(h/t) to Geraghty at NRO for pointing this out.
On the heels of my analysis of polling and their value, we have this poll from Pennsylvania showing Obama leading McCain 50% to 46% among likely voters.
Whoa there Batman! That looks a lot like Rasmussen’s results, doesn’t it?
As they say in the Hertz commercials, “Not exactly”.
Their partisan breakdown for this poll is 53% Democrat, 39% Republican, 8% Independant. Yeah, you read that right. In their infinate wisdom, they have given a 14 point advantage to Democrats for this poll, despite the 2006 exit poll party breakdown of 43D/39R/20I. Now if you take a look at those Rasmussen trends I linked to earlier, you see that nationally there has been a 2% shift toward the Republicans since 2006.
For this poll to be accurate, Pennsylvania would have had to buck the national 2% trend with a 10% trend in the opposite direction.
I can think of only two reasons why ARG would do this:
1) They took a poll of PA which showed McCain in the lead, which frightened either them or their sponsor, so they fiddled with the results.
2) They are just sloppy and didn’t bother to weight their sample.
I now await the inevitible troll to come and explain to use why ARG is the “gold standard” in polling.