Mindless Philosopher

Sometimes the universe freaks me out a bit. Monday I guest taught "Ms. Stollwerk's" world history class in the South Bronx. I volunteered to teach the scientific revolution since I took an excellent class (from the inimitable Prof. Manning) on that period of history at MIT. In the course of one of my lectures to these high schoolers, we talked about gravity and why objects fall toward the Earth (i.e., why doesn't gravity pull us upwards?). And I explained that everything has a gravitational pull, but that small objects, such as this marker or that sticky-note pad, have tiny, tiny gravitational effects that are way too small too feel. Only massive objects, such as the Earth, tug at us with enough force that they affect our lives. And, lamely attempting to connect to the students, I followed up this statement with "I'm sure there's a Yo Mama joke in there somewhere..."

Yes, okay, I'm at least a year too late with the joke. And yes, okay, no one laughed. But, I get home from NYC and check my RSS feeds and what is the XKCD comic for the day?



Freaky.

I think it's ridiculous that a bare majority of voters can amend a state constitution, but that's not the question at issue today in California. Instead, Prop 8 opponents (i.e., the side of justice and equality) have filed suit saying that Prop 8 is a "revision" to the constituion rather than an "amendment." And since I figure you all are just as ignorant as I am about the distinction, here's a good primer.

<b>Update</b>: After mulling this over for a day, I'm against this lawsuit. I think the appropriate action is to put the issue on the ballot in 2010 (either in the primary or general, whichever one gay rights activists think will be better). And in that case, we would want to argue that striking down Prop 8 would be an "amendment" and not a "revision" (so that we wouldn't have to garner support from 2/3rds of the legislature). Filing this suit undercuts that claim and also undermines the credible of the movement with the voters. Convincing 52% of Californian's to support gay marriage in 2010 will be a lot easier if we don't attempt to throw out their votes with a lawsuit today.

For the better part of a year I worked for Google but couldn't tell people the details of my work because of an NDA. Over the past few months the details of the project have emerged and this weekend the true promise of the project came to fruition. So, let me introduce you to the Voting Information project.

The goal of the Voting Information Project (VIP) is to free election data from the confines of state and county databases. This pursuit fits nicely with Google's mission statement of "organizing the world's information" and the project is a partnership of the search company, Pew, and JEHT. We started in early 2007 with the long-term goal of creating an open format that states could use to publish their election geographic and ballot information, and that organizations like the League of Women Voters could layer candidate questionnaire responses over.

For this election, we focused on election geography, specifically informing citizens where their polling location is. Past attempts at a nationwide polling place locater had failed because not all states or counties were willing to share information. However, with the advent of nationwide voter files, one step of the process was complete: every registered voter in the U.S. could be matched to a precinct. Google devoted many resources this summer filling in the second piece of the puzzle: matching precincts to polling location. Now, every registered voter should be able to find his or her poll location here. (The Obama campaign also used this strategy.)

But, what about new registrants (or even same day registrants)? If they live at an address without another registered voter, Google's strategy will fail. That's where VIP comes in. Participating states and counties, distribute their street segment infomation (in a common format), which matches all addresses in the state to a precinct and then a polling location. These street segments are exactly how registrars themselves determine the precinct of new registrants, so they are very reliable.

I'm happy to announce that this past weekend, Google combined these two strategies at maps.google.com/vote. Test it out! One of my Princeton colleagues hadn't realized that her polling location had moved until she tried the application, You can even put the app on your own website as a gadget. Also, the website got some press play from the NYTimes and Te The Today Show <video>. So that's what I've been up to.

Also, go vote!

In an odd convergence of the universe, my work received three shout outs today -- all on completely unrelated topics. So I figure since this site will get a slight boost in traffic today, I should use the attention to better the world.  I have two requests:

1) Give to your favorite charity today, please. FYI, this month is breast cancer awareness month. Also Charity Navigator is a great resource.

2) In case pollsters are reading this blog, please help settle the "Bradley Effect" debate by releasing individual-level data aggregated by the race of the interviewer.  There have been several articles in the past week about the so-called "Bradley Effect" and whether pollsters are overstating Obama's lead. But I have yet to see an in-depth analysis of current, individual-level data. With this data we'd be able to see whether whites and blacks answer the presidential preference question differently by their race. Yes, as Prof. Krosnick rightly points out, we wouldn't immediately know who is lying to whom (i.e., are blacks lying to blacks more than whites are lying to blacks?), but I think this type of data could be useful in several ways:

• Even without knowing which way the error went, we could put bounds on the problem under somewhat innocuous assumptions (e.g., there is not a racial stigma when whites respond to a white interviewer).
• Comparing these type of data from the primary to the actual results of the primary (and possibly exit polls) might shed some light into which way the errors occur.
• Comparing live interviews to robo-interviews by Rasmussen and SurveyUSA might also lend insight into which way the errors occur.

So, experts! Please help stop the rampant speculation.  Let's get some hard data. (Also, I would be remiss to talk about the Bradley Effect without a shout-out to my friend Dan Hopkins, who wrote a great paper on the Effect. His research indicates that we don't have much to worry about.)

Thanks for stopping by.

As a lifelong Orioles fan, I've always loathed the New York "Evil Empire" Yankees. The only time I would root for them is when their winning would help the Orioles in a penant race (remember, back in the 1990s, the Orioles were in penant races). In essence, I was rooting for the other team to lose.

But, today, with Mike Mussina on the mound for the Yanks, I'm rooting for them to win and for "Moose" to get his 20th win.  Mussina has had 19 wins in a season twice, and 18 wins thrice, but has never crossed the 20-win plateau. The last time he was a win a way from 20, he was pitching for the Orioles (where he started his career), but our bullpen blew it for him (on two separate occassions!). Even though Mussina is in his 8th year for the Yankees, I've never really thought of him as anything other than an Oriole.

So today, I hope the Bronx boys can support Moose the way the Birds never did. Go Mussina! Go Yanks!

(Cross-posted at the CSDP Election Blog. Extra note below.)

Every morning one of my friends IMs me to say either, "We're winning!!!!" or "We're losing :( :(." He bases these conclusions on websites such as electoral-vote.com, which aggregate polls and produce a point estimate of the presidential electoral vote. While I'm sure my friend enjoys following the ups and downs of tracking polls (as sports fans enjoy watching games instead of just reading box scores), his ultimate question is who will win the election, not who is currently ahead.

But how well do current polls predict the outcome of the election? Professors Andrew Gelman and Gary King addressed this question in their 1993 work and found that polls are very poor predictors of the final vote. They even went so far as to call voters' responses to pollsters "not "rational." Wild swings in polling returns during conventions or after candidate gaffes often do not translate into long-term effects.

A recent survey by the University of New Hampshire raised the possibility that pollsters, not citizens, are to blame for these poll fluctuations. UNH asked half of respondents the standard preference question:

<blockquote>"Suppose the 2008 presidential election was being held today and the candidates were John McCain and Sarah Palin, the Republicans and Barack Obama and Joe Biden, the Democrats. Who would you vote for?"</blockquote>

These voters slightly supported Obama (46% to 45%), with 8% undecided (and 1% not responding). The other half of respondents were asked a seemingly similar question:

<blockquote>"Thinking about the presidential election in November, would you vote for: Republicans John McCain and Sarah Palin, Democrats Barack Obama and Joe Biden, someone else, or haven't you decided yet?"</blockquote>

While the vote margin for these respondents was approximately the same (McCain +1), the percent of undecideds more than doubled, to 20%. This finding underscores the facts that citizens realize they might change their mind before the election, and that polls are just snapshots in time, limited in their ability to predict outcomes.

The mantra that "polls are just a snapshot" has been repeated often. And often, polls are used appropriately. For example, snapshot polling is helpful in determining who is benefiting from current events, or the effectiveness of a shift in campaign messaging (examples here, here, and here).

On the other hand, websites (some examples here) have no business tallying the electoral vote before the election. Who has "won" the September 25th electoral vote has no bearing on public policy, and provides noisier estimates of candidate momentum than so the current snapshots of the popular vote.

One of the smartest applications of polling (from a campaign junkie's perspective) is to analyze which states are "pivotal" (i.e., the closest state that tips the election--FL in 2000 and OH in 2004). If you have followed Nate Silver's analyses on 538, you may have noticed how Obama's win percentage (top left of the homepage) fluctuated during the conventions, while the list of pivotal states (middle right) barely moved. The analysis of pivotal states has two great features: (1) it provides campaigns with actionable intelligence about where to direct their resources, and (2) it is much less vulnerable to the minute-to-minute fluctuations of the news cycle.

So the next your preferred candidate is behind by more than the percent of undecideds, instead of freaking out, remember that a large chunk of the electorate is still persuadable. And then channel any remaining nervous energy into volunteering in the nearest battleground state.

[Addendum on this site only, since I was over CSDP's word limit: Yes, Gelman and King address the issue of question wording (page 426). They verify that question wording does not affect the percent of the two-party vote as measured by polls (and I agree that there is little effect). But, I think that question wording and the level of undecideds goes directly to the heart of their claim that voters are not behaving rationally. A voter bouncing back and forth between preferring one candidate and being undecided is more reasonable than the picture Gelman and King paint of the electorate swinging wildly between the two candidate.]

With Obama's VP announcement imminent via text message, reporters are taking a renewed interest in the study that Allison Dale and I completed last cycle. We found a 3 percentage point intent-to-treat effect for mobilization text messages (with a 4 percentage point treatment-on-treated). For those of you not familiar with the field experiment lingo, those results mean that campaigns who use text messaging to get-out-the-(newly-registered)-vote can expect a boost in turnout of 3 percentage points. On the individual side, if you receive a text message, your probability of voting increases by 4 percentage points. Those numbers are different because not every text message reaches its intended target.

Garret Graff started off the (small, academic version of a) media feeding frenzy by alluding to our study in his NYT Op-Ed piece last week. Then, our study landed in the news section of the Times, with a slightly innacurate article by Brian Stelter. The quotes from Allison are accurate, but:

• 4.2 prectage <b>points</b>, not 4.2 percent. Those two phrases mean different things.
• The study had over 8,000 subjects, about 4,000 of whom received text messages.
• "The Obama campaign may be running the biggest text messaging experiment" is an abuse of the word "experiment" unless Obama has a control group that no one knows about (unlikely).
  

Then came the UPI article and, like a game of telephone, the errors were compounded. Note the completely innaccurate use of "millions of cell phone numbers" and incorrect attribution of the "experiment" line to Allison. The article has thankfully been corrected (without a correction notice which is odd). The AP story was much more careful in its reporting.

I'll try to update this blog post more often as futher articles are published. After the texts go out and the convention starts, this story will go away fairly quickly. But this was/is a fun week of near-fame :)

Update: Two more stories: one in the SF Chronicle and one in the National Journal.

For all you NH Deaniacs who read my blog, don't miss today's WaPo article about the Obama campaign using the organizing model of voter contact. The article will make you all warm and fuzzy when you run across terms such as "one on one" and "house meeting".

One of my biggest regrets over Dean's loss in New Hampshire was that I didn't think our style (i.e., Marshall Ganz's style) of campaigning would be validated. Even though NH voters were subjected to a week long media story on "The Scream", we still managed to come in second in the Granite State. Unfortunately, it was a distant second, and the media gave us little credit for garnering over a quarter of the vote in a crowded field with a candidate who had already imploded. The media narrative of the "Dean Machine" stopped on that Election Night.

It shouldn't have. Anyone on the ground will tell you that energized Dean Democrats helped flip NH's governor and congressional representatives to Blue over the next two cycles. Most campaigns tap into existing party resources, cajole partisans to volunteer and myopically focus on getting votes on Election Day. In contrast, the organizing model empowers the partisans to take control of their own neighborhoods and create ties to other Democrats that last much longer than a fall campaign season. As Lavalee said, "We left more than just yard signs".

Thankfully, in 2007, ex-Deaniac Jeremy Bird hopped on the Obama wagon and ran the organizing model in South Carolina. Check out the part of Obama's Philly Speech on Race about Ashley to see the fruits of always asking "What's Your Story?" And now the WaPo article indicates that this organizing model is now nationwide: huzzah!

I am a bit disappointed that the article didn't make the Labor organizing->Ganz->Dean->Obama connection, but it's probably too much these days to ask reporters to use Lexis or Google.

I griped to Janeite about this WaPo story when it came out last week, but now that we have more fundraising numbers, I feel confident enough to gripe to the world about it.

The article says that because Obama is asking donors to give to Clinton (to help retire her debt), his fundraising machine is showing "signs of wear." The only empirical evidence in the article is displayed on the graph on the right, which shows Obama's per-month fundraising numbers falling from February to May. Yes, that's right, Clinton's debt is to blame for Obama's "poor" fundraising ($22 million in May still seems high to me) during a period when Clinton was still in the race. When the Post decided to run this article on July 11th, five days before June fundraising numbers are reported, they really didn't know what the true impact of Clinton's debt on Obama's fundrasing would be. The data don't say one way or the other. Maybe they would get lucky and Obama would have a terrible June in terms of fundraising.

Not so much. Obama raised $52 million, more than doubling his May take. Oh, and that measly $22 million that Obama raised in May? It's the exact same amount that McCain raised in June. Still waiting for the Post article about how the McCain machine is lackluster.

Late Update: After years of being behind, Dems finally have fundrasing parity with the GOP: Obama+DNC=$92m CoH, McCain+RNC=$95m.

Congrats to Samidh, whose company Pluribo launched their first product a couple days ago! In a nutshell, it's a Firefox plug-in that takes Amazon product opinions (just for electronics at the moment) and summarizes them into one sentence via an automated algorithm. To see it in action, take a look at this screencast, with narration by Samidh.

I've tried it out, and it's very cool. Actually, now that Pluribo is released, maybe the team can turn Pluribo on itself and summarize what people are saying about it!