One frustrating aspect of political analysis for wannabe math nerds like me is that the data available about who voted for whom is pretty poor. John Judis, for example, in putting together his smart post-election analysis is forced to rely on some very imperfect proxies:
The Democrats also made gains among a critical subgroup of independents--the white working-class voters known as Reagan Democrats. In the Midwest, Democrats won these voters (most clearly identifiable in the polls as voters with "some college") by 50 to 49 percent. . . .
For instance, professionals--best identified in exit polls as voters with postgraduate education--backed Democrats by 58 percent to 41 percent in congressional races.
I'm willing to believe that those are the best available proxies. They're not, however, very good proxies. By this metric, neither I nor scarcely any of my friends count as professionals, even though we're all professionals, and lots of people who aren't even white will count as members of the white working class.