Trump +2 in CNN/ORC National Poll

[Alcon]

Good lord, those education gap numbers are incredible (perhaps literally).  Using the ANES 2012 survey as a baseline, and pretending the electorates were static, this would be a 28-point Clinton swing among college-educated whites and an 18-point Trump swing among non-college-educated whites.

Again pretending that the electorate is static, that would mean Clinton is losing a net of 36% of non-college white Obama voters, and Trump 37% of college-educated Romney voters.

[Alcon]

LOL

Dems on Atlas, you are not alone! MSNBC are with you, my dear unskewers!

#UnskewCNN

That isn't unskewing, that is actually forecasting based on how the actual electorate will look. Gallup got into trouble last cycle because they had a bad forecast of the electorate.

Because we know for a fact that the 2016 electorate will look exactly like the 2012 electorate, right?

This is exactly what the unskewers have done in the past - weigh the poll to the results they want instead of the results the poll gets.

Did you not learn anything from 2012? You were predicting a Romney win, because the electorate surely couldn't look like 2008 and would be whiter. I don't have to remind anyone how that election turned out, but the fact of the matter is the electorate never regresses in presidential elections.

The electorate never regresses in Presidential elections?  What does that mean?  We've certainly had years when the electorate did shrink, and keep in mind that party self-ID is fluid.  You're absolutely doing the same kind of unskewing Romney supporters did in 2012.

[Alcon]

http://www.dictionary.com/browse/regress

You must be logged in to read this quote.

You're welcome.

I know what "regress" means (as my next sentence made obvious), but what are you claiming?  That it's impossible for the electorate to shrink in a way that decreases the proportion of non-white demographics more than their 18+ growth rate?

With that said, unskewing was totally based on the concept of party ID and nothing else. If you are talking about the actual demographics of the election, then that is where polls can end up being wrong. I'm talking about what Gallup did when they assumed that the 2012 electorate would look like 2000. It has nothing to do with party and everything to do with demographics.

Unskewing was not just about party ID -- quite a lot of it also involved insisting that, regardless of what Likely Voter polls showed, the electorate would revert to have many fewer black voters (and non-whites in general) because 2008 was an unusually elevated turnout.  You're also doing a touch of that.  It's reasonable to weigh by demographics, but it's not reasonable to assert with certitude that the electorate will look like 2012 even if, say, polls are showing depressed enthusiasm and higher rates of failing LV screens among certain groups.