Reuter’s “polling explorer”: http://polling.reuters.com/#poll/TM651Y15_26/filters/LIKELY:1,CP14A1:2
I find this “polling explorer” awful. So someone please check my findings.
Nonetheless..
In their sample of “likely voters” 23% voted for Romney, 30% voted for Obama. That’s not at all close to the actual relative percentages in 2012, it favors Democrats by a lot.
But here’s the kicker....
47% of their “likely voters” didn’t vote for ANYONE in 2012!
What kind of “likely voter” sample is that?
from the link you put up, it looks like there are TWO factors in the demographics: 1. Likely voter *and* 2. voted for Romney.
Of all of those voters (voted for Romney and likely to vote this year), 81.4% say they’re voting for Trump, 7% say Johnson, and 3% say Hitlery.
I did another scenario, except changed the second factor to “voted for Obama” instead of Romney.
Results: 77.1% vote for Hitlery, 10.3% for Trump, 4.5% for Johnson and 3.3% for Stein.
Looks like former Obama voters are more likely (by 7.3 points) to vote for Trump than Romney voters are for Hitlery.
Hope that makes sense.