... also, with a 3,000 person sample size, random fluctuation is minimized. Just the two days since the convention make up 950 responses.
They ask anyone 18 or older, and do not ascertain if anyone is a registered voter, or what party, or likely voter. So it’s a fairly strange method.