« Statistical challenges in estimating small effects | Main | Yes, it's rational to vote in presidential elections »
November 6, 2007
Mapping State and Congressional Ideology on a Cross-Institutional Common Space: An Update
An updated version of my (Boris Shor, Harris School, University of Chicago) paper with Nolan McCarty (Princeton) and Christopher Berry (Harris School, University of Chicago) on state legislative ideology is available here. A prior version of the paper was featured in an April 2007 post on this blog.
Basically, the idea is to try to understand the ideology of state legislators along some common scale. We have good estimation techniques (like NOMINATE and item response models) that exploit roll call votes to line people up along ideological dimensions.
There have been some attempts to apply this at the state level, and researchers have come up with ideological scores for legislators within individual states. But a major problems remains: how can we make sure scores are comparable across states? We know that legislative agendas differ. We also know that the meaning of Democrats and Republicans is different across the states, too. Illinois Republicans probably aren't the same thing as Florida Republicans. But while we may suspect that Democrats and Republicans in the various states are more or less liberal or conservative with respect to each other, we have never been able to find out if this was actually true. Nor have we been able to tell if the divisions within states are comparable in size across states.
We address this well-known problem with a trick: it turns out that quite a few former state legislators have gone on to serve in Congress. So we use the well-understood congressional ideology of these people to rescale the ideology of all state legislators who didn't go on to Congress. So, in essence, we have rescaled all the ideological placements of thousands of state legislators over a decade to be on the "Congressional scale." We use this approach for California, Colorado, Florida, Illinois, Michigan, New York, Ohio, Pennsylvania, and Texas. We start with these large states because their congressional delegations are larger, which means there is more chance for former state legislators to become our institutional "bridges."
Therefore, we can now make valid comparisons of various measures of state legislative ideology across across states. Read the paper for the full details, but here's a little preview of the result.
Below is a boxplot of the two parties in each state, but pooled over a decade and across upper and lower chambers. So this gives you a basic flavor of the ideological orientation of these states, but doesn't show you all the juicy details of what is going on in each chamber over time (see the paper).
Scores are on the vertical axis. Positive numbers indicate conservatism and negative numbers liberalism, and 0 is "moderate." The dark lines are the party medians.
What is interesting is how different the states are from each other (and Congress). California is highly polarized (parties are far apart and there's no overlap), while New York and Pennsylvania's are far less so. State parties often move in tandem in ideological directions: Illinois Democrats and Republicans are more liberal than, say, Michigan Democrats and Republicans.
This paper is still in development, so we welcome your comments here or by email.
PS: Corrections in the text of the post made, thanks to Lee Sigelman (GW).
Posted by Boris at November 6, 2007 12:01 PM
Trackback Pings
TrackBack URL for this entry:
http://www.stat.columbia.edu/~cook/movabletype/mt-tb.cgi/1268
Comments
Posted by: Andrew ![[TypeKey Profile Page]](http://www.stat.columbia.edu/~cook/movabletype/mlm/nav-commenters.gif)
at November 6, 2007 6:09 PM.
Interesting. This is the same technique used in the original works to bridge the senate/house-of-representative data, right?
The section discussing the possible sources of polarization seems like a separate paper.
Two marginally related questions. Is there a convenient place, yet, for looking up a legislator's score? Is there a good source of examples, i.e. given law X those with scores below Y will likely vote against.
Posted by: Ben Hyde at November 7, 2007 8:15 AM.
Andy: what do you think would be a better order? Perhaps overall chamber medians?
Ben: yes, that's right, this is the same bridging technique to create common space scores for Senate and House. Except there are a lot more Representatives who go on to the Senate!
After we publish the paper, we'll put up individual scores, right now we're showing distributions.
Posted by: Boris S. at November 7, 2007 10:46 AM.