Regarding polls...
4 years ago I said landline polls should largely be ignored. Models shouldn't even bother with them. They're useless and, dare I say, not scientific polling!
This isn't unskewing. I don't believe they cross the threshold that a poll requires...that is being a truly random sample. It's not.
In separate news, I personally don't understand the argument for poll shifting. In a highly partisan electorate, what could cause Hillary to go from, say +8 to -2 or whatever? This does not make any rational sense. People are not switching their votes (Hill to Trump that is).
The only things that could change it are:
1. different electorate makeup
2. some hillary voters moving to Gary Johnson because Hillary is viewed as "safe."
For the latter, this will auto-correct plus I think when it actually comes time to vote, people wake up. Plus, people like voting for the winner, not some 3rd party loser.
For the former, I have seen no real plausible argument for it. I'm not talking about just a different electorate from 2012 but one from the polls just 3 weeks ago. Because people are NOT SHIFTING VOTES FROM HILLARY TO TRUMP. It's just not happening.
At the end of the day, the question still remains the same. What will the electorate look like? Obama's or not? And the real question is youth turnout. I believe enough of them will turnout in swing states that Trump is sitting dead (and the losses are offset by higher college educated vote favoring Hillary).
In the primaries, until the end, Trump was mostly overrepresented in the polling and I think this will turn out true on election day, as well (lower educated voters are less likely to vote).
I think Hillary's early voting turnout machine will be formidable compared to the non-existent one of the GOP. They'll get votes that the GOP won't.
So, I'm comfortable right now regarding Hillary's win. And I think once the first debate comes up, you're going to see an "electorate shift" in these polls, again. Response rates, if you will.