You have the best race car driver in the world at your disposal.
You give him an F1 car, a sports car and a Jeep and tell him to bring his own personal car to the track as well.
He will have varying times won’t he, depending on the car.
Same way the pollsters are only as good as the car they are given, which in this case is information.
Modern campaigns have an access to a plethora of information about individual voters which they can use. This is expensive as fuck and only the highly sophisticated resource rich political campaign can get them.
Which is why they can see trends which may not be apparent elsewhere, having such detailed information about individual voters.
There is no question this has been a huge miss by pollsters at least in certain states. Trump seems to have won Ohio by 8, same as last time. The polling averages showed it as a +1 Trump state so a 7 point error.
In Wisconsin the polling average was +8 for Biden and even if he pulls it out, his margin will clearly be a lot smaller.
Selzer nailed it with their +7 in Iowa which drew a lot of comment. I am wondering why they don’t expand into nearby states which should great improve poll quality in the Midwest.
Honestly. I think it’s all bullshit. It’s just a way for a campaign to deflect a topic or justify something. Internal polling and public polling are the same, MAYBE it’s more hyper focused on a local region, but the basics are the same. When a candidate or a surrogate says “well, our internal polling says something different…” to a reporter it’s basically a brush off. It’s them saying, you’re wrong and I’d prove it but it’s “internal” so I can’t.
Before you decide whether you trust the polls, you need to decide whether you trust the result.
And potential hacking aside, note that the 538 model explicitly did not factor in all the voter suppression and nullification efforts - legal and illegal - the Republicans were openly engaged in across the country, from fake ballot dropboxes to intimidating robocalls to severely limiting polling sites etc etc etc - there’s no doubt it had a significant effect (and, sadly, proved to the GOP that it was effective - look for more and more each election).
It’s all very well to talk about the odds-on favorite of the Kentucky Derby but if there’s a couple of guys in the bushes with rifles sniping at the horses it’s a little unfair to blame the bookies for getting the odds wrong.
People want to know projections, so poles will always be sought out.
A campaign’s private polls are not ‘more accurate’ than public polls, they are more targeted. A campaign knows who the swing voters are (or who might vote) and will focus their polling on those groups so they better know who to target for rallies, phone calls, ads, etc.
The difference between this time and 2016 is that the national polls were fairly accurate last time and only a few critical swing states had larger margins. Now, national polls are off by way more, and the polling error seems to be everywhere. Just looking at states that have reported almost all their vote, OK is off by 8, DE is off by 8, IO is off by 7, FL is off by 6, and OH is off by 7 and a half. Those are some pretty big misses.
This is why I can’t escape the idea that something about the results are being fucked with. Either the polling is wrong in a extremely consistent way, or something else is going on.
How do people think Nate Silver got his job? Through some meritocratic process, or some long track record of accurate predictions? He has never, at any point, been a more accurate predictor than anyone else looking at the RCP poll average and making a guess.
He told a story that sounded good – “like moneyball but for politics” – and he has continued to draw clicks. He has done, and likely will continue to do, the job he is paid to.
As far as polling in general – generally speaking the pollsters who were most accurate in 2016 were most accurate in 2018 and again in 2020. The pollsters who were least accurate then were most accurate now.
If your media sources use consistently use inaccurate pollsters – and especially if they disparage the ones who are more accurate – you may want to think about why that might be.
The polls might or might not be utter crap, but why the fuck is Nate Silver at fault?
He’s not even making predictions. His team simply (this is oversimplified, but based on the conversation so far, I’m leaving it at that level) has models that take polls, weigh them historically for accuracy, apply those weights to current poll results and spits out weighted results. If the polls all suck, the results output from his model will be wrong, but that’s a flaw with polling, not the model. If the polls all sucked equally, we wouldn’t need the Nate Silvers of the world. If the polls that have done historically well are currently doing poorly, then there’s a discussion to be had. On 538, they are quite clear:
Of course, there’s always the chance of a polling error, which tends to be correlated from state to state when it happens. Trump needs a bigger-than-normal error in his favor, but the real possibility that polls are underestimating Trump’s support is why he still has a path to win reelection.
He also has blog posts specifically laying out this possible outcome:
You can argue that there are better methods, but you should at least bring these methods to the table. You should also educate yourself as to what the various people involved are responsible for. Nate Silver isn’t involved in the polling at all, he’s simply attempting to interpret it in a scientific manner.
I don’t think that the polls were really that far off in 2016; Hillary won the national vote, which had been predicted. What was less predictable was how the state-level vote would unfold, and that was where the surprise was. Nate Silver was among the few who saw the potential for a stunner.
But yeah, this time, the polls really could be off. By how much remains to be seen, but normally a 9-10 point polling advantage would be a landslide election. Instead it’s a nail-biter even if Biden wins.
I don’t blame Nate Silver. He just takes the data available to him and works with it.
The problem is the data itself.
Like the old expression from the earliest days of computing : GIGO–garbage in, garbage out.
My guess is that the whole concept of polling may have become a tool that is irrelevant due to changing technology and social customs…
We’ve come full circle. The first polls in the 1940’s were inaccurate because they used the telephone…and thus missed half the population who did not yet have phones.
Now, the polls use the telephone…and miss half the population because we have too many phones, and use them way too much. We block calls from unknown sources, and are very,very suspicious of any callers asking for personal info…
Yep, everyone’s freaking out a bit prematurely.
According to their Secretary of State, in Pennsylvania alone, there are over 1.4 million mail in ballots that have yet to be counted. Of the mail in ballots already counted, 78% were for Biden (Biden - 879,768, Trump - 252523). I’m not saying that the ratio will hold, but if it does and if there are no movements in the in-person ratios, Biden actually takes Pennsylvania, which he might not even need depending on how a few other states play out.
I’m with Doors on this one (not worried, but aware of the possibility), even though the Steelers suck. 
While you didn’t call him a pig-fucker or anything like that, you did say the following, which looks like a dig at 538:
Either way, I was mostly calling out stuff like:
They were also wrong for the senate. So far the Ds have 47 and reasonable chances in 2 more (ME and GA special). With only 49 senators, even if Biden wins he won’t be able to do anything, even name a cabinet without McTurtle’s agreement. And his only interest will be in holding on for 2022s and beating Biden (or Harris or ?) in 2024. Actually getting anything done is not in his agenda.
And no, I will probably ignore polls henceforth.
I can confidently say that I sure as fuck won’t ever trust them again.
I think a lot of people don’t understand uncertainty. Even the best polls can’t tell you what will happen. And good polls are upfront about that.
If a poll tells you that there’s a 95% chance that something will happen, and the thing does not happen, that doesn’t necessarily mean the poll was wrong. If things that had a 95% chance of happening invariably always happened, that would be wrong.
I don’t recall ever hearing that internal polls are more accurate than public polls, and I don’t think anyone really thinks they are. My understanding of internal polls is as follows:
Polls are not done by just polling a lot of people and counting the results. They require a lot of manipulation before being released, where the pollsters look at various aspects of the sample and adjust it to ensure that it’s representative of the voting population. That manipulation is more of art than science where a lot of room for error comes in. Because what you think the voting population might be incorrect.
In general, internal polls tend to be more favorable to the candidate than neutral polls. This is in part due to a desire to paint an optimistic picture for the candidate. But I think it’s also in part because the campaign people tend to genuinely believe in these optimistic scenarios and the pollsters go along with that. For example, suppose the campaign has a strategy to up their ground game and increase turnout of their people. Their goal is to increase turnout of their likely voters by 2% and they think it’s likely they will accomplish this. If the pollsters go along with that assumption and up their assessment of the electorate on that basis, their data manipulation will be different than that of another pollster who wasn’t buying into that notion. Or perhaps they have a message tailored to a certain slice of the electorate and this that will influence their turnout. And so on.
This, I think, is mostly due to selection bias. The candidate isn’t going to release internal polling results that look bad for him.
[quote=“DMC, post:32, topic:924714”]
weigh them historically for accuracy
[/quote]He weights the polls for a number of factors, including accuracy, but also his subjective opinion about their process.