Mathematically we should not give that much extra weight to a 70k sample vs a 1k sample. But I wonder if having a sample size this large might help to counterbalance some of the fundamental issues polling faces today. Or, if instead it just magnifies those problems over the larger group.
Skimming the crosstabs I do not see anything truly egregious, but parts of it look decently off from what I'd expect too. The gender gap is small: men are Harris +1, women Harris +8. It has Trump with 22% of the black vote but only 51% of the white vote. Trump wins independents by 3 while losing nationally by 4.
I'll give them credit for doing something different, at the very minimum.
That's a fascinatingly large polling size.
Mathematically we should not give that much extra weight to a 70k sample vs a 1k sample. But I wonder if having a sample size this large might help to counterbalance some of the fundamental issues polling faces today. Or, if instead it just magnifies those problems over the larger group.
Skimming the crosstabs I do not see anything truly egregious, but parts of it look decently off from what I'd expect too. The gender gap is small: men are Harris +1, women Harris +8. It has Trump with 22% of the black vote but only 51% of the white vote. Trump wins independents by 3 while losing nationally by 4.
I'll give them credit for doing something different, at the very minimum.