(Phenomenal work. Promoted from the diaries with an edit to the title by DavidNYC.)
Lots of people in the blogosphere want to primary Bad Dems. Used wisely, that's a good thing. It can wake up a Blue Dog on cruise control in a dark blue district (Jane Harman); sometimes, it can even lead to a victory and a Better Dem (Donna Edwards).
Unfortunately, I've never seen any sort of metric that helps us decide who's a worthwhile primary target, i.e. who's way out of whack with his or her district, versus someone who's an odious Blue Dog but who's the best we can manage in a red district and someone to be welcomed into the big tent (if sometimes secretly grumbled about). So, here's my stab at it.
It's based on this proposition: the representative in the most liberal district should have the most liberal voting record. The representative in the most conservative district should have the most conservative voting record. The representative in the 217th most liberal district should have the 217th most liberal voting record. And so on. Of course, in reality, it doesn't always work like that. But finding the representatives whose voting records are severely mismatched with their districts' lean helps us find the Bad Dems who need some prodding. And as a bonus, it also helps us find the Republicans who should theoretically be extra vulnerable in a general election: the wingnuts hiding in moderate districts.
To that end, I've developed the PVI-Voting Pattern Index. It simply rates every district from most Democratic to most Republican in its presidential preference, and rates every representative from most liberal to most conservative, and looks at the differentials.
