I saw somewhere between 1-6%, depending on the state. PA is higher than normal because the naked ballot issue.
plausible number but man I do NOT want to have to sweat AZ/NV out there
I’m not 100% clear but I think they only had absentee ballots in 2016 and early in person voting. No universal mail voting. They’ve only had 4 days of early in person voting so far this year. Maybe he will elaborate in his blog but I think the 2 weeks he references might have been the entire early voting period in 2016.
https://twitter.com/ElectProject/status/1318996737810464768?s=20
https://twitter.com/ElectProject/status/1319001318296608772?s=20
I’d think this situation kinda necessitates a reevaluation of polling methodology or at least poll reporting, but the situation is complicated by the known fact that dems and repubs have different plans on when they will vote.
Biden +4 FL and +10 PA in new CNN poll.
12 Likes
35 Likes
Starting to see some effects from early voting:
1 Like
Nate’s ripping off my points but phrasign them better. (And hence the title change this morning.)
1 Like
538 Nate more or less admits (without admitting it’s a shortcoming) that his model doesn’t properly account for early voting. He says that the pollsters account for early voting in how they weight the electorate, and then kind of waves his hands and says that’s good enough. BUT this only accounts for the part of the model that deals with polling error.
It does not account for the part of his model that tries to forecast the uncertainty surrounding the economy, scandals, or anything else that can happen between now and November 3rd. If you took it to some extreme where say, 90% of the people voted early, then clearly you are all but eliminating all the uncertainty surrounding news cycles up to including Nov 3rd. And Nate’s model simply does not account for this.
So what’s all that mean? It means that all else being equal, Biden should have higher odds than what his model accounts for. Note that not all else is equal because Nate’s model does not account for cheating and other rat-fuckery.
5 Likes
Hey guys apparently we’re all wasting our time says this Canadian business man tv dude with BES degree (seriously?).
CNBC: Pollsters are ‘irrelevant’ and ‘do not know what’s going on anymore’ ahead of U.S. election, Kevin O’Leary says.
In other words, he’s reading this forum.
3 Likes
Likely voter screens are likely about to become problematic. They generally treat anyone who has already voted as a likely voter. This means that there are about to be a lot more democratic likely voters than republicans because they are voting so much earlier.
1 Like
It’s been many years since I took stats but shouldn’t “I’ve literally already cast my vote” like do something do the margin of error?
2 Likes
Yup, this is definitely right. As of Thurs 10% of the 15% Trump odds (as of that day) was uncertainly. With this much early voting and more favorable polls, Biden should be like 93-95% currently.
1 Like
I don’t think it’s exactly what you pointed out. I still think you have to dig deeper into the make up of the people that voted early (which is his point about how pollsters may be treating this differently).
Na, the sampling error is, I think, inherent in polling, as with any data analysis based on a limited sample. Most data is not “uncertain” the way likely voters are somewhat uncertain, it’s just actual results from a random subpopulation, which is reflected in the p value.
2 Likes