“Quite simply, they are oversampling Democrats”
The oversampling is supposed to reflect the fact that in the past election, a higher proportion of Democrats turned out to vote than Republicans. Is that correct?
If so then the oversampling concept is valid, correct?
But one can argue whether the amount of oversampling used in current polling is realistic. It’s possible that in the upcoming election Republican voter turnout will be higher than Democratic, in which case they should be oversampling Republicans.
Yes, but not by 6, 8, or 10 points. If you look at the polls, if they oversample by three or four (which is the traditional number) then Romney is up by one.
Some polls they oversample by 10 now and put Obama up by *one*.
It is almost as if the pollsters are being paid to find data to fit a desired result.