Lots of people have probably heard that most polls don't call cell phones. I've always assumed that this doesn't bias the outcomes very much. (Yes, more cell-only users may favor Obama, but there aren't many of them.) Over the weekend, Nate Silver had a post suggesting that it actually does have a significant effect. Nate compared the findings of polls that call cell phones to those that don't, and found that the latter tend to depress barack Obama's share of the vote by 2.8 percentage points compared with the former. That's a huge difference.
I emailed Nate and asked him how this effects the overall average of polls. (Since the average includes cell-users as well as landline-only-users.) He replied that the effect is around two percentage points. If Nate's right, and the polling average is underestimating Obama by two percentage points because it omits cell phone-only users, then that finding deserves more attention. This isn't to say that Obama has it in the bag, or that there aren't other polling biases that could swing the other way. But few events in the election are big enough to swing two percent of the electorate.