“All we heard about was “That poll is crap. D+8? YEAH RIGHT!!!” All we heard yesterday was story after story of record turnouts.”
The “polls are oversampling Dems” argument was really one of the more frustrating things to see repeated over and over, here and on every conservative site. It’s like no one understood that the vast majority of pollsters do not weight their polls to a certain D/R/I sample, and it would not make sense to do so (since party ID is flexible, often changes with one’s voting preference, and is one of the things pollsters are trying to measure - weighing to a certain party ID target makes only slightly more sense than weighting directly to a certain result).
The simplest explanation for why virtually every poll this season had a party ID spread of D+5 to D+10 is that the party ID spread of the electorate was between D+5 and D+10. Occam’s Razor.
Sadly, writing anything like this before yesterday would have been enough to get people zotted around here. Hiding from the facts does not make them disappear. Denying the facts does not help us win. That must be one of the primary lessons we take away from this loss.
Exactly. I remember some on here trying to explain that (that polls just report the weight...not determine it). But...people didn’t listen...and unfortunately I was one of them. I repeated that mantra to my friends...and now appear to be a big ‘ol idiot.
Perhaps, but it is reasonable to question whether the polling was correct and if there was some common flaw in the polls done by the various polling places.
Lets try a thought experiment. Lets assume I did a random sample of the US population and the results were that I received responses from 90% men and 10% women. Although, such a result is possible for a random sample I would need to question whether my method of selection was in fact random.
When the current crop of polls showed a stronger democrat response than what would be expected, it is perfectly normal to question whether the sampling methods were flawed. Unfortunately for us we do not have access to the actual methods these pollsters use in order to see if they are applying a different methodology. We also have instances in the past where these pollsters have erred.
One such reason that made sense to me that could have caused a common error for all polls was the greater emphasis on cell phone contacts. Being a newer method I could understand that wrong assumptions on who the population was could have an unknown effect on the sample.
Did Nate have a statistical method to determine if the sample internals were representative? Were the samples actually flawed like many thought and the polls were actually wrong but the result was affected by other factors?
Nate ended up being right.