Monday September 29, 2008
Here is a quick look at the way the bailout bill failed on the floor of the House of Representatives today. It failed 205-228.
(although at the moment my read of Jeff Lewis’ site has it 204-227; that will get updated at some point). (Fixed by Jeff; thanks).
If we project House members’ voting histories in the entire 110th House onto one dimension, how well does the recovered dimension predict the vote today? Answer: ok, but not great, since opposition came from both parties, and the parties are well separated on a one dimensional continuum these days.
Where in the ideological spectrum did the bill run into problems? Answer: all over, but slightly to the right of center.
See the graphs, below (click on the thumbnails). The first graph shows a probit model fit to the Yeas and Nays, with the recovered one dimensional left-right position of each legislator as the predictor; the fitted curve slopes down (movement to the right generates less support), but the fit isn’t terrific (area under the ROC curve is .67, treating the ideal points as known).
I wonder what this looks like in 2d, or with other predictors introduced to the model. Update: Nate Silver says marginality helped scare legislators away from voting for the bill.
Update: The 1st graph has estimated “ideology” of the House members on the horizontal axis (the “ideal points” of the legislators); left is well, “left”, and right is conservative, although with so much party-line voting there is some debate as to how we should actually interpret these estimates. Vertical axis is probability of a Yea vote on the bailout, with the curve coming from a probit model. Actual “Yea” votes are shown at 1.0 on the vertical axis; actual “Nays” at zero. Color indicates party.
2nd graph: simply plots the ideal points of legislators conditional on party and how they voted.
Sunday September 28, 2008
What twaddle. George Will’s performances on the last couple of Sunday mornings (This Week) have seen his stock store here in this corner of Palo Alto. But then we came across this nonsense:
…surely the quality of the electoral turnout declines when the quantity is increased by “convenience voting.”
A word describes most of the people who will vote only if a ballot is shoved through their mail slot: “slothful.”
So maybe we should insist people pass a civics quiz before going into the polling place? I guess we shouldn’t be surprised that conservatives are the only people in America entertaining the idea that voting is too convenient.
A word describes people voting by mail, the same word that describes those who vote any other way: “voters”. Democracy is more about process, less so about outcomes, so we don’t get to criticize citizens who exercise their democratic rights through the mail.
Saturday September 27, 2008
A recent comment reads:
Nice font you are using in your CV, syllabi, etc. — What is it called?
How nice of you to notice. Its FFMeta. I’m almost over it: it is a little corporate, used by Bank of America, the Queensland Government, the ANZ Bank.
There are some nice “hint of a serif” humanist typefaces out there I could look at too, but life is short. I’ve got Meta nicely integrated with my LaTeX setup, and unlikely to ever have the time to go through that again. Oh well.
I’m even wondering about something oldish like Baskerville or Didot. Sometimes technical papers — or topical, timely papers — can look all that more authoritative (!) when set in something older looking.
Tuesday September 23, 2008
If it is Week One at Stanford, then it means coursework is running like treacle. Why can’t Academic Computing do a better job of anticipating this, preparing for it, testing high load conditions on their system. Ugh. Hardware is so cheap, load-balancing this shouldnt be hard, unless of course it is a poor database design (gee, could it be?!?).
I’m hosting the content for PS 13N, my Freshman seminar on the 2008 election over on my own web site for the time being, under classes.
Update (Wednesday): the system seems much snappier now.
Monday September 22, 2008
My former student Sunshine Hillygus has been running some polls for a Yahoo/AP joint venture this summer, with Knowledge Networks handling the fieldwork.
A recent wave of the Yahoo/AP panel saw my colleague Paul Sniderman (
and perhaps also Jon Krosnick and also possibly with support from IRiSS) field some items on race in the 2008 election.
A report appears here. The findings mirror some of what we were seeing much earlier in the year in analyzing voting in the Democratic primary.
A useful follow-up from Larry Bartels.
Gary Langer from the ABC Polling Unit says that including cell-phone only respondents produces a negligible impact on results
Across 100 response categories the biggest difference was 2 points, which occurred in three cases. The rounded difference was 0 in 56 of the comparisons, and 1 point in 41.
This is smalll where the “margins of error” (classical 95% confidence intervals assuming SRS, I suppose) are +/- 3pp for RVs.
There are two reasons for the negligible effect. One is that cell-only respondents are not different enough from the landline population – and at 15 percent, there are not enough of them – for their inclusion to make much difference in overall estimates. The other is that customary weighting to Census norms helps account for their absence.
Nonetheless, Langer says that including cell-phone only respondents should be considered experimental.
Its good to see some research on this topic, since it is often talked about as a problem with phone-based surveys research these days. I wonder where they are getting lists of cell phone numbers, or indeed, if they know they are cell phones before they call, etc etc.
Update: Mark Blumenthal at Pollster.com reports on some work on this by Scott Keeter et al at Pew, which seems to suggest that there are some differences across phone type.
Friday September 19, 2008
I’m offering a Freshman seminar on the 2008 election this Fall, with our first meeting at 9.30am on Monday morning. Alas, or thankfully (depending on your perspective), this class has a capped enrollment. I’m unable to review applications until after 5pm today, Friday September 19.
So, to any applicants: please keep an eye on the Freshman seminar website over the weekend re the status of your application.
Update (Saturday): I have 108 applications. Wow. I’m so sorry to the 90+ applicants who I won’t be able to take this quarter. Thank you for your interest in the class, in the election, in American politics, in political science… Clearly we have some demand going unmet here.
Update (Sunday): Of the 108 applications, 81 are from Freshman who said the class was their 1st preference. Its a very good bet that my 16 admits will come from that subset of 81. Again, my apologies to other students.
Update (Sunday night, 8.40pm): So I made my selections and entered them into the FSP computer system. Students should be able to see the results now, I believe. The process was extraordinarily painful in that there were so many wonderful looking students, with incredibly diverse backgrounds and experiences (if not political orientations!), and so many who are already quite politically active in different ways. The sense that this is an important election for these new adults is crystal clear; so much for the myth that Stanford undergrads are apolitical.
It really is so bittersweet. On the one hand, I’ve never been more sure that Stanford is doing a fabulous job with admissions (!). On the other hand, I had to somehow pick 16 Frosh to be in this seminar, and resorted to some randomization given the impossibility of selecting among so many outstanding prospective students. Good luck to us all.
Some wisdom on this from Jason Santa Maria (with a lovely looking blog, btw) in turn referencing Mike Davidson:
When there are a few hundred people sitting around at a conference, chances are good that they know where they are and what they’re attending. Usually, the most vital piece of information they don’t know is everyone else’s name.
Too right. My own design, from PolMeth 04, rendered in Zurich-BT Ultra-Black Extended:
2000 turned us into FL experts; 2004 turned us into OH experts. And now, CO?
I blogged on this last week. Stuart Rothenberg agrees, as does Nate Silver, with considerably more detail than I offered.
Wednesday September 17, 2008
My contributor’s copy arrived today. I contributed an essay on “Measurement”; Andrew Martin did the Bayes essay. There are many other contributors from political methodology, including Keith Poole on the legacy of psychometrics on political methodology.
It was very nice of Jan Box-Steffensmeier, Henry Brady and David Collier to ask me to be part of what is a very distinguished line-up of contributors.