Friday, October 23, 2009

Agenda control and state legislative polarization

The more I think about the Shor/McCarty graph showing legislative partisanship in the states, the more concerns I have. Don't get me wrong -- it's an amazing graph with a ton of important data behind it. But I'm increasingly concerned about the hidden role of agenda control.

The graph below is a density plot of the percent of the chamber voting yes on all the bills that came before it in the 1999-2000 legislative session. It charts this distribution for both the California and Florida assemblies. (The data come from the Representation in America's Legislatures project.)What do we see here? In Florida, the distribution is bimodal. Most pieces of legislation either get around 40% of the vote or around 80% of the vote. Why would so many bills get 40% of the chamber voting yes? They are minority party bills. The minority party usually loses, but they can still get their bills to a floor vote.

Contrast that with the distribution for California. Again, the distribution is bimodal, but the peaks are around 60% of the chamber (a narrow win) and around 90% (near unanimity). Almost no bills get below 50% of the chamber voting yes. In other words, the minority party's bills almost never make it to the chamber floor. The majority party makes sure that they die in committee.

This can have a substantial effect on the appearance of party polarization. In California, bills that might otherwise give the chamber more of an appearance of bipartisanship are being killed by the majority party before they can reach a floor vote. This may account for California having the most divergent parties in the Shor/McCarty graph. Or maybe it really is due to the legislators being so ideologically dissimilar. It's hard to tell, and I don't know of any multi-year measures of state-level agenda control that would allow us to factor this effect out.

One other issue. As some of the commenters at 538 noted, it's kind of weird to see California's GOP further to the right than any of the southern states' Republican parties. Now this is certainly possible. The Shor/McCarty data compresses everything onto one dimension. I'm guessing that the southern Republicans would look more conservative on cultural measures than the CA GOP, although the Californians are probably more conservative on economic measures. But I'm not sure.

Update: Boris Shor responds.

Later update: Boris is probably right that agenda control isn't biasing these results all that much. As he notes, the U.S. House and Senate have very different agenda control rules, yet members show remarkably consistency in their ideal points when they move from one chamber to the other. However, I don't know whether the differences between the House and the Senate are on par with the differences across state legislatures. The differences between Florida and California strike me as pretty huge.


Anonymous said...

You can get some answers by computing nominate scores (or other preference estimates) only from npat responses. This should get you pretty darn close to the electorally-induced preferences of legislators.

If you make those scores for 99/00 and then take the R mean minus the D mean (the data I have lying around right now) then the CA S is the most polarized chamber in the US, followed by the OH S and the CA H. Upper chambers often have pretty terrible response rates; if we look at lower chambers the order for 99/00 is CA, AK, CO, TX, WA, AZ.

If you just look at the conservatism of Republicans, then looking again only at lower chambers the order from most conservative down is CA, GA, OK, TX, CO, MO, NV, AK.

(Jim B again)

Anonymous said...

I'm guessing that the southern Republicans would look more conservative on cultural measures than the CA GOP, although the Californians are probably more conservative on economic measures.

Okay, so we made 2-D nominate scores based only on npat responses, for everyone who took the 1998 npats including losers. As it happens, the libertarians cluster in the northeast "corner" of the unit hypersphere. (Yes, spheres have corners. What?) If you accept that that "corner" defines the stereotypical libertarian position, then the first dimension looks a lot like econ or "standard" L/R and the vertical a lot like social L/R.

Anyway, if you buy this, then among lower chambers the most "socially" conservative Democratic parties are in LA, NM, MS, OK, KY, AL, SC, NC. Which seems pretty reasonable. The most "socially" conservative Republican parties are then MD, VA, MS, HI, NY, OK, CA, GA, NC. Which seems a little bit weird on the face of it.

Again, all for 99/00, not for the much broader set of data that Boris has.

Anonymous said...

sorry, that last was Jim again.

bshor said...

Seth, I responded to your questions in a new blog post.