Kudos to Markos and others for pointing out the heavy "skew" introduced into the Gallup numbers by Romney's ridiculously high lead in the South. It raises a strong possibility of a Romney popular victory and an Obama EV victory.
But there is one problem: The states where Romney is likely to lead by very high margins (e.g., Mississippi, Arkansas, Alabama) aren't that big (populous). And there are big southern states where the margins are likely to be much closer. So, if proper weight is given to those states, its hard to imagine a scenario where there is a 22 point margin.
I don't know exactly which states Gallup considers "South", so I did a quick "worst case scenario" calculation using the following states: AL, AZ, AR, FL, GA, KY, LA, MS, NM, NC, OK, SC, TN, TX, UT, VA, WV.
FL, NC and VA are obviously competitive, and they are 3 of the 5 largest states listed above (TX and GA are the other two).
Lets assume a 5% margin for Romney in those 3 states.
New Mexico is lean-D, so assume a tie.
Texas and Arizona seem likely to be single-digit, or near single-digit states (large Hispanic populations), but assume a whopping 15-point margin for Romney.
For the remainder, assume either (a) a 2.5-fold larger margin than 2008 or (b) a 30-point blowout, whichever is smaller. Except for OK and UT, where margins were over 30 in 2008, assume a 40-point margin.
Multiply by the voting population in 2008 and get your worst case scenario: Romney wins the "south" by 15.
I hate to be in the business of "unskewing" polls, but this seems to be a problem. It does call Gallup's geographic weighting into question. For instance, when weighting for geography, is a OK voter the same as a FL or NC voter?
Just a quick Excel calculation. Let me know what you think -- especially if you know something about election polling.
I'll save my worksheet and modify numbers if people give me additional info (e.g., which states are in Gallup's "south").