General polling question. Obviously we know pollsters have adjusted based on how 2016 turned out. I'm wondering if perhaps they have OVER adjusted, thereby inflating how Trump is doing in certain areas. Thoughts?
Combined with unprecedented turnout thus far, this could be a bloodbath if the case.
There is that possibility. People don't ever mention that polls were off in 2012 as well. But in 2012 it as a very close race, Obama outperformed the polls and so it wasn't a shocking moment. So we hope that the fix in education weighting has helped them to be more accurate, but there is always a question as to how effective that will be. Some have said that those changes have resulted in Biden looking worse with Hispanics but a bit better with white voters. I dunno.
But I have stated many times that I think Biden will outperform the polls. I hope that's the case, but then have to temper that with voting complications. So many mail ballots will result in more votes being rejected, which will have an effect. Half a percent? more? Again, no one knows. thrown in other shenanigans and court rulings and I just hope Biden outperforms enough to counter them.
anitram makes a good point about independent voters. You get a lot of different numbers here. Some states Biden is winning them by 20, some by 5. You could guess that northern state ind. voters swing a bit harder to Biden than the southern ones. Either way, they are not split down the middle. I think Biden is winning them in every state by varying degrees.
I keep bringing up Ind. voters cause I think it's really important. They make up about a third of the electorate and when people are looking at ballot returns by party, they often forget about the ind's.
I do think the polls have been pretty good about having representative samples of ind. voters, so not sure it throws polling off much.
I still think we need to trust the averages. And I also think we have to be careful about what polling aggregators we give credence to.
This is why when looking at the big 3, I would rate them like this.
The Economist
538
RCP
Unfortunately most media outlets use the RCP average to do their reporting. And they are just a place that takes a limited amount of polls, with no reasoning on what polls are included and which aren't, and average them. No real analysis or figuring in "house effect" for different polling houses. They also use polls sponsored by SuperPacs, which The Economist does not allow.
For example the "house effect" for a left-leaning pollster like PPP is about .5 to 1point, Qunnipiac about 1-2. While for the right leaning Trafalgar has about a 5-6 point one. And the problem is that there a LOT more right leaning ones. I think THIS is probably the biggest issue right now in looking at the averages, even more so than if the polling this year is more accurate.
Lets look at PA for example.
The Economist - Biden +6.8
538 - Biden +5.2
RCP - 3.9
So I look at 538 more because I feel it's a bit of a balance between the other two, but I think Economist is probably more accurate
To more directly answer your question, There has been chatter amongst the likes of Nate Silver and Nate Cohn about state pollsters possibly shaving a little bit off their numbers to avoid any big mistakes like 2016. Again, conjecture, but I think we can be somewhat comfortable that the state polling averages we are seeing is fair, accurate and hopefully a little bit low.