At one point, I thought it might be good to use this method as a grading tool -- maybe deducting points for the ideological distance between students and the members they're portraying. However, the standard errors generated by only 30 roll call votes are huge, so I don't think it would be proper to base grades off these scores. That said, the ideal points are suggestive. For one thing, the students are doing a pretty good job; the scores correlate at .965. But, of course, that's what you get when all the data are in the extremes, which points to a second inference: my class appears to be more polarized than the actual Congress.
I'm not sure why they're so partisan. I mean, the TAs and I try to instruct them in the importance of partisanship and issue warnings when they vote against their party (or district) too much. But I'm not sure how much of this is us and how much of it is the dynamics of legislative life. I give the parties time to caucus before floor sessions, and they actually use those times to develop strategies for screwing the other party. They're really quite crafty this way. I sometimes worry I'm creating a small-scale Stanford prison experiment -- the students really do internalize their roles well -- except that they're still keeping it civil with each other in committee and on the floor and, as far as I can tell, they don't carry their partisan roles outside the classroom. They're just voting against each other.