The most recent (October 16) breakdown of the LV spread, the same one that shows the South as producing all of Romney's national LV spread, also has data for non-hispanic white & non-white spreads for Romney and Obama respectively. A simple equation for the white share of the vote, x, shows that the LV model has white(non-Hispanic) turnout at 80%!
0.22x = (100-x)* 0.68 + 4
Solving for x, x=80.
No fricking way (2008 was 74%) is that possible judging by turnout in OH and NC early voting. In point of fact, if white turnout falls to 75.5%, Romney's LV advantage (4 on Gallup, October 16) disappears totally!
So much for agonizing over Gallup's numbers!
Note on Gallup's regional breakdown below the fold.
As for the regional breakdown:
Here is what should guide the regional weighting of the RV's, from each state's total House electoral votes (Total state EV minus 2):
East 93
MidW 94
(incl. OH,MI,MO,KS,NE,SD,ND)
West 99
South 152
(incl. VA,WV,FL,TX,OK)
Seems this is how Gallup regionated, the weighting in the LVs seems approximately right to produce the LV spread of 4 for Romney.