It's nice to see how many other people now notice things only a handful of us used to.

The latest national Marist poll topline shows Harris gaining three point and now shows her up by five. But Trump leads Independents by 10 and my own modeling shows their poll recalculates to 49.78 to 49.64 when accounting for Demographics.

Now that would be if the Democrats actually end up with a nearly two point turnout advantage. Anything less than that and Trump starts leading, and by the time it might get to even or GOP +1 he leads by a decent margin.

The crosstabs that they show that they are looking at a 40/33/26 (D/R/I) breakdown. They claim that those are raw numbers and are adjusting, but guess what? I enter their numbers into the model and put it at 40/33/26 and I show Harris with a five point lead (exactly as the poll suggests). So they are not adjusting from very far from 40/33/26 or they would not still be seeing Harris with a five point lead.

Now I imagine this was a poll that put some wind into the sails of some Harris supporters, as in "what momentum she just gained three points in the latest Maris poll". But with the way so many people now look into these polls deeper, I would guess that most will realize that if is quite literally a "fake poll" at this point. These pollsters cannot get away with this sort of thing as much as once did.

Similarly (but not quite as egregious) is the Quinnipiac poll that shows Harris up three points in North Carolina. In North Carolina there is a GOP advantage that supposedly grew in registration. According to the 2020 exit polls, the GOP held a 37/34/30 advantage. Plug that number in the Quinnipiac demographics and you see Trump actually would lead by 1.2 points (or in line with the RCP average there). That is over a four point difference, which can 100% be attributed to faulty demographic calculations.