No need to lecture me on cell-phone users, it's obviously a laundry list of things that can go wrong with a poll.
That's actually only one theory behind the
Bradley Effect. If you count all the people who are really decided for McCain and instead tell the pollster that they are undecided or voting Obama, that
might equal 1% states in Pennsylvania or possibly even more. I think it's pretty ridiculous, biased, and arbitrary to give a flat 0.5-1% automatically for McCain off of the official poll results though.
Because, even if you are right and this many people lie, combined with racist undecideds, resulting let's say 1-2% breaking for McCain, you are still assuming:
a) The poll is 100% accurate statistically
b) The poll has 0% problems methodologically, including oversampling and undersampling
c) No one else is lying about not supporting Obama when they really do
d) Both candidates have completely equal ground game
I don't think any of those things are true. I know in Pennsylvania, ground games tend to neutralize one another, but an extra 1% this way or the other wouldn't be unheard of. Not to mention polling accuracy in many other respects.
To isolate out just the Bradley effect in order to give a boost to your candidate is misleading.
Why do you assume Obama is more like Steele and Patrick and less like Ford? It seems Obama has higher amounts of information-saturation, meaning less people feel the need to lie, combined with the fact that Steele and Patrick were both Republicans, further differentiating themselves...