[LS]

I don’t do polling. I found that in 2016 a much better indicator of voting trends was voter registration changes. No, this didn’t work in 2018, largely because a) the GOP had a LOT of candidates who just didn’t really run campaigns, b) probably 3-5 seats in CA were stolen; and c) we lost 20 seats by >2%. That likely won’t be the case in 2020, especially with President Trump on the ballot.

Voter registration DID look much better predicting the senate races, where I got all but one seat (McSally, again >1% and possibly stolen) right.

[Jeff Vader]

I’m in AZ, that seat wasn’t stolen, she was just a terrible candidate who’s probably going to lose the seat again in 2022

[nwrep]

You mean lost 20 seats by <2%, not >2%. Same with McSally, she lost by <1%, not >1% like you say?

[DoodleBob]

Here is an interesting study on voter turnout by the Rising American Electorate (RAE) of unmarried women, Millennials (ages 18-34), African Americans, Latinos, and all other people of color, from 2012 to 2016 and beyond. What's interesting (assuming they got the analysis correct) is that:

-RAE as a block grew more than non-RAE contracted,
-the same is true in PA, MI, and FL -battleground states that Trump won.

Could this mean that "minorities" voted more for Trump than we are being lead to believe?