The Victory Lab (47 page)

Plouffe’s decision in early summer to strike a deal to buy data from Catalist was met with suspicion by many in Chicago, since it meant bypassing free access to a national voter file that had been assembled by the Democratic National Committee at significant cost. But being one client among many in Catalist’s portfolio of progressive institutions allowed the campaign to create seamless links across the activist left, including outside groups with whom candidates were legally prohibited from coordinating directly. When Democracia USA collected a new voter’s registration form in Florida, Obama’s targeting team often knew about it before the local board of elections. Democracia would create a record in its databases, which synced daily with Catalist’s servers. When the Obama campaign conducted its daily download from Catalist’s database, per its contract with the information vendor, the new record would show up in the VAN. The campaign could start treating the person as a voter—assigning model scores, canvassing her, communicating by mail and phone, or getting her an absentee ballot—even before the registration had been officially processed. Republicans wouldn’t have any idea the new voter even existed until she went on the books.

Moffo set out to build a system that would allow Obama’s field staff to extend its continuous profiling of the electorate straight through the moment the polls closed. Even the campaigns that were best at tracking their voters usually lost the trail on election day. The Iowa Democratic Party, known for one of the country’s most efficient general-election vote-rendering operations, had for a generation relied on multiple stacks of cards printed with the name of every Democratic voter. At the polling place, as the judge of elections marked the name of an arriving voter, the person’s card would be thinned from the deck. Seven times a day, a runner would get the cards and deliver them to volunteer door-knockers who, three times over the course of the day, did sweeps of houses. Cards still
left in the pile at the end of the day would get passed to callers at a phone bank for one last contact. Other local party organizations had their own versions of Iowa’s cards.

Moffo believed that new technology had become possible for the names of those who had voted to instantly disappear from a GOTV walk list or call sheet, and thus was born his new system, Project Houdini. (Four years earlier, campaigns tried to automate the procedure with Palm Pilots, but the devices had to be plugged into wired computers to download the information and move it into databases.) Each voter in a precinct was given a four-digit code, and as soon as he or she cast a vote, an Obama worker assigned to the polling place would punch in an update by phone to automatically tag the voter’s name in the VAN. As a result, field directors could continuously hone their election day GOTV programs in real time, ensuring that volunteers were dispatched to rouse only nonvoters. As that process freed capacity throughout the day for more contacts, field directors could move past their list of initial targets to include voters with lower turnout scores who were thought to be more of a reach.

Houdini worked magnificently during the morning rush on the East Coast, and a half million people who voted before 9 a.m. were disappeared from future target universes. Hundreds of thousands of people across the country were helping to turn citizens into voters on Obama’s behalf—some walking their neighborhoods, others joining the National Call Team from their home computer and phone—and none of them were wasting their time talking to people who had already cast a ballot. By midmorning, those monitoring turnout operations from Chicago saw Houdini begin to slow: as voters began arriving at West Coast polling places, they overwhelmed new phone lines that had been installed at headquarters to handle the expected heavy volume. By 10 a.m., the whole system had locked up, turning away new callers. Moffo hectored the phone company, which told him it was too late to lay additional cables. Others concocted work-around fixes, like having poll workers call a local field office so voters’ records could be updated manually by computer there.

On the other side of headquarters, the campaign’s new-media team watched National Call Team volunteers continue to log on to the MyBarackObama site to make GOTV calls. Their numbers barely slowed even as the wave of states going blue moved west across the country. News organizations would not formally call the election before polls closed on the West Coast, but their coverage assumed a foregone conclusion: Obama would be the next president, and would get there by vastly exceeding the necessary 270 electoral votes. The technologists were amazed to see that, even as it became evident that the race had been decided, the number of volunteer callers surged. The minute the last polls closed in the continental United States, all the TV networks simultaneously declared Obama the victor. The eleventh floor echoed the exultations made by the quarter-million people celebrating in Grant Park a mile away.

Eventually Uday Sreekanth, the deputy chief technology officer, looked back at his computer screen and saw that MyBO was still clogged with National Call Team volunteers looking to do their parts. He performed some quick calculations and thought about the number of minutes left for voting in the country’s newest states. Hawaii, where Obama had been born, was likely to deliver his largest margin of victory, but onetime target Alaska had fallen off Obama’s map when its governor had joined McCain’s ticket. “We have the capacity to call every voter in Alaska three times, and really embarrass Sarah Palin,” Sreekanth alerted his colleagues. “Should we pull the switch?”

They decided not to, and walked away from their screens filled with data, down the elevator, and out onto Michigan Avenue, where they were captured by a stream of people—each one believed at that very moment to be 100 persuasion, 100 turnout, and 100 openness—all headed in the same direction.

PUSHING THE ENVELOPE

B
arack Obama’s election took place exactly one decade after Alan Gerber and Don Green had taken to the streets of New Haven to run their pioneering electioneering field experiments at Yale, and even forward-thinking Republican operatives found something to cheer in the way his ascendance represented a triumph for new-wave empiricism over the reactionary, clubby old campaign world. Finally an industry that had been chronically unreflective about its failures was obsessed with learning from Obama’s success. Naturally this swung all too quickly to its own thoughtless excess: any operative with the most fleeting connection to the Chicago operation was reborn as a political celebrity, and any tactic that could be marketed as “Obama-style” found immediate global demand. Meanwhile, the earliest revolutionaries—the social scientists and statisticians who had laid the intellectual foundation for Obama’s victory even as they stood apart from it—continued to work away at a fundamental
question the field had not yet fully answered. By digging even deeper into psychology and the behavioral sciences, they hoped to finally crack the code of what can turn a person into a voter.

In the spring of 2007, as the Obama campaign was opening its first Iowa offices, Alan Gerber had written to Hal Malchow to let him know he was coming to Washington, and the two arranged to get dinner. The Yale political scientist arrived at the downtown steakhouse Morton’s accompanied by a surprise guest, a Harvard graduate student on whose dissertation committee he had just served. But the presence of Todd Rogers wasn’t the only evidence that Gerber’s attentions had recently swerved dramatically toward questions of voter psychology. Once they were seated and had specified the temperatures for their steaks, Gerber detailed an experiment he and his collaborator Don Green had just completed with Mark Grebner, the Michigan voter-file manager who had gleefully threatened to out nonvoters. Malchow’s eyes opened wide as Gerber described the four different versions of Grebner’s postcard, and the staggering result that the “neighbors” approach had delivered—three times better than any other technique in the bulging mental archive Malchow maintained of hundreds of voter contact experiments.

Malchow had been one of the first political consultants to fully embrace the use of randomized field experiments, but his relentless commitment to following their findings had not always been good for his business. Even if Gerber and Green’s experiments had shown mail could be effective, their preferred efficiency metric of dollars-per-additional-vote highlighted how costly it could be to get the desired results. (This was supposed to measure the price of mobilizing a marginal voter, calculated by dividing the return from a given get-out-the-vote technique over the cost of delivering it.) When the academics compiled dozens of varied mail experiments, they found that it took 333 pieces of mail to turn out one new vote. Malchow added design, printing, and postal fees and realized that the product his firm marketed to campaigns cost them eighty dollars per new vote.
Instead of masking this unpleasant fact, Malchow loudly chastised employees who persisted in preparing get-out-the-vote mail, even though much of his firm’s revenue depended on it.

But the Michigan experiment showed it was possible to improve that math dramatically: you had to show only twenty citizens copies of their neighbors’ voting histories to convert a new vote. This arithmetic warmed Malchow, who saw in Grebner’s social-pressure breakthrough both redemption and opportunity. Here at last was get-out-the-vote mail that worked, and fabulously so.

“Alan,” Malchow said across an expanse of white tablecloth. “I will pay you a hundred thousand dollars if you won’t publish the results.”

Gerber turned away Malchow’s only half-joking proposal, but he agreed to show his dinner companion the experimental materials well before they would become public in the
American Political Science Review
. Malchow rushed to take advantage of the head start he had been given to master social pressure. The next election on his firm’s calendar was a Dallas municipal election, working with a gay and lesbian group doing independent advocacy on behalf of a candidate running to be the city’s first openly gay mayor. The group refused to embrace the approach of revealing neighbors’ vote histories, for fear of inflaming an already delicate contest; the best Malchow could do was persuade them to mimic Grebner’s “self” mailing, which includes only voting records for those within the receiving household. Even so, the group braced for an unfavorable response, creating a front group with an anodyne name and a decoy return address. They were right to fear a backlash; a local Fox television affiliate tracked the mysterious letters to a Mail Boxes Etc. location that Malchow’s clients had used as their address, and parked a camera crew outside on an ultimately fruitless stakeout. “I don’t think anyone went back there for two months,” Malchow says.

Malchow knew that the candidates, party committees, and major institutions that made up most of his clientele would not find similar glee in resorting to such stealthy tactics. The social-pressure technique needed refinement if he was going to put it into wider use, and so Malchow turned
where he often did when he had an idea he wanted to test. The group Women’s Voices Women’s Vote had been established by Page Gardner in the wake of the 2000 election, when strategists in both parties began to prioritize turning out their known supporters instead of hunting for swing voters to win over. Gardner had been startled to see exit polls expose what she considered a “marriage gap,” a staggering split in voting behavior between married women and unmarried ones. The latter were among the most loyal Democratic blocs but among the least likely to actually vote. Commercial database vendors did not maintain reliable lists of who was married and who was not, and so Gardner hired Malchow to see if he could use statistics to predict whether an individual was single. Malchow liked the puzzles this challenge posed: if two people of similar ages live at the same address, how do you tell if they’re married and not roommates or siblings?

With the 2008 campaign looming, Gardner expanded the group’s mandate to focus on other parts of what she called the “Rising American Electorate,” including not only unmarried white women but Latinos, African-Americans, and young voters of both genders. Together they comprise 53 percent of the voting age population, but they are chronically underrepresented at the polls. The question wasn’t whether they would vote Democrat or Republican, but if they could be made to vote at all. (Even though the group was officially nonpartisan, for tax purposes, there was no secret that the goal of all its efforts was to generate new votes for Democrats.)

To placate the Women’s Voices donors who would have to back the new social-pressure technique, Malchow set out to find more delicate language that could maintain the implied threat without making the recipient feel like he or she was under investigation. He designed an experiment to take place in the Kentucky governor’s race that fall, replicating the original Michigan experiment. Malchow, however, added a new twist that went beyond the one-way communication of Grebner’s mail. One group would get phone calls asking if they were planning to vote in the November election. Those who said yes were sent a simple letter restating that commitment followed by a robocall just before election day reminding them that they
had pledged to vote as part of a study that would check on their follow-through afterward. It ended up being a small group—only 30 percent of those initially contacted said that they intended to vote—and a complicated technique to execute. But while the sequence of calls and mail was expensive up front, the return was so good that it proved a relative bargain, producing new votes at eighteen dollars each.