Hail to You uses Reuters' America Mosaic polling explorer to check out his theory that the reason Episcopalians voted for Obama more than other Protestants did is because they are concentrated in the Northeast. So he looks at whites' voting by religion for the Northeast and the South.

Mostly, it looks to me like whites in the Northeast went about 15-25 points less for Romney than did whites in the South and that holds for religious subsets. For example, Romney won 29% of the Jews in the Northeast and 46% of the Jews in the South. Romney got 45% of the Episcopalians in the Northeast and 66% of the Episcopalians in the South; 52% of the white Catholics in the Northeast and 72% of the white Catholics in the South. 

In general, the Reuters-Ipsos results are so beautiful in terms of how perfectly they would fit into a multiple regression model of how people vote that I sometimes fear that Reuters-Ipsos is pulling my leg. Maybe they didn't really survey 41,667 voters online. Maybe they just started with a multiple regression model with reasonable weights for race, gender, marriage, religion, region, state, homeownership, education, and so forth, and then just made up the data to fit the model?

I'm just being paranoid. I have zero evidence that this is a hoax. Indeed, a few times I've seen anomalous results from Reuters that represent small sample sizes that wouldn't appear in mocked-up data. For example, there is a reverse gender gap among white working class non-college voters in those Slippery Six upper Midwestern states that Romney lost fairly narrowly.

By the way, for polling wonks only, I figured out a way to get a reading on groups too small to get a readout on the Reuters polls. Reuters’ American Mosaic Polling Explorer is set up to not let you see the results for groups with a sample size of less than 100 respondents, such as Southern Mormons. But, you can figure out the numbers by combining groups below the cutoff with groups a little above the cutoff. For example, in Hail's 12 Southern states, there were 180 Jewish voters (excluding 3rd party voters), who went 45% for Romney. If I select Jewish _and_ Mormon, now I get a sample size of 266 that went 63% for Romney. This suggests 266 Mormons+Jews – 180 Jews = 86 Mormons in the South. Romney carried 63% of the Southern Jewish+Mormon group, so that would suggest he got 78 of the 86 Southern Mormons, or around 90%. (I’m doing the arithmetic in my head, so I might be off by a little.)