So I was thinking about the really interesting question of whether Obama's deficit among white voters (40-60) was a result of the rural-urban gap. I haven't seen any exit polling that breaks things down that far, so I've resorted to cobbling together an estimate from a mix of exit polling and census data.
If I were to spend more time on this, I could incorporate assumptions about voter turnout and the like. But my basic conclusion is that urban whites are probably slightly Romney-leaning and are, at best, split 50-50. The reasoning is fairly straightforward: urban minorities went for Obama around 80-20, and supply most of the margin of victory. I end up with an estimate of around a 20% margin for Romney in urban areas, which is probably around 5-10% when you take into account the various assumptions I make that would overestimate the % minority vote.
In order to do this I have to make a few assumptions, which means I'm certainly not 100% accurate. I assume that I can use national census data to estimate the % of each minority that voted for obama. Since minority turnout was slightly lower than white turnout, and this may have been even more the case in non-competitive states, I am probably overestimating by a few % the pro-Romney margin. I also assume that african-american, latino, and other minorities had the same pro-Obama gap in both rural and urban areas. If I had assumed that urban minorities were more pro-obama, this would mean a larger Romney margin among whites. Lastly, I assume that the census and CNN have roughly similar definitions of "urban," but I suspect that CNN classifies far more people as "rural" than the census does.
I estimate that Romney's margin of victory among whites was: 32% in rural areas, 25% in suburban areas, 20% in urban areas. Since Romney only won whites by 20%, it's clear that the differences between the census and exit poll data are shifting the results towards Romney. This is looking at the national level though--it seems likely that Obama did better among whites in Ohio, for example, since he won that low-minority state.
How do I get these number? To use urban voters for example:
19% AA * 87% margin of victory = 17% of Obama's margin comes from African American voters. 23%*44%=10% margin from Latino voters. Other minorities->3.5% margin. Basically, Obama probably won urban minorities by around 80-20, and they accounted for 30% of his margin of victory. However, Obama only won cities by 20%. Given that white voters brought Obama's margin down by 10% and are about 50% of city-dwellers, that means about a 20% pro-Romney margin.
One other major source of "error" could be if the exit polls used a different definition of urban. This, like lower minority turnout, would mean that I overestimate the percentage of minority voters and therefore assume a larger Romney margin among whites. There is no way to check this directly, but let's assume that I at least got the minority margin of victory (80-20, or 60% margin) right. And let's say I really think that Obama and Romney tied among whites in urban areas. In that case, whites would compose 66% of urban voters.