I've been planning to lay out exactly why and how I think polling is fucked at the moment but here's the main points in super broad strokes.
1) Flooding the zone is a thing. We know they tried it in 2020, and we know they've redoubled their efforts in 2024. An entire bullshit industrial complex of right wing polling firms have popped up since 2016. This fucks up the poll aggregators when they willingly or not let bullshit data into their averages. Something I've noticed in the last few years is that aggregators like 538 who used to track the quality of polling agencies by their accuracy in predicting the outcome of the election they were polling, have stopped doing that.
2) This is made important because MoE is a thing that can be jobbed if you don't give a fuck about accuracy. Run a truly excellent poll with a ton of effort made to remove various sampling biases and you're still looking at a 2% MoE. But people kind of misunderstand what a 2% MoE means. A full 2% shift in one direction is really 4% off. 51/49 turning into 53/47 is technically within the 2% MoE but that's a pretty massive move that is actually a 4 point swing in total. Fwiw, the MoE I'm seeing in most of the polls I've looked at recently are in the 2.5% range. If you were so inclined you could easily massage your polling numbers to show Trump +3 out of data that if unmassaged was actually a dead heat or even up to Biden +2. It's a fairly easy thing to do as well when you understand sampling bias and how (or how not to) control for it...which leads us to:
3) Landline data collecting is still the backbone of the industry. I'm serious. When was the last time you met someone who still had a landline? Who are these fucking people? We know who they are. They're old and white. Which is fine. Some of my favourite people are old and white. But when you look at the cross tabs, old and white are the only demos where the data doesn't seem to generate some massive fuckiness. Like the mail in poll data I showed yesterday that showed 18-29 year olds going 20 points over to the GOP since 2020 and with 41% of the overall poll respondents claiming to not have voted in the 2020 election. At what point does the craziness break your sample of less than 500 people? This was a poll used for the weighting of other polls ffs. Which leads me into my last main point:
4) The 2020 census was fucked. Everyone involved in it knows it was fucked. The data is shit. Minorities were heavily under counted, white home owners over counted by like 2-4% depending on where you were. All....all of the population modeling the polling agencies use to underpin their data collection is based on 2020 census data.
View attachment 21132
Informing a Smarter World
www.prb.org
So yes, I know this seems like wishful thinking but it's wishful thinking in the way of going into a 7 game series as the team with the lower xG but thinking you're still actually the better team because you know the flaws in xG and how your goaltending is way better and you generate way more slot shots, have better special teams, etc and xG isn't properly capturing the story.
