Read another poll yesterday that had Trump ahead in Nevada.
Still, there’s that vote fraud issue.
A look at Atlas turnout model
Penn They sampled 9% black. That is an undersample.
And 6.2% Hispanic, also slightly undersampled.
D/R/I 45/37/17 this looks like a Dem oversample, especially with recent GOP registration gains. Probably under sample of Independents.
Women/Men 50/48.7/1.2(other) Probably in line with norms.
Overall this Penn sample looks like it averages out the over and undersamples.
Hitting the important items:
Wisconsin has a surprising Woman/Man turnout model of 54/45. This is at least a bit credible. Farmers may be harvesting rather than voting. But it looks like an oversample. Blacks undersampled about 1% at 6.4%. Does NOT register by party.
North Carolina D/R/I is 29/35/36. She leads with Independents. Black 17%. This looks like an undersample. 53.5/46.5 college degree vs none. This looks like an oversample.
Michigan Woman/Man 52/47 slight oversample of women. D/R/I 29/33/38. Michigan does NOT register by party. Cannot compare to registration data. Atlas seems to be the only pollster who read the Forbes January survey showing Independents ascendant. Blacks 15%, that’s an oversample.
CollegeDeg/None 60/40 clearly an oversample of degreed (these break for Harris).
Georgia sample was 26% black, this is close to the 30% of total population, maybe slight undersample. D/R/I 32/36/31 perhaps a very small undersample of R and I. Woman/Men 55/44 clearly an oversample of women. The Trump +0.3% looks inaccurately low in GA, not by much.