I think I found the actual study and their methodology is unsurprisingly dogshit. They don't mention until the very end that the poll was conducted via an online web panel. They hide the fact that they're inflating the numbers by not counting undecided voters. Also hoo boy this footnote