The Blowout No One Sees Coming

Posted by thomascountz 10/27/2024

The Blowout No One Sees Coming(app.vantagedatahouse.com)

35 points | 48 comments

docdeek 10/27/2024|

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.

dragontamer 10/27/2024||

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.

015a 10/27/2024||

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.

aaronbrethorst 10/27/2024||

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.

gonzo41 10/27/2024|

Well nate silver is still trading off his past success. Now 538 says it's a toss up, so what's the point of that site if it offers no new insight.

caconym_ 10/27/2024|||

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.

RajT88 10/27/2024||

538 is now owned by ABC news.

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)

caconym_ 10/27/2024||

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.

Arubis 10/27/2024||||

A little tangential, but 538 is no longer affiliated with Nate Silver; he sold it to ESPN in 2013.

whimsicalism 10/27/2024||

it’s true that 538 is no longer affiliated with nate silver but he left in 2023, not with the espn sale

frgtpsswrdlame 10/27/2024||||

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.

fluidcruft 10/27/2024||||

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.

PaulDavisThe1st 10/27/2024||

> 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

https://www.nytimes.com/2024/10/23/opinion/election-polls-re...

fluidcruft 10/27/2024||

I'm talking about his Silver Bulletin writings, not things filtered by NYT editors.

whimsicalism 10/27/2024|||

which past success? he’s had so many compared to most pundits

HumblyTossed 10/27/2024||

I just always assume polling is propaganda. There are clearly other countries involved in manipulating this election.

frgtpsswrdlame 10/27/2024||

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.

bryanlarsen 10/27/2024||

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.

https://www.natesilver.net/p/the-state-of-play-in-the-7-stat...

parpfish 10/27/2024||

If polling is as wrong as they think here, I wonder how much effect it’d have in places outside of politics.

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.

andrewflnr 10/27/2024|

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.

MostlyStable 10/27/2024||

I mean...in Nate Silvers' model, the two most likely outcomes are one or the other candidate sweeping all 7 swing states.

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.

xhkkffbf 10/27/2024|

I'm pretty sure it's going to be Harris or Trump. One of those two.

Izkata 10/27/2024|||

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.

xhkkffbf 10/27/2024||

True. Boy that would be fun to unravel.

Izkata 10/27/2024||

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.

https://en.wikipedia.org/wiki/Contingent_election

4b11b4 10/27/2024|||

true but by 1 swing state? or all of em

doctorpangloss 10/27/2024||

Everyone wants easier to predict US elections.

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?

whimsicalism 10/27/2024||

Yes there are plenty of free elections that are easy to predict. The Mexican election for instance was easy to predict

ghosty141 10/27/2024|||

German elections are also rather consistent except for a few years.

ofcourseyoudo 10/27/2024||

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.

whimsicalism 10/27/2024|

Yada yada pre-registering their high-variance bet so they get credibility if it happens to come true.

It’s pretty much a dead heat with maybe slight advantage for Trump.

sherburt3 10/27/2024|

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.

More comments...