During an April 25 conference call to unveil the Obama campaign's 50-state voter registration drive, deputy campaign manager Steve Hildebrand sounded genuinely exasperated as he tried to steer reporters to the topic at hand.
Despite his best efforts, those on the call didn't request any further details about the effort. No one posed questions about the drive's goals or how the campaign thinks it could affect Obama's general election prospects. Instead, they brought up the latest distractions: What does the campaign make of Jeremiah Wright's new comments? Are advisers worried that Obama has a problem attracting working-class white voters? And so on.
The media's lack of interest is hardly surprising. While voter registration drives are invaluable, they're also dry and tedious -- not that exciting to participate in or to write about.
But some new data indicates that, come November 5, we may find ourselves looking back at this year's intensive voter mobilization efforts as what put Obama -- and down-ballot Democrats nationwide -- over the top.
The Poblano Model
This new data was generated by the anonymous 30-year-old author of the website FiveThirtyEight.com. A statistician and analyst, "Poblano" (who asked that his real name be kept private for professional reasons) has developed a sophisticated regression model that uses state-by-state polling data to assess possible general election outcomes in individual states.
"We weight the average based on the reliability of the pollster," Poblano told me. "The polls that have a better track record are obviously going to be emphasized more heavily. We look at sample size. We look at how recent the poll was. We also skew various demographic factors as well."
In the Frequently Asked Questions section of his site, Poblano lists the demographic variables that his model takes into account in each state. Included are John Kerry's 2004 performance, the candidates' respective fundraising efforts, and other factors such as income, race, religion, age, and education level.
Using the resulting baseline averages, Poblano runs simulations -- 10,000 at a time -- to determine the probability of either Obama or Hillary Clinton beating John McCain in the fall.
"The simulation takes the polls we have now and recognizes that this far out from the election there are a lot of uncertainties," Poblano explained. "There was a point in time [in 1984] at which Walter Mondale led Ronald Reagan by 17 points or something. So we look at how much polls move over time, plus the margin of error in each poll, plus the fact that polls are never as good as they claim to be And we just simulate around that."
Poblano's simulation engine has produced some fascinating results. According to his current data, the model predicts that Clinton would win four states against McCain that Obama is favored to lose (FL, AR, WV, OH). Meanwhile, Obama wins eight states where Clinton would likely fail (MI, WI, IA, CO, NM, NV, WA, OR).
Using his simulation engine, Poblano went way against the grain last Monday, predicting that Obama would win the North Carolina primary by 17 points (most expected him to win by a single-digit margin). The following day, Obama took the state by 14.
Poblano's Indiana projection was even more spot on. He predicted a 51-49 Clinton victory. And sure enough, that's what happened.
Later in the week, pollster Mark Blumenthal devoted his National Journal column to the "Poblano Model," noting that it had outperformed five major national polling operations in predicting the results of Tuesday's primaries.
POST COMMENTS BELOW:
Alternet.org
http://www.alternet.org/election08/85549/
