The problem with a sampling universe in this is far more complex than raw numbers ~ when you are down to a 9% response rate from the public at large you have to give up all those additional factors ~ else you'd have to call hundreds of thousands of phones, and when that happens the influence of respondents with multiple phone lines grows ~ which it may well have in Romney's case ~ (the rich have more phone numbers than the poor and middle class ~ and even then we found ourselves with 5 active numbers!).
The pollsters ~ who are not as stupid as we imagine ~ stepped back and began running what can only be called AUGURIES!
I think some of my original Freeper posts on AUGURIES are still accessible through Google.com by asking for FREEREPUBLIC MUAWIYAH AUGURIES.
Except for pre-selected targets I"m afraid random sampling polling may be a dying art. There are other methods that need to be brought to play.
Sampling strategies will be altered once it is proven that current strategies are faulty. However, current strategies are for them most part accurate as polling prior to the 2012 election was within the margin of error.