But unlike African Americans or college students, WVWV targets were not geographically clustered in specific neighborhoods whose sidewalks could be stalked by clipboard-carrying canvassers. Gardner knew she would have to seek them out individually, but it wasn’t obvious how to do that: Databases used by campaigns and parties usually included voters’ ages and flagged them by race, but did not identify their marital status. But Gardner happened to be launching her new group just as the political world was discovering the value of commercial data warehouses initially created for credit agencies and consumer marketing. She bought lists of all voting-age adults and then cross-referenced them with voter-registration rolls: The names that appeared on one and not the other became a potential universe of targets. She then had analysts develop statistical models that mined demographic variables for each woman on that list to predict the likelihood that each was unmarried. Women’s Voices Women’s Vote didn’t have to wait for these voters to come to them. They sent the women on their list a letter.
During campaign years, the group issues quarterly mailings.* Each mailing has a randomized experimental design built in, allowing the group to test various ways of packaging a registration form, adjusting everything from cover letter language to envelope typography. By paying the postal service to track which of its forms were sent along to registration authorities, it is possible to determine which packages get the best response. In a 2010 test, for example, WVWV found that Ohio-specific forms sent to targets in that state got returned at a rate 1 percentage point higher than a standardized national alternative. Analysts can also apply statistical models to identify the attributes of those recipients who have been most responsive to each appeal. With time, WVWV learned that those who have recently moved, or just turned 18, are twice as likely to respond to any mailed registration appeal as the broader population.
After the group’s June 2012 mailing, WVWV analysts and designers made 13 improvements to the way it builds a list of new targets.* Each was designed to slightly increase the odds of response, and thus lower the cost of a registering a voter. A decade of such tests and tweaks has cut the price of a new registration more than half from the $15 that Gardner estimated it cost a decade ago to administer traditional site-based programs using paid canvassers. “We have kept driving the cost of registering a person down cycle after cycle,” she says. “Our lists are getting better, our models are getting better, our targeting is getting better.” With that data-centric approach came something unimaginable at the card table: the ability to put on the rolls only those people you wanted to see cast a ballot.*
As ACORN became the most visible practitioner of site-based registration drives, WVWV perfected a more refined régime largely out of public view. When the group did surface to broader attention, it was usually because one of its registration packages had been addressed to someone who should not have gotten it: a corpse, a pet, or someone already registered. The explanations tended to turn less on nefarious objectives than shoddy data sources: The Social Security “Death Master File” had failed to deliver a complete index of the deceased, for example, and anyone who has complained about seeing mail for a pet almost certainly put the animal’s name on a warranty or marketing form that fed into one of the commercial databases where WVWV hunted for names of voting-age adults not already on the electoral rolls. Just last week, the Virginia elections board rejected a request from Mitt Romney’s campaign to investigate the group’s mailings to those ineligible to register.
Last year, Gardner’s group renamed itself the Voter Participation Center to reflect a portfolio formally broadened after 2008 to cover what it calls the “Rising American Electorate,” adding unregistered minorities and young people to its mailing list. Gardner emphasizes that the officially non-partisan VPC chooses its targets because they are underrepresented on the voting rolls, not because they are likely Democrats.* But its new universe now neatly covers the core Obama coalition. This year, collaborating with groups like NAACP and the National Council for La Raza to handle their targeted mail registration programs, Gardner expects to help put 800,000 new voters on the rolls for $6.32 each. “We knew from our previous work that the way we registered single woman worked as effectively with other people,” Gardner says. “We knew a mail-based, list-based approach can reach the people who need to be registered in this country.”
* * *
Obama’s team does not appear to have sent out a single registration appeal by mail this year, according to sources inside and outside the campaign familiar with its practices. The choice to stick with the card table exposes an identity crisis that continually vexes decision-making at Chicago headquarters: An empirical campaign built for efficiency can’t always afford to follow the data.
Obama has chosen to adopt the more labor-intensive approach to registering voters, as opposed to the capital-intensive method of acquiring data on prospects and screening through the mail. In large part, that’s because human capital has arguably become the defining obsession of Obama’s approach to electioneering. When the New York Times recently questioned the campaign’s profligate early spending, an Obama spokesman said it was in the service of the campaign’s priority: “building and maintaining our grassroots foundation.”
Much of that grassroots foundation is structured toward the goal of enlisting even more volunteers. A week ago in Northern Virginia, when Mitt Romney’s callers and canvassers found a voter who supported the Republican ticket, he or she was asked if they wanted to submit an absentee ballot request. When Obama canvassers nearby found their backers, the follow-up question was not about voting but volunteering: did the supporter have time to help out?
As a result, the campaign has had a surfeit of free labor, and for much of the year needed activities to keep it occupied. Local organizers worried about losing committed volunteers if they were not continually engaged, even if it was too early in the campaign year to dispatch them to turn out voters or persuade those still undecided. To some degree, in-person registration drives became a make-work project for a campaign looking to test itself.
There are advantages for Obama’s team to be bringing in its own registrations and not solely relying on allies like the Voter Participation Center to do so. By entering data forms immediately as they are collected, Obama’s targeters can begin dispatching mail and phone calls to pending registrants and treat them as voters, even if the board of elections is struggling to clear a backlog of new forms and updated its public voter file—a period in which the Romney campaign may not even know the person exists. At the same time, the campaign’s sophisticated voter-targeting work has given local field organizers newly granular data that can be used to direct site-based programs to locations likely to be have the most plentiful concentrations of registration targets.
But the campaign’s tacticians are aware that, even with such improved geographic targeting, they are in the unusual position of willfully disregarding state-of-the-art campaign science in service of a larger mission. “Everything is about building ‘the biggest grassroots campaign in history,’ ” shrugs one consultant affiliated with Obama’s operation. Accomplishing that goal, campaign officials appear to have decided, may require putting a few new Republicans on the rolls.
Corrections, Aug. 15, 2012: This article originally stated that the Voter Participation Center sends out its registration appeals monthly. (Return to the corrected sentence.)
The article originally stated that the Voter Participation Center made 13 changes to the design of its mailed packages. The changes were to the way it builds its list of targets. (Return to the corrected sentence.)
The article originally inaccurately reported that the Voter Participation Center used microtargeting models to individually identify potential voters most likely to be Democrats. (Return to the corrected paragraph.)
This sentence was revised to clarify that the Voter Participation Center describes its work as reaching out to groups underrepresented on the voting rolls. (Return to the corrected sentence.)