I’ll try explain it better, and also know that the question could be irrelevant if I am not understanding this particular poll.
Is the poll, +4 nationally lead, for the popular vote. If it is, is that +4 lead good enough when compared to Bidens and Hillary leads in 2020 and 2016. If it has nothing to do with popular vote, then ignore the question.
Yes it is and no it isn’t. (good enough)
I might be wrong, but I think I remember reading an analysis that polling in general is more accurate than it has ever been, it’s just that the minor quirks in the landscape e.g. around electoral votes is complicating the picture a lot, and people place too much value in the statistics without understanding the margins of error.
I’m not sure he has ever been honourable, even prior to Trump.
Your post from above linked to a poll.
It seems like you are taking this one quite seriously, but now you aren’t.
The problem with the popular/national vote results is they don’t tell you the winner. Polls have been more accurate getting that right than the winner. A +4 could be a resounding Harris or a remarkably solid Trump win
The lead is bigger than the final 538 polling average advantage for Biden (not that such a comparison is particularly useful, but it I get it is intuitive). People also forget how much the polling average in 2016 tightened right at the very end and remember that race based on the summer and autumn projections that set the narrative rather than the actual final data
Examples
Of course hillary lost so a 4 point national popular vote lead is not a guarantee, but that is not what polls are supposed to be used for anyway. The main relevence of this poll is that it their methods have produced the most favourable results for Trump of any of the Grade A independent polls and it is showing an advantage for Harris that is not just approaching the size where a lot would have to go wrong for it to not be enough to win, but is also showing a good trajectory in both the top end results and the underlying data explaining why she is now pulling ahead.
But Ben Wikler, the head of the Dem Party in WI, someone with a fantastic record of success in this sort of toss up environment, has a great phrase for races like this - the polls are not within the margin of error, but margin of effort, so go out and knock some doors and volunteer in GOTV efforts
With the caveat that all the data over this cycle suggesting the electoral college advantage for Republicans has shrunk.
Another issue for the polls is that there are many more polling “organisations” that have no interest in being right. They have a narrative to push and place on various news channels to discuss the results of their clearly fradulent poll.
An egregious example is “Patriot Polling”. Ran by two high school kids with a pro-Trump agenda and consistently overstating Trump’s support. Of course the genius Nate Silver goes and throws that into his algorithm.
A less obvious, for the neutral, but equally bias source is Rasmussen.
I’d bet good money that a lot of these polling companies aren’t even talking to anyone, probably just working with betting companies to set narratives that are profitable.
I don’t know how many there might be that simply are fudging the data, but certainly the ‘moving aggregate’ method is now far less reliable because of the increasing number of bad-faith pollsters. It is not that hard to use framing questions to skew results even for legitimate samples of 500-1000 respondents.
I hope you’re right about that because as it stands his polls have been the most consistent , and encouraging , for Harris.
Opinion polls are not published to inform public opinion, but to influence it.
Oh, but he “adjusts for house effects”, making all actual statisticians piss their pants at the fundamental violation of thinking you can adjust for inherently bad data.
He also rejects the premise that there is any other incentive other than to be accurate, which anyone who pays attention to the actual politics of it all knows is bullshit, and responds that “if Dems think that is that bad then they can just go out and respond with equally bad polls”, which if he really believes is a valid response makes a mockery of his entire practice.
I dont think there are any that are purely fudging, but as hard as it is to develop methods to produce a representative sample, it is incredibly easy to develop methods that give you exactly the results you want.
The one drills down, while the other is national lead (popular vote) which I find uninteresting as Harris has a huge probability of winning this. Some polls are more interesting than others, same goes for people if you looking for a comparison.
Thanks. As you mention, polls tightened for Trump in 2016 towards the end, and as far as I can see this also happened in 2020 (battlegrounds). The tightening started at about this time.
I was under the impression that he adds weights polls based on their historical performance. So poll quoted above (pro rep) would be weighted as weak, along with pro dem polls. I think a good example of a consistently left leaning poll would be Bloomberg. As stated way above somewhere, a poll that is accurate this year, might have been inaccurate in 2020, 2016, or might be accurate in one battleground and inaccurate in battleground.
I don’t know any other way other than weighting polls that would help get a more accurate result. Maybe there is?
I think 2020 needs to be taken with a grain of salt as the circumstances were extraordinary, and will most likely (hopefully) not be repeated in a lifetime.
Anyway, a month to go. I got 2008 right, ignored 2012, got 2016 and 2020 wrong, so maybe 2024 just to get me closer to a coin-flip in predictions.
If all you look at polls for is to see who is winning then it is best you ignore them altogether as you will very rarely get what of value they have to offer to understanding what is happening.
Yes, this is what I said. Importantly though, House effects (systemic lean towards one side) and accuracy (how far are they from actual results) are not the same thing. Most of popular averages use weighting on one or both of these factors, which means a weighted average of two polls that both report a 50-50 race will not necessarily average out at 50-50. But the thing is, those factors have already been taken into account in those averages. Reapplying them to the averages in your own personal mental model is nothing short that sticking your finger in the air and guessing at an outcome and pretending your gut feel is the result of a rigorous process.
It is reasonable to question the methods in calculating the weighting, or even question the validity of the data that goes into those averages (what the quoted part of the discussion was about), but it is one thing to identify those as issues. It is another to be able to apply an accurate adjustment and no one here has anything like the knowledge to do that. those are sophisticated conversations that no one on here has any real insight into solving.
You may think you’ve stated it. You may have actually stated it. But it isnt at all how your express your thoughts about what is happening. You routinely “account for past errors” by simply applying those misses to the current polls in your own mental model to make predictions on the outcome, which is as intuitive as it might be is hopeless wrong. Like all “polling” it might still allow you to back into predicting the correct binary outcome by complete luck, but it doesnt reflect a particularly meaningful understanding of what is happening.
This is what I do. I just cant bring myself to ignore the poll inaccuracies from the past, especially if they are consistently wrong. I agree it might be statistically/mathematically incorrect to use past inaccuracies, however I don’t know of a better intuitive method.
If for example, Bloomberg poll is overstates dems by 5 points on average for the past 4 elections, I think it will be overatate by close to 5 points this time.
I believe what the soon-to-be-underwater member of this forum is saying, is that no matter how intuitive it might seem to you, it would still be wrong except by pure luck. And therefore it would be a little wrong to keep making such strong assertions about it.
