When people were discussing the CA-36 special election, some of them called the Democratic candidate, Los Angeles City Councilwoman Janice Hahn, a moderate or even conservative Democrat. I usually replied by referencing this statement from Gene Maddaus on Calbuzz:
1. Janice Hahn is no moderate. Along with Jose Huizar and Richard Alarcon, she is one of the three most liberal members of the (quite liberal) LA City Council.
Ok, but that's just someone saying something. Is there a way to check that it's true?
To find out, I decided to use the DW-Nominate method developed by Poole and Rosenthal (the program also uses the PSCL package developed by Jackman). If you're not familiar with DW-Nominate, the basic idea is that you input a bunch of roll call votes for a legislature, and you get back a coordinate for each legislator. For example, if that's how the votes split up, an economic and social liberal might be in the top right, with an economic liberal and social conservative in the bottom right, and an economic and social conservative in the bottom left. Of course, liberalism and conservativism are about more than voting records--how a Councilmember works as a local power broker, their public statements, and so on, are all very important too. But let's just see what we can get from voting records.
The Los Angeles City Clerk maintains a pretty good database of city council votes. I decided to use all of the non-unanimous roll call votes covering the period when the City Council consisted of Alarcon, Cardenas, Garcetti, Hahn, Huizar, Koretz, Krekorian, LaBonge, Parks, Perry, Reyes, Rosendahl, Smith, Wesson, and Zine. That was 256 roll call votes from January of 2010 to June or July of this year, when Englander was elected to replace Smith (Smith and Zine are the two Republicans).
I think the easiest to understand is with two dimensions (I lost this version of my worksheet, but as far as accuracy goes, it got over 90% of the yea votes and something like 64% of the nay votes). Here is the ranking for the first dimension:
1. Koretz 0.93719816
2. Rosendahl 0.81596053
3. Hahn 0.78070289
4. Alarcon 0.76844990
5. Huizar 0.68893397
6. Krekorian 0.68403184
7. Wesson 0.41718554
8. Reyes 0.34818116
9. Garcetti 0.28581905
10. LaBonge 0.24602304
11. Cardenas -0.05470585
12. Zine -0.17166252
13. Parks -0.78856319
14. Perry -0.78784156
15. Smith -0.85530823
and here is the ranking for the second dimension:
1. Cardenas 0.8761148
2. Alarcon 0.6399099
3. Hahn 0.5396467
4. Parks 0.4402239
5. Huizar 0.3588524
6. Perry 0.3016354
7. Wesson 0.2420398
8. LaBonge 0.2045514
9. Reyes -0.1883332
10. Zine -0.2451584
11. Koretz -0.3487975
12. Garcetti -0.4932601
13. Smith -0.5181195
14. Rosendahl -0.5781076
15. Krekorian -0.7294521
What this means is that there are some issues where people like Alarcon and Rosendahl vote against people like Parks and Smith --and there are other issues where people like Alarcon and Parks vote against people like Rosendahl and Smith.
In fact, let's break it up that way. I'll call a "liberal" someone who is among the first 8 in the corresponding dimension, and the others are "conservative". Keep in mind that this is just mathematical, and I haven't yet discussed what the dimensions actually mean, so these are just labels for now. Councilmembers in each category are listed alphabetically.
Liberal/Liberals: Alarcon, Hahn, Huizar, Wesson.
Liberal/Conservatives: Koretz, Krekorian, Reyes, Rosendahl.
Conservative/Liberals: Cardenas, LaBonge, Parks, Perry.
Conservative/Conservatives: Garcetti, Smith, Zine.
The two Republicans on the Council at the time--Zine and Smith--end up in the Conservative/Conservative category, so that is a good sign, analytically speaking. (Oddly enough, so does the well-liked Garcetti, perhaps due to his position as Council President.)
LaBonge, Reyes, Wesson were both close to the center by both estimates, and LaBonge and Reyes would have changed categories if I used top 7 for "liberal" instead of top 8 (and, roughly speaking, Wesson, who was described as a "charming seat-warmer" in this entertaining and acerbic look at the Council, which you should definitely read as a companion piece to this diary, seemed to vote "yes" on almost everything).
So we can see that Calbuzz was right, provisionally: Alarcon, Hahn, and Huizar were the most "liberal" members, taking both dimensions into account. Which makes me wonder if they've seen an analysis like this--or, more interestingly, if the analysis just reproduces, in a purely quantitative way, what reporters like Calbuzz know qualitatively.
Of course, that still leaves open the question of what, exactly, these dimensions mean. Maybe they don't mean liberal/conservative in any familiar sense! But I think I will look into that in a follow-up diary, as I am curious to see what people think they might mean, I am getting tired, and this is probably long enough for now. I am not sure how often DW-Nominate is used for nonpartisan municipal roll calls, but it is pretty interesting. I'd be happy to look into it with other reasonably small city councils. I don't think even my hyperfocus is enough to get me through the New York City Council with its 50+ members.
(Note: This is using all contested roll call votes--when I set it to try to exclude roll call votes that were 12% lopsided or more, the order changed a little in both dimensions, but not much. Notably, Rosendahl and Smith moved closer to the center, leaving Hahn as the second most liberal and Parks as the most conservative in the first dimension. That version got 92.7% of the yea votes correct and 73.1% of the nay votes correct. Of course it eliminated 92 of the 256 votes--Los Angeles City Council votes are seldom contested very much.)
(Citations: Wnominate program is from "Scaling Roll Call Votes with wnominate in R", Keith Poole and Jeffrey Lewis and James Lo and Royce
Carroll, Journal of Statistical Software, http://www.jstatsoft.org/.... PSCL is from Simon Jackman (2011). pscl: Classes and Methods for R Developed in the Political Science Computational Laboratory, Stanford University. Department of Political Science, Stanford University. Stanford, California. R package version 1.03.10. URL http://pscl.stanford.edu/)
Tue Jul 19, 2011 at 1:18 PM PT: On the advice of davidshor, I ran the votes through a similar program, Ideal, in the PSCL package. This yielded some other first and second dimension rankings, one of which was similar to the diary's, but some of which are very different--see the comments. I'll leave this up, but there's much more to figure out.
Thu Jul 21, 2011 at 3:54 PM PT: davidshor pointed out in comments that some of the "different" dimensions might be the same basic plotting, just rotated. Having run ideal on the roll call data again, with 150,000 iterations, I got a d=1 plotting of: Koretz, Rosendahl, Alarcon, Hahn, Krekorian, Huizar, Wesson, Garcetti/Reyes, LaBonge, Zine/Cardenas, Perry, Parks, Smith. Using this dimension to orient the d=2 map gives more or less the dimensions in the original post. The main change is that, if the classification into four groups was repeated, Garcetti is now in with Rosendahl, Koretz, and Krekorian; while Reyes is in with Zine and Smith. Which makes more sense--and perhaps explains Meteor Blades' dissatisfaction.