stay home, it seems)
I find that very unlikely, not least because from their perspective that will probably actually reduce their page views.
"But I don’t think you should put any value whatsoever on anyone’s gut — including mine. Instead, you should resign yourself to the fact that a 50-50 forecast really does mean 50-50. And you should be open to the possibility that those forecasts are wrong, and that could be the case equally in the direction of Mr. Trump or Ms. Harris." - Nate Silver, NYT, last week
https://www.nytimes.com/2024/10/23/opinion/election-polls-re...
It’s pretty much a dead heat with maybe slight advantage for Trump.
Who didn't have much of an answer when I asked him who it was that enacted Medicaid, which paid for the major surgery his son needed when he was born that the parents could not possibly have paid for. He also didn't have much of an answer when I asked him who constantly wants to nibble away at Medicaid, as a testimony to their fundamental opposition to such a "safety net".
The Democrats have lots and lots of problems, including deference to the rich and corporations. But for all their failings, they have consistently attempted to make the lives of regular people a little bit better and sometimes even succeeded.
If you have nowhere to turn, start a new party, instead of imagining that the role of the existing parties is to cater to your particular political outlook. And yes, that won't work, so first work to change the electoral system.
That doesn't imply that both parties are fucked, nor that even if they are both fucked, that they are both fucked in the same way.
If you want a different electoral system (great idea!), argue for that rather than this silly both-sides nonsense.
i would take the other side of that bet - ie. i believe that we will still be in a democracy 4 years from now.
Basically any company that tries to do market research or any organization measuring public opinion is wrong in systematic ways.
there will be some very exciting opportunities for whomever can come up with a polling methods that works with modern communication.
This is certainly true. the response I see nowadays if an NYT sponsored poll shows Trump is ahead is… not pretty. People are treating it like a sports game nowadays, wish they would stick to football.
Anyway, the polls haven't actually been swinging much this year, I don't know where you're getting that from. The only significant movement was when Biden withdrew from the race, which you would expect to make a big difference.
Are there free elections anywhere that are easy to predict?
In a world where $1.00 in Instagram ads yields $1.02 in donations, does it matter?
The way polls actually work is that they build up the "typical voter" (a graph of race, ethnicity, age, past behaviors, etc. etc.), and then sample those groups, and then add up the numbers at the end. This is the dirty truth about polling, its impossible to get clean data so these "cleanup" steps have substantial room for errors.
Worse: pollsters talk to each other, meaning a lot of them share methodologies. So we're likely to see every pollster make the same mistake in the same direction. Its just the nature of how communication works.
---------
This blogpost has numerous claims:
1. Senate Race predicts the Presidential race -- It looks like everyone's senate race numbers match up. But the typical media have
2. Independent voters -- I dunno about this one. I can believe its true, but I'm not seeing how they picked out (or the methodology) behind independent voters. I've also witnessed a lot of behavior where my friends tell me they're independent but then suddenly spout off extreme political viewpoints. I don't understand why people like to pretend they're independent, but... if they do that kind of make-believe or pretend to a pollster, the independent vote number will be wrong.
So the argument almost entirely lies on the Senate vs Presidential race numbers. So there-in lies the question. Are we about to see unprescedented levels of split ticket votes, and are they all going to be for Trump for President / Democrat for Senate?
That.... seems unlikely to me. The Senate Race correlation with the Presidential race is a very strong argument to me. At least within my social circle, I cannot imagine anyone voting for a Democrat in the Senate but Trump for President.
-------
So the blogpost's argument is that we use this new metric (ie: Senate polls) and try to calculate out the correlation to the Presidential race. Its... a new theory but one that I can largely get behind.
I have to imagine that the people are reasonably consistent between Senate Races and Presidential Races with regards to party affiliation.
There are 7 swing states, and Nate Silver says that the odds of all 7 states going to the same candidate is 40%. 25% chance Trump takes all 7, and a 15% chance Harris takes all 7.
Essentially, a broad 2% polling error in either direction means a blowout. What are the chances of a 2% polling erorr? Pretty darn high.
https://www.natesilver.net/p/the-state-of-play-in-the-7-stat...