I disagree that weighting can't fix aggregators when it comes to **** pollsters like Trafalgar. You can evaluate their methodology and determine how rigorous and believable it is. You can then evaluate their accuracy from past polls and assign that a value. And finally, you can weight miscellaneous things like funding, minimum threshold of polls conducted, etc.
This appears to be similar to what Morris is doing at 538. He combines these into a weighting scheme that seems to reflect this from what I've read. He also gives the truly garbage outfits like Rasmussen (post Rasmussen) the boot entirely.
now, we can argue until we're blue in the face whether places like Trafalgar should be counted, but looking at their rating gives me the impression it's almost ignored in the aggregate.
Anyways. I think Morris is the best of the bunch and by a fairly wide margin. He seems to care about the numbers and doesn't pontificate like most of the rest. He is active on social media answering questions. He's also be very open that the polls might be horribly broken in a way that wasnt seen before this cycle and isn't understood.
Alright I will amend to better explain:
You are correct that weighting can
help but it doesn't really fix the problem. They can give lesser value to polls like Trafalgar (to think Trafalgar isn't even close to the worst anymore) but they still get weight, and as even crappier polls show up en masse even weighting can buckle under it all. We saw this back in the day with the BCS where a lot of the computer polls that were suspect (why they were included was always a mystery) could lead to some really weird outcomes that flew in the face of a lot of the results we saw. It also lead to some teams/conferences learning how to game the system a bit. Hockey had a similar problem which was why it seemed like every year they had to make a tweak to the formulas to counteract it.
Plus again, the polling itself is not really reliable because it takes tens of thousands of calls (sometimes hundreds of thousands) to get enough respondents to even publish. And those respondents over-represent certain demographics. I mean if 20k Gen Zers register to vote today they are not going to be part of this both because they are too new and because their phone will likely block the call anyways. I found it interesting in the thread I posted how pollsters in Washington knew cells would be a problem back in '07 and still have never found a way to really combat it.
So the problem is compounded. Now I will assume you are correct that the 538 dude appears to do a much better job (I have to remember to follow him) but that isn't saying much when you have Silver who is openly rigging his aggregate to screw with the Polymarket betting numbers for his Lord Peter Thiel and you have RCP which hasn't been reliable in quite some time. (it gets better as we get closer but sometimes I wonder if they can even do basic math) Even 270 to win is suspect as I believe they just moved Texas to a toss up which has to be a troll move I mean I am optimistic but not
THAT optimistic.
Taking a step back, to me it looks like all of this rather nerdy stuff that we used to all love has been co-opted in ways that make it not only way less fun, but also twists it into something it is not supposed to be. The more public they became and the more the news media started to rely on them as a means to try and predict elections the more easily corruptible they became. I think most of the people at 538 are still fighting the good fight (hell they used to refute Nate even back when he was the Golden Boy) but you can tell some realized the same thing a lot of podcasters, radio hosts, news orgs, comedians and celebrities realized...there is way more money in making things look bleak and tilting Right than there is in calling it down the middle. (RCP was the first I remember that really lurched that way) They used to be a tool people like us used to
disprove the BS horserace narrative now they are tool
for the horserace. And as the market gets flooded with more polls every hour it seems even the best have a hard time keeping up. I swear every year polling grows exponentially...and it isn't for accuracy's sake.
(Just my amateur opinion)
I suspect there are three reasons for this that could be ascertained from public records:
