Saturday, December 15, 2012

Legislator Performance Scores

In a previous post I discussed methods to measure legislator effectiveness and challenges in doing so.  Imagine people sitting in bars discussing the averages of legislators like we do with pitchers and quarterbacks now.  But as of right now, we can't.  One measure that jumps to mind is sponsorship of passed bills comes to mind, but the seniority of the legislator has an impact there, and the bills might be ineffective or damaging!  Since we don't measure bills and legislator performance, diffusion of credit, and post hoc propter hoc behavior on the part of the electorate (when they're not just identity-voting) makes free ridership an especially bad problem.  It's amazing good bills ever get passed at all.

One thing we can measure is partisanship, although it bears immediate emphasis that this is not a measure of effectiveness.  I'm only using it to illustrate the point that politics is clearer when we measure performance numerically, which of course is exactly what politicians don't want.  To illustrate why partisanship is different from effectiveness:  imagine if an employee is told by his boss that he isn't meeting certain important performance metrics that were important for the company, and he would soon be let go if he didn't shape up.  Imagine further that the indignant employee's response is "yes, but I consistently support efforts that align with the goals of the VP of marketing's internal coalition!  Isn't that enough?"  That would be stupid, but this is exactly what voters are doing when they equate partisanship with effectiveness.

So scoring legislators is a way to make sure we have knowledge about something more than just tribal instinct.  For example, recently these measures have been in the press largely to put a value on political polarization over time that's based on more than just feelings in our little hearts. It's the political scientists who blog at (my new favorite blog) who make these nifty partisan plots.  There are multiple "dimensions" along which partisanship can be measured, as you can see below, but the one I pulled out for this analysis is the left-right axis here, which is basically the traditional American, de facto 2-party system conception of the conservative-liberal spectrum.

We can look at these measurements against the party makeup of state representatives and executives.  First, I looked at the D vs R composition of Senators and Governors as well as presidential electoral votes in 2012.   There are 26 states (a quarter) whose senatorial contingent, governor, and vote for president were entirely one or the other:  13 of each.  (All R: ID, WY, UT, AZ, NE, KS, OK, TX, TN, MS, AL, GA, SC.  All D:  HI, WA, OR, CA, CO, MN, DE, MD, NY, MA, RI, CT, VT.  The independents in the all-D column caucus with the Dems so they were included with them.)

First, do yourself a favor and watch the migration and clustering of partisanship for all Congresses through the 111th in this Quicktime movie.  (Note:  I've used only Senators here but a more thorough analysis would involve all of Congress.)

Then, I used their measurements for the 111th and ranked states by the average of their senators' voting, liberal to conservative (more negative liberal, more positive conservative).  The trend is clear in the table:  more liberal senators mean more likely Democratic Senators, governors, and presidential vote. What doesn't jump out right away is that states with senators more liberal than "neutral" will vote Republican for president, but not so much the other way around. The tendency is even stronger for governors.  I imagine it hasn't escaped notice at GOP HQ that Chris Christie is the Republican governor in the state with the most liberal senatorial voting record, who could benefit most from this effect in a national contest.  Whether he's pure enough to make it through a GOP primary remains to be seen.

State Sens Gov Pres Part Score
VT DD D D -0.5625
IL RD D D -0.5455
RI DD D D -0.5295
MA DD D D -0.51
NJ DD R D -0.498
NY DD D D -0.482
MD DD D D -0.457
DE DD D D -0.456
MI DD R D -0.451
CA DD D D -0.4465
HI DD D D -0.4405
OR DD D D -0.425
NM DD R D -0.413
MN DD D D -0.397
WA DD D D -0.3725
CT DD D D -0.371
WI RD R D -0.367
ND RD R R -0.32
CO DD D D -0.278
WV DD D R -0.2755
VA DD R D -0.249
MT DD D R -0.244
AR RD D R -0.214
OH RD R D -0.206
PA RD R D -0.133
IA RD R D -0.0595
AK RD R R -0.026
NH RD D D 0.0395
MO RD D R 0.045
ME RD R D 0.0495
IN RD R R 0.0615
SD RD R R 0.1
NV RD R D 0.112
NC RD R R 0.1575
LA RD R R 0.1935
NE RR R R 0.2195
UT RR R R 0.352
FL RD R R 0.361
TN RR R R 0.372
MS RR R R 0.396
KS RR R R 0.4295
TX RR R R 0.4765
AL RR R R 0.511
GA RR R R 0.511
AZ RR R R 0.5295
ID RR R R 0.53
KY RR D R 0.594
WY RR R R 0.6295
SC RR R R 0.652
OK RR R R 0.826

Legislative partisanship scores over time give us a number with more spatial and temporal resolution to assign to states.  A continuously measured integer that gets down to Congressional districts sure beats R or D at the state level once every four years.

That it's effective to think of politics this way is strange and troubling.  If people are really voting rationally, they'll look at how the actions of legislators and executives have impacted them materially, rather than just punishing them for not using certain in-group codewords or being in office during an economic downturn that they had nothing to do with and possibly even made better.  But Democrat or Republican are brand names that the party organizers use to paper over the cracks in coalitions, so that people can use them as a proxy for identification as their own tribe. Culturally, Native Americans, West Coast moderates, and Northeast union members have very little in common other than voting Democrat; and big city financiers and evangelicals don't cross paths an awful lot either, except maybe at the polls, voting Republican.  We've become so used to the absurdity of being a country of 315 million viewpoints that we assume can be represented by two parties that we don't think about this very much.

To re-emphasize, the whole reason we should be interested in scoring legislators is to measure their performance - that is, their productivity, not their partisanship.  So far we don't do that, although it's true that there are "report cards" issued by public interest groups scoring legislators (example here.)  But even there, the report cards only measure what the legislators and their bills were trying to accomplish, not what they actually accomplished; what we really need to fix the broken loop and help democracies is a way to measure what they actually did, relative to expectations, so that governments are accountable.  I imagine that lobbying organizations (or their smarter industry clients) have individual tools for this, but they're not available to individual voters.

No comments: