Your math may be a little bit off.
Only 5% of Democrats are going to vote for Romney? That’s supposed to be impressive?
So 95% of Rats are going to stick with Bonzo.
Are we to believe that 100% of Republicans are going to vote for Romney? How naive is that?
In reality, there will proably be roughly the same % of crossovers both ways. You can find polls (even reputable ones) to document that. So take off that 5 from your equation.
Is R turnout really going to be +6? Not even the most delusional poll has R enthusiasm THAT high. +1 or +2, maybe.
11 percent oversample? Where is it written that there are exactly the same number of Ds and Rs in the USA? Maybe there really *are* more Democrats (no ‘maybe’ about it) — not 11% more, no way, but perhaps 5% more? Perhaps 7%?
If there really are 5% more D than R then a D+11 poll is “oversampling” not by 11% but by 6%. Still pretty bad.
So let’s say Mitt wins by 5 or 6 at most, not..... 22.
Even 5 or 6 sounds high, but as long as he wins by 1 in the electoral college that’s good enough.
My understanding of the ‘rat oversampling is that that is what their sample actually was. This just means that they didn’t get a good sample, probably because R’s don’t answer pollsters on the phone.