These may not be the best numbers, but they are below average errors, i.e., as I suggested previously, the average error in the last week's national polls for all presidential elections between 1952 and 2020 is 2.5 points, and it is larger still in the states.

Whichever way those voters skew, politically, is far less likely to be picked up accurately by a pollster. Selzer, for example, famously refuses to contact voters via any method besides a live call — in an age of seemingly nonstop spam calls, it’s a method that just feels inherently dubious.