The Victory Lab (49 page)

But it was that “stuff”—including running more than one hundred field experiments since the group’s founding—that won Malchow’s heart: Women’s Voices had impressed him as perhaps the most empirically minded of all the institutions and candidates for whom he had worked in twenty-five years. “They have the best agenda of everybody out there,” said Malchow. When he shut down his firm, Malchow saw no reason to cut
his ties to Women’s Voices. In fact, its research agenda seemed the best way for Malchow to relieve himself of the corporate hassle while keeping his hand in the one part of politics where he believed he could still make a difference: the semisecret cabal of social science experimenters that he considered the only hope for imposing accountability on the multibillion-dollar industry that helps Americans choose their leaders.

Yet when Malchow arrived at Gardner’s 2011 summit, he was quickly bored by what felt like another PowerPoint-intensive reminder of how poorly Democrats had fared in the prior year’s midterm elections. Pollster Celinda Lake navigated slides showing that, while the Rising American Electorate represented 46.6 percent of the electorate when Obama was elected in 2008, the share had fallen to 41.9 percent two years later. Women’s Voices registered one million people in 2008, but two years later few of them bothered to cast a ballot; only 3 percent of the people Women’s Voices contacted during the 2010 elections had ended up turning out. Everyone in the room knew the implication of those numbers: unless Democrats could figure out by 2012 how to reanimate this core part of their coalition, Barack Obama would have a very difficult time winning a second term. “The key fight,” says Gardner, “is: who’s in the electorate?”

The presentations dragged on, and Malchow perked up only when he heard an old friend describe a novel tool for getting registered voters to turn out to cast ballots. In his presentation, Alan Gerber told of stumbling upon a question in a 2005 Michigan survey in which nearly one-fifth of voters said they believed that their vote choice was not secret. Other national polls showed as many as 27 percent of people shared that view. When they were asked, “How difficult do you think it would be for politicians, union officials, or the people you work for to find out who you voted for, even if you told no one?” only 12 percent said it would be impossible. The finding was consistent across three surveys, and it floored Gerber, who had begun his career by studying the introduction of the secret ballot to American politics in the late nineteenth century. The end of public voting coincided with
a dramatic, and not fully understood, drop in turnout rates. (
The share of the population voting dropped from nearly 80 percent in 1896 to 65 percent eight years later, and never recovered.) Gerber wondered if nonvoters simply didn’t trust, or understand, the idea of confidentiality at the heart of the process. Gerber commissioned his own survey. Among adults who had never voted, 20 percent expected their ballot to be marked so it could be identified as their own, and 12 percent thought that upon arrival at a polling place someone would ask for whom they were voting.

Gerber went on to tell his audience how he had then set up an experiment to measure whether he could bring those skeptics to the polls. Just before election day, he had randomly divided a set of Connecticut voters into five groups. Three of them were sent slightly different letters from the Connecticut secretary of state, each describing how the voting process keeps individual votes secret. A fourth group got a placebo election reminder, and the last control sample got no mail at all. After the election, Gerber saw that his letters had no impact on those who had been to the polls before. But when they reached people who were registered but had never voted, participation spiked: the letters emphasizing ballot secrecy created 2 or 3 new voters for every 100 people who received them.

It was only the latest example of someone using twenty-first-century tools to assimilate lessons from nineteenth-century politics. More and more, those who looked anew at the act of voting were beginning to think of it in altogether different terms. Maybe what stopped people from voting wasn’t a lack of information about the candidates or a feeling that the outcomes of races didn’t matter or a sense that a trip to the polls was inconvenient. What if voting wasn’t only a political act, but a social one that took place in a liminal space between the public and private that had never been well-defined to citizens? What if toying with those expectations was key to turning a person into a voter? What if elections were simply less about shaping people’s opinions than changing their behaviors?

MALCHOW HAD SPENT
a quarter century living off the mail, but not until his revelatory steakhouse dinner with Gerber did he give much thought to envelopes. After he had overcome his initial awe at the power of the social-pressure tool used in Michigan, Malchow looked closely at the mailers themselves. They were simple copy paper, laser-printed and crudely folded, the result of Grebner scrambling to produce them cheaply in his own office rather than hiring a professional copy shop for the job. They looked appropriately amateurish, unlike anything Malchow had put out in his years of sending political mail to raise money, persuade voters, or turn them out. By the traditional standards articulated by direct-mail vendors, valuing high-impact visuals that “cut through the clutter” of the mailbox, Grebner’s bland letters should have been a dud. But, of course, they hadn’t been, and now Malchow began to wonder whether their success owed something not only to psychological tricks but to their humble packaging as well.

One of Grebner’s letters didn’t even try to exercise social pressure, instructing a voter merely, “Remember your rights and responsibilities as a citizen. Remember to vote.” Such generic “civic duty” messages rarely made any impact on turnout, starting with the first Gerber-Green experiment in New Haven. In fact, the only reason they had included it in the Michigan test was as a baseline against which they could measure the various social-pressure effects. Yet in Grebner’s hands the civic-duty message increased turnout by nearly two points over the control group, and the only reason Malchow could find to explain it was the primitive format. He thought about the other pieces of paper that shared those aesthetics: a jury-duty summons, a letter from the taxman, the homeowner’s association announcing a policy change. What if, Malchow wondered, an unstylized simplicity had become a signal at the mailbox that something was to be taken seriously?

So he started testing. He ran experiments pitting letters against glossy brochures, black-and-white against full color, slick against clunky. The evidence piled up, all pointing in the same direction: toward plain, official-looking communications. Others at the Analyst Institute reported experimental findings that seemed to confirm the virtues of simplicity. A
group called Our Oregon, which runs state ballot initiative campaigns for progressive causes, found that it could increase its vote tally in select precincts by five points by replacing its glossy mail with a bland, text-heavy voter guide devoid of endorsements from politicians but instead featuring the validating logos of groups like the PTA and the League of Women Voters. Rock the Vote found that e-mail and text messages arriving from unexciting senders like “Election Center” often do better than those with livelier “from” lines, like the names of celebrities. “If you believe this, it says we’re doing everything wrong,” says Malchow. “There’s a principle underneath this. When people see the fingerprint of Madison Avenue, it becomes advertising—and advertising is not important to them.”

Throughout 2011, Malchow was eager to press these concepts further, and to deploy new modeling tools that combined microtargeting and experimental methods to predict which individual voters would best respond to a given appeal. He and Page Gardner, along with others at Women’s Voices, assembled a list of twenty experiments they wanted to run that year. Malchow conjured a single layered design that would mix and match treatments in different combinations and test their compatibility with proven techniques, such as social pressure. He wanted to tweak Gerber’s ballot secrecy reminder the way he had the social-pressure menace, changing the language or the presentation to see if he could squeeze more new votes out of the electorate. He and Gardner planned to administer the experiments in Kentucky, which would be selecting a governor in the last scheduled election before the 2012 election cycle began, affording enough time to analyze the results and deploy the best new tactics nationwide to boost Obama’s reelection.

But when Gardner went out to pitch donors on these tests, she couldn’t find any willing to sponsor them. This had to do in part with the changing dynamics of the Kentucky race, which for largely local reasons had lost the interest of liberal donors nationally, but more with a changing set of priorities for Democratic strategists. The question of how to most efficiently get large numbers of new voters on the rolls had moved to the top
of their list of concerns. “We want to do some research as we go into 2012,” says Gardner. “What is the most appealing way to make the process of registration easiest for people most underrepresented in our democracy?”

So Malchow shifted his own focus, too. Through weekly Tuesday strategy calls with Women’s Voices staff, Malchow arrived at fourteen variations to the group’s standard voter registration package to test. In some cases he and Gardner decided to fiddle with the format (adding a fake Post-it note to direct a recipient to the fields she needed to complete) and in others made more substantive adjustments (would putting an NAACP return address on the envelope make black Mississippians more likely to register to vote?). Malchow produced each of the test mailings, assigned treatment and control universes across twenty states, and in mid-September, a little more than a year before Obama would reappear on a ballot, postmarked just under a half-million registration forms. “What seems to be a very small difference in response rates … becomes a difference in cost per net vote,” says Gardner. “All these nitty-bitty things have magnified effects.”

Registration was the first step in the process of winning a vote, but it rarely earned sustained attention from campaign operatives. The tax code treated it as a civic function rather than a political one, something that was unquestionably good for democracy rather than a tactic to back one candidate. Because outside groups would take on this work, especially on the left, campaigns rarely did it themselves. To the extent there was expertise in the art of registration, it had belonged to groups like ACORN and its many satellites, who specialized in overwhelming minority neighborhoods with unskilled workers who were typically paid for each form they returned. The workers knew their turf, and would often do little but set up a card table outside a well-trafficked grocery store, with a stack of forms handed out indiscriminately to passersby who said they were unregistered. The going rate for each new registrant was about fifteen dollars, covering labor and administrative costs. Often there wasn’t a computer in sight, or any way of matching the names to the voluminous databases that could help predict what party that voter was likely to join, or—if it was not a first-time
voter but merely someone registering in a new location—document his or her history of political behaviors with great specificity. Groups, parties, or campaigns that had learned to rigorously target their mail and phone communications to maximize votes and lower costs would often mindlessly use their registration programs to put new supporters of their opposition on the electoral rolls at great expense to themselves.

Gardner had long yearned for a more refined approach. Her experience with Women’s Voices had taught her that the group could not just adopt a card table strategy, since unmarried women did not segregate in particular neighborhoods. She also fixated on the obvious inefficiencies in the standard process; lowering the cost per registration meant expanding her program’s reach. So years ago she and Malchow had turned to new commercial databases that covered the entire adult population, pulled out unregistered women who appeared to be unmarried, and started sending out registration packages by mail. Through those tests, he had found he was able to bring the cost down to around eleven dollars per registration.

In part because of this success, Women’s Voices’ contributors pushed Gardner to start hunting for other targets, such as African-Americans and Latinos (regardless of gender or marital status). Malchow built a model to predict which members of this expanded target universe would be most likely to register through a mailed appeal. He refined the statistical models each time, increasing the response rate and bringing down the costs. (Malchow even paid the post office for the mailers it returned as undeliverable, and used their common attributes to develop a model that could predict which addresses were likely to be bad so they could be preemptively removed from future lists.) In 2008, Women’s Voices had sent out nearly 20 million applications, and registered just under 1 million of them in time for Obama’s election. Thanks to Malchow’s modeling and experimentation, they had cost around seven dollars each.