I think people were expecting more of a ‘bounce’ post-convention (I know I was) , but it appears hers came before , and most analysts seem to agree that the new norm for a post-convention uptick is actually less than a percentage point. Following the RCP averages though , the trend is unmistakeable and she’s on course to go ahead in all of the battleground states pretty soon.
6 Likes
Polls from 538 have had a negative bias towards republicans (ranging from a % to 5+) for the past 4-5 elections, except for 1 or 2 states that have had a negative bias towards democrats (less than a % bias). Harris needs to be up a few points and then it will still be a tossup with the bias incorporated. She has a huge amount of work to do now that the convention is over. I believe Walz has done a few pressers, while she has only done “1” with Walz.
2 Likes
538 doesn’t do polls. It does statistical models in which polls conducted by other companies are one source of information.
There is a wide variety of results from the various polls and so you cannot make a blanket statement like “they have a negative bias towards…”. I dont know whether you are trying to reference general polling error (how much did the polling average miss the actual result by) or electoral college bias (how many points does a Dem need to win the national popular vote by to account for the pro Republican bias in the EC) but neither of those is what happening here. As I’ve already said, the model assumes both candidates will get a small temporary bounce from their convention. That means it is viewing Harris narrow lead at the moment not as a marker of an advantage, but as a predicted situation in this immediate post-convention period that it expects to dissipate in a few weeks. With no change in the polling as soon as the model deems the period of being able to see a post-convention bounce is over it will immediately flip back to putting her in the lead.
A relevant point on polling error though is that the presidential elections of 2016 and 2020 both had polling errors that undercounted the Trump vote. In 2016 it was an error within typical limits, but in 2020 it was really big. While covid was surely a part of that, no one is really quite sure why they were so badly off which has made fixing it a challenge, or even if it represented something that is repeatable and so needs to be accounted for in the polling. Potentially relevant though is that the polls were back to being pretty good for the 2022 federal elections. In short, making assessments based on what we think the polling error is going to be is pretty much guess work even for the experts.
3 Likes
No not the polling error, this is over and above that. The 538 has had that bias for at least the past 4 elections, meaning the results skew against Republican’s in most states (ignoring solid dem/rep states as it’s irrelevant), and then also undercount Democrats in about 2 states.
Not sure that is correct is it, I might be misremembering but I thought Nate Silver was one of the first voices in the 2016 election cycle to say that the polls underrepresented Trump votes?
1 Like
Nate’s argument was more that people were under appreciating how unsurprising it would be for an outcome with a 30% chance actually happening and so people shouldn’t be treating a supposed 70-30 race as certain (was actually closer to 60-40 by Election Day I think)
2 Likes
I might not be explaining myself correctly, so for example….
538 predictions vs reality.
Georgia, Biden by +1.2, actual, +0.23.
So that “error” for all states against reps (different numbers per state), and opposite for approximately 2 states against dems.
I wish I had bookmarked the article. It went back as far as BO election. I spot checked a few of the numbers, and they were correct, so I have no reason not to think the analysis was done in good faith.
2 Likes
Presumably, they will adjust their future prediction model based on the reality vs their previous prediction?
2 Likes
It’s probably self-evident to most everybody , apart from ShitferBrains himself , that racist attacks are probably not such a great idea. This piece digs into why exactly that is so ;
2 Likes
They’re a good idea if your target audience is racist. Same goes with the misogyny.
It doesn’t seem to matter how vile and repugnant Trump and his puppy are, his support remains at 44/45%.
3 Likes
There’s a link at the end of that piece to a study done in 2020 that suggests that by then Trump had already exhausted the most racist part of his base in 2016 ;
‘Using racist appeals and racialized wedge issues won him support most clearly in the 2016 GOP primary. But there are key differences between that contest and the 2020 general election — differences that may make racial appeals less impactful this year.’
Considering he only lost by a whisker in 2020 , it makes it all the more surprising that he would still be trying to scrape that barrel with all the associated downside. You’d have to conclude that he just can’t help himself.
Yeah, maybe he’s just a racist, sexist, homophobic asshole after all.
4 Likes
lol … who’d have known ?
3 Likes
538 nor the other popular models do not directly account for past polling error. Polling error is largely random (in a statistical sense) and so not accountable for other than in your interpretation of the data, but they can and do make adjustments for elements of the data that can produce systemic error. The polling average they all use is not a pure arithmetical average, but a weighted average that adjusts each data point used for “house effects” (e.g. the fact Rasmussen is always 3-4 points for favourable for Republicans than any other quality poll), and weighs each result according to polling quality - a measure of not just how historically close they have been, but how robust and well disclosed their methods are (a small improvement from an A grade pollster will have more of an effect on the average than a big improvement for a C class pollster).
But you cannot just take a past cycle’s polling error and impute it into the interpretation of the current results because as said above the vast majority of polling error is statistically random. 2020 was unusual in that there appears to have been a systemic error across the board that undercounted Trump, and the pollsters themselves have put a lot of effort in trying to understand what happened, but as I’ve already said, no one has great answers. There was surely a unique covid element that can probably be ignored in terms of things that need to be fixed, but the big debate is in how much was due to what they call negative response bias. Pollsters have never believed the press fascination with the so called “quiet Trump voter”, the idea that people hide their preference for him to avoid being socially condemned by polite society, because, well, have you ever met or spoken to a Trump voter? They are not shy about telling you about it. The concern today is more over the system breakdown of trust and unwillingness to participate in society in general of a set of MAGA voters making them harder to reach. If you believe this there are adjustments to your data that can be made, but there is not agreement among pollsters on whether it is real and if so what the appropriate degree of adjustment is.
5 Likes
5 Likes
And Pence will probably not disagree.
Weirdos.
2 Likes
I’m surprised they aren’t attacking him more on senility. If Biden had to step aside, Trump is, if anything, less coherent.
3 Likes
Senility is an acceptable inconvenience for Project 2025.
3 Likes
Maybe they are letting it slide for the public to pick up, and if not, the dr doing his check in the chance that Trump does win the election. I could think of anything less embarressing for him personally, to be told that he is mentally not fit to be president. And also how embarressing would it be for the GOP to find out that they have all been hoodwinked by someone that has lost their mind?
1 Like
Well, that’s plain to see.