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.

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!

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.]

• 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.

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.

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.

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!

I'd like to add that Tyson validates his model in an extremely poor way. He proposes an inference method (the "median poll method") for using early polls to predict the Novemeber winner. Thus, he should test this model on early polls in modern presidential elections. Instead, he applies his model with only October polls and only for the 2004 election. Not surprisingly, he finds a "good" model; the median method only mispredicts the winner in one state (Hawaii). Heck, here's a model that misses only two states in 2004 (NM and NH): just predict the winner by who won last time. Clearly that model sucks historically (cf., predicting 1992 from 1988), and I'm guessing so does the "median poll" method (if you examine the historical May state polling).

As Gelman says, Tyson should consult political scientists (such as James Campell, who has done a lot of work on this topic) before penning an op-ed for the Times. If he does that the next time around, I promise not to write my next blog post on Bell's Inequality.

But I want to take a moment to pause and reflect on just how close this nomination election was. For first time since the introduction of superdelegates, the distance between the top two candidates, in terms of pledge delegates, is less than the number of superdelegates. Only 31 pledged delegates are up for grabs tonight and Obama needs some of these delegates to secure the nomination. (In contrast, the last primary day in 1984, which Mondale needed to garner a majority of delegates, included California and New Jersey.)

Obama's popular vote margin (with MI allocated by the exit poll) will end up being about 70-80K, or 0.2% of the popular vote (48.1% - 47.9%). Wow.

I'd also guess that this nomination race is the last time the eventual winner needs the last day of elections to secure a majority of delegates before the parties decide to reform the system and move to a semi-national or regional primary process. Something to tell the grandkids about :)

Next Morning Update: Including Michigan's 30,000 write-ins and allocating Michigan by the exit poll, my final popular vote numbers are 18,087,958 (48.0%) for Obama to 18,004,743 for Clinton (47.8%), a difference of 83,215.

Grad student 1, undergrads 0.

My home-brew TiVo is dead.

(What? You thought I was talking about something else?)

I "called it" at 12:46pm on May 24. Cause of death: hard drive failure. *Sniff*. The MythTV-based DVR lasted exactly three years, as one of the first things it recorded was the Champions League Final of 2005. It blew out just before it could record this year's final, and two weeks of recovery efforts ended in failure.

Ah, well, I'll have to buy a new hard drive. I think I'll go with Ubuntu instead of Fedora for the Linux distro this time around, as Ubuntu seems to be the new standard. Hopefully this installation will be easier than the first go-round.

TJ: You should get a tatoo!
Me: Um...ok.
[We walk over the fake tatoo station.]
TJ: Get the pirate one!
Me: Ok.
TJ: Put it on your forehead.
Me: ...okay.

(I'm not a sucker for kids at all! /sarcasm) Thus I walked around with a pirate on my forehead for the rest of the day. I received some distressingly weird looks at the D-Bar later last night; I miss MIT, where no one would have batted an eye.

*name changed so I don't violate any Internet laws.

Maybe the fault lies elsewhere
And obama's behind this scare
There's no act too low
Where they wouldn't go
Since his wins are increasingly rare

No matter for this Hoosier state--
It's really a matter of fate
There's no better pal
HRC's our gal
Tomorrow night we'll win by eight!

6:46 PM L: were you really canvassing all day?
7:20 PM me:[just] this morning
7:26 PM L: you are a good man [What's your prediction]
7:27 PM me: I think HRC by 11

Back in 2006, Taren and I canvassed MD before the general election and I was dissappointed to find that the campaign was still creating turf manually, with scissors and a highlighter. In the summer of 2006, I wrote TurfCutterGM as a "proof of concept" program to show how you could canvass without resorting to such crude, 20th Century techniques. I promised Taren that day that this would be the last election cycle that people "ran (for office) with scissors" and that in the future campaigns would use online mapping tools to "cut" their turf.

Flash forward to 2008, and I am happy to report that the prophecy came true. I'm not sure what system the Bristol HRC campaign was using, but I assume it was the VAN, since it seemed to integrate so seemlessly into their voter file (which is VAN). VAN bought a license to the TurfCutterGM code in early 2007 for $1 (still haven't seen that check!), so it's possible that my canvass route was created with some of my own code, but my guess is that VAN just took my concept and wrote their own code.

And, I must say, canvassing with a computer-generated, unobscured-by-highlights map with each house displayed as a pin point was a very pleasant experience and made the whole process much more efficient from the canvasser's perspective. We'll all find out shortly if my efforts paid off!