The Victory Lab (30 page)

When the final week of the campaign arrived, Rogers believed everything was in place. But he couldn’t actually reach his campaign contact—a dozen e-mails, all unanswered—and Rogers grew increasingly panicked. Eventually the bad news reached Rogers through other channels: his experiment had been killed. The campaign leadership had decided that Clark, who had skipped Iowa to contest New Hampshire, couldn’t afford to take voters out of his contact universe before a crucial primary just to maintain the sanctity of Rogers’s research, although Rogers never heard it directly from the person he thought was responsible for the decision. “The decision makers never told me,” says Rogers. He adjusted to the reality that he would spend the closing days of the campaign as a volunteer knocking on doors and not as a social scientist presiding over a potentially groundbreaking experiment.

That weekend, Rogers had stopped in a supermarket in Derry for some basic shopping. As he rounded the edge of an aisle, he faced the campaign official who had first signed off on his experiment but then dodged his e-mails, thus bringing the project to an ignominious end. Rogers begged for an explanation. Why had the campaign given up on an experiment it once saw as so promising?

“There is no day after the election,” the official told him.

Rogers insisted that Clark’s operatives in other states would be able to integrate the findings to adjust their GOTV tactics in later primaries. “You’ll learn in time for it to be useful in the next primary,” Rogers pleaded.

The two argued about the project’s fate near an end display. “They just shut it down,” Rogers recalls, before wryly noting that Clark’s third-place finish
in New Hampshire effectively ended his presidential campaign. “Turns out they were right: there was no day after the election.”

Rogers returned to Harvard with mixed feelings from his first encounter with politics as a social scientist. If anything, the resistance he had encountered—and the institutionalized lack of curiosity it appeared to reflect—made Rogers all the more interested in studying the frailties of political communication. But he wondered whether he had the fortitude to continue begging campaigns to effectively sponsor his research. “My experience with the Clark campaign prompted me to do experiments on my own,” he says.

A colleague recommended that Rogers contact the Yale political scientists who had developed something of a franchise running these kinds of experiments. Rogers wrote to Gerber, and shared with him a copy of the social norms paper he had written the previous spring and his idea for how to structure an experiment to test his thesis in a live campaign. Gerber already had a busy research agenda for 2004, but he and Rogers made plans to take the turnout messages into the field the next year, when a few states would have governor’s races. Within weeks they had started to design an experiment testing messages that they could run in the following year’s New Jersey gubernatorial election.

Rogers first had to prepare the terrain. That summer, he visited the Democratic National Convention in Boston, and with a clipboard prowled the corridors of the Fleet Center looking for party operatives. He ended up surveying fifty of these “self-identified experts in voter mobilization”—as Rogers later identified them with as much sarcasm as a dissertation would permit—to ask whether they thought it would be more effective to use a message that warned that turnout in a coming election was expected to be high or low. The survey had no statistical power—the sample was too small, and it had not been assembled randomly. But Rogers thought he was on the cusp of undermining a matter of standard political practice, and before he destroyed that conventional wisdom Rogers wanted to enshrine it for the historical record. He was benchmarking the opinion of so-called
experts, so that after his study none of them could say:
Duh, that’s obvious, we knew that all along
.

Rogers knew that his research path was putting him on a collision course with political professionals he began to describe as “shamans.” From his psychological training, Rogers instinctively resisted the idea that the decision to vote was the self-interested choice described by political scientists. Instead he thought of it as “
self-expressive social behavior.” He didn’t visualize the mathematician in the privacy of the voting booth, coolly calculating how to maximize his democratic utility, or the accountant attaching a price to civic engagement. He wondered what happened to that person in the days before and minutes after casting a ballot, about interactions that brought shame or pride or a desperate need to fit in. For Rogers, this seemed like a modest intellectual breakthrough, “just taking a behavioral science understanding of how people conform to the behavior of others and applying it to politics.” As a result, Rogers ignored the fine demographic distinctions that consumed many working in politics and instead focused on questions that went deeper into what he considered the universal human condition. Were there things campaigns could say to people that—regardless of their ideology or background—would nudge them to the polls?

MARK GREBNER WAS STUMBLING
to a similar conclusion from a different starting point. Since 1974, he had run almost continuously for two-year terms as an Ingham County commissioner, always representing a district containing the Michigan State University campus. Grebner was a familiar college town archetype, an ungainly hybrid of perpetual student and public intellectual, of the local institution but with few credentials to attest to his ties. On the wall of his office two blocks from the campus’s edge, there is no diploma to commemorate the eleven years it took Grebner to complete the requirements of a bachelor’s degree, a path detoured in large part because he insisted on taking graduate courses instead. A
more impressive token is framed just above Grebner’s most comfortable chair, which offers a natural opening for any conversation he starts there: a primitive megabit chip, one million bits strung together, the size of an open magazine. When the Michigan State computer lab retired the part, which was rumored to have once cost four hundred thousand dollars, in favor of more compact processors, it was given to Grebner, the way a renovating pub might save a familiar bar stool for a prolific patron. “It’s on loan, technically,” he says.

Grebner came of age with the technology. “I met computers when I was eighteen,” he says, referring to the day in the spring of 1971 when as a college freshman he enrolled in a computer class. He had turned eighteen just in time to take advantage of a constitutional amendment granting him the vote, and as the MSU campus was being radicalized by Vietnam. He volunteered for local Democratic campaigns but grew impatient with their culture of endless deliberation. “Very quickly I looked around and realized that the thing I could do that was the most valuable was not sit in these meetings and argue about what color the campaign lawn signs should be, but was to computerize stuff,” says Grebner. Campaign records were kept in file folders, with typed lists of voters or legal pads scrawled with the names of those who had received lawn signs two years earlier. Grebner began buying computer time at the university lab, coding punch cards to feed a mainframe database. Even though he had converted the information into digital form, it was no less bulky: the records of ten thousand voters meant five boxes of computer cards weighing fifteen pounds each. Moving data from the lab to a campaign office required pulling a hand truck across campus. “It was almost closer to threshing wheat,” says Grebner, for whom hardship became a teachable lesson he likens to Soviet scientists who developed a knack for improvisation because their labs were often short of glassware and other basic materials. “Most of what I know I picked up when computing was relatively difficult.”

One day, frustrated by a message strategy meeting for a state legislative candidate that ended without an agreed-upon message, Grebner
walked out and decided he would just write his own literature for the candidate. The three-panel brochure proved memorable only for the fact that Grebner avoided the campaign’s internal bureaucracy to get it copied, folded, and into the hands of volunteers to distribute it. “People said, ‘But that wasn’t approved?’ ” Grebner recalls. “No, it wasn’t. But unlike the other ones, it actually got written.”

In 1974, he ran for office the first time, for a part-time seat on the county’s commission. At first, Grebner did almost all the canvassing himself, a one-man GOTV operation trundling from dorm to dorm with a list of voters. He would remind the students that election day was approaching, and ask if they planned to vote. They would almost always say yes, and following that, Grebner would usually give them his pitch. After rounds like this, Grebner felt unsatisfied, as though he was not having a meaningful interaction with voters; the yeses seemed mindless, as were the inert nods he saw as he talked up the Democratic ticket. So he started asking people if they were going to vote, and then when they said yes ended the conversation. “Oh, good, thanks,” he would say, then mark a check next to their name and move down the corridor. “This leaves the person who you interview with this very unsettled feeling,” Grebner says.

Grebner liked that unsettled feeling, and he felt especially good when he saw it reemerge on election night. He had developed a computerized system with punch cards for each of the voters who had self-identified as his supporters and became his GOTV targets. As they showed up at the polls on election day, Grebner would remove their card from the pile, and then at around 6:30 p.m. he would use the remaining cards to track down those who had yet to vote, and hustle to their dorm rooms to pull them out. Grebner showed up at the doors of his targets, and even before he could reintroduce himself he started hearing excuses and justifications. “I changed my mind, you can’t tell me what to do,” one nonvoter told him. Another declared defiantly he had just gotten back from the polls. “As soon as they see you with the deck of cards, it already bothers them so much that they’re telling you,” says Grebner. “That is so different from the normal experience
of GOTV, so it was clear to me I was onto something.” A few years later, Grebner felt that his hunch was vindicated by something of a natural, if unscientific, experiment. He split campus GOTV responsibilities with another campaign but found out after the fact that his partners had failed to follow his peculiar script to abruptly short-circuit the conversation. After the election, Grebner saw that the six precincts he had canvassed significantly outperformed the other five. “We had dedicated the same resources to it. The only thing different that we had done is they screwed up my interview,” he says. “I almost thanked them for having done that.”

In 1982, Grebner had an epiphany. He was reading the book
Who Votes?
by Raymond Wolfinger and Steven Rosenstone.
Their work relies on an unusual data set that few other political scientists had ever fully appreciated.
The Census Bureau’s Current Population Survey questions
a rolling panel of fifty-six thousand households every month to prepare the regular labor market report. But every two years, it adds a battery of post-election questions to the mid-November survey that yields the unemployment figures announced in December. Commerce Department interviewers do not ask people to say how they voted or why, but they do ask if respondents are registered and if they voted in the election that took place earlier that month. Often the number gets reported in the press as a first estimate of nationwide voter participation that year, and it almost always prompts stories about how unusually high turnout had been. Months later, when researchers are able to fully examine updated state voter rolls, the Commerce Department estimate gets revised downward by as much as 10 percentage points. As this pattern became evident, academics who had relied on such self-reported turnout figures (including those working on the University of Michigan’s national election studies) began to worry that their numbers were inflated by inaccurate survey samples.
So they started matching respondents to the voter rolls. The survey methodologies were right, but as many as 10 percent of its respondents were wrong. They hadn’t voted.

“The fascinating thing for me is they conduct this survey ten days after the November election, and already ten percent of the population
is lying about having voted!” says Grebner. “They didn’t forget, right? If there were a forgetting process it would be gradual. I refuse to believe that people honestly don’t remember if they voted ten days earlier. Bullshit! A whole bunch of people lied.”

This piqued Grebner’s attention, in part because the political science literature had offered a lot of explanations for why people vote—their civic duty, their desire to pick a winner, their calculation that a particular result would serve their self-interest—but no one had proposed an incentive for lying about having turned out. “Hypothesis: Why do people fail to vote? This is the classic answer:
Because they didn’t know it was election day! They’re too fucking stupid to know it was election day!
” says Grebner. “Crap! If you had done a survey of severely psychotic people, people living under bridges, people with IQs below sixty, and you asked them, ‘Is today election day?’ I’d bet you eighty percent of the vegetables in this country knew it was election day. If you asked psychopaths, ‘Is voting a good thing or a bad thing?’ I’d bet eighty percent of the psychopaths in this country—people who’ve completely rejected all social norms—would be able to tell you the right answer about that social norm,” says Grebner. “So reminding people ‘It’s election day, you should go vote’ or telling people ‘It’s a good thing’—are there people who are unfamiliar with this norm? That norm is universal; the knowledge it’s election day is universal.” But campaigns kept using the same language to beseech people to vote. “Why do people go out and sing Christmas carols before Christmas?” he asks. “Because they’ve always done it!”