The Victory Lab (45 page)

He intensified his focus on mobile phones, which were prevalent among youth across class lines and often served as a substitute for a traditional computer. In early 2007, Goodstein had convinced the campaign to spend one thousand dollars a month to rent Obama’s surname as a text-message short code, so people could submit information from their phones by typing just 62262 instead of a longer, area-code-specific number. For far less money, Obama could have shared a nondistinctive code with other customers, as Clinton and Edwards had, where messages were routed to a particular destination by a unique keyword. Goodstein successfully argued to the campaign leadership that having a dedicated code would avoid confusion and hassles with the other (usually nonpolitical) customers who would share it. But he eventually realized that it had another benefit: Obama could segregate incoming messages by their content, like those who wrote JOIN from those who wrote GIVE, and track them separately, at no further cost. When he wanted to expand the campaign’s list of mobile numbers, Goodstein invited supporters to text the word
Go
to 62262 to win a sticker that read “Got Hope?” When the campaign bought thirty seconds of Super Bowl time in targeted states it used a different code (“text
Hope
”) to measure the response of the unconventional ad placement.

Still, Goodstein was disturbed by how rare it was for the campaign to embrace such novelty in its pursuit of a coveted yet elusive demographic group. In June, Obama’s media team developed plans for a $30 million television and radio buy targeting young voters. It struck some of those already skeptical of the campaign’s youth-marketing efforts as an old person’s idea of a young person’s ad buy. Most of the spots would go on MTV and Comedy Central, and the aesthetics seemed too conventional to reach nontraditional voters. “What’s the difference?” Moffo asked anyone who would listen. “It’s the same old ad but has some kids in it and rock-and-roll music.” Field director Jon Carson convinced Grisolano to hold off on the buy as he looked for ways to persuade the campaign’s leadership to spend the money differently. He realized that there were no bounds to what the campaign would be willing to fund; Organ was messing around with placing ads in movie theaters and kiosks in shopping malls as part of an effort to expand Obama’s reach through what the industry called out-of-home advertising. But Carson also knew that if he was going to help Goodstein and Organ commandeer $30 million (or even a chunk of it) toward heretofore unproven communications techniques, they were going to have to devise new empirical methods to prove their effectiveness.

Goodstein’s ability to compare text-message response rates offered the metric they needed to redirect the $30 million in the youth-marketing budget. He and Carson began toying with a proposal: before committing the money to TV and radio, how about spending less than 1 percent of it on a field experiment? Moffo put the details into a Jigsaw-logoed presentation, describing a test that could be conducted in August, before the conventions, so that the findings could dictate how money would be spent in the fall. The campaign would budget thirty thousand dollars each to a series of different treatment groups, assigning every medium its own text-message keyword and unique 800 number. They could measure responses-per-dollar-spent, so that the effectiveness of TV ads could be fairly pitted head-to-head against less traditional media. None of the media would be personally targeted, so there wasn’t a way to randomize their assignment
individually. Separating the buys geographically and comparing Obama’s movement in polls across various areas wouldn’t help much, either, since the goal wasn’t persuading undecideds but rather mobilizing the type of natural Obama supporters the campaign needed to turn out.

Grisolano signed off on the experiment. Now Moffo and Goodstein needed a place to run it. Moffo turned to Ohio, in part because the campaign was less concerned with registering new voters there but understood the value of turning out Obama’s base. But most of the reason was the state’s leadership: Bird was known as someone who was eager to try new things, and had earned the loyalty of local field staffers, who would have to be counted on to execute some of the nontraditional parts of the experiment.

Bird helped to isolate two Ohio cities reflecting the two major types of young populations Obama was targeting: Columbus, the state’s dominant college town, and Cleveland, with its permanent population of young minorities. In each market, Organ would buy cable TV and hip-hop radio ads, as per Grisolano’s original plan, and online and mobile-app ads, which could be restricted by geography and specific categories of digital content. Goodstein placed ads in student newspapers and alt-weeklies, as well as on outdoor billboards in carefully selected neighborhoods, and assigned “street teams” of field volunteers to heavily trafficked areas, where they would blanket available surfaces with posters and the polypropylene sheets known as ClingZ. Every medium featured had a “Get a Free Ohio for Obama Sticker” appeal, redeemable by text-messaging 62262. Unlike most candidates, Obama had only rarely given away basic items like stickers; the campaign preferred to sell them, which allowed them to count purchasers as donors and pad their contributor reports. Even so, the prize for the Ohio experiment was something that the campaign thought was appealing enough to lure passive consumers to turn to their phones and send a text message.

As soon as the test was scheduled, Goodstein contacted Shepard Fairey, the street artist who had previously contributed an instantly iconic portrait of a posterized Obama with the word
Hope
. Goodstein wanted
permission to use a lesser-known image Fairey had developed specifically for the campaign, with a serious Obama shown in profile above the word
Change
, on the stickers the campaign would be giving away. After a few tweaks, Fairey signed off on the request, and Goodstein had the stickers printed. (Later in the fall, Moffo would prepare a $1 million poster campaign and, relying on internal opinion research that Obama could have an “angry black man” image problem, conclude that Fairey’s “Change” likeness was “too severe for GOTV.” It became Goodstein’s job to tell the artist that the campaign was retiring his famous portrait; Fairey quickly designed a third Obama “Vote” portrait that showed the candidate smiling.)

As it got under way, the experiment became a bit of a bake-off; Goodstein relayed text-message responses hourly, with staffers rooting for their favored mode of contact. When they found that online ads and street teams did best, just about everyone involved in the project exulted. Grisolano erased the television line-item from the youth-advertising budget altogether. Moffo and Goodstein, for their parts, felt as though they could claim victory in a broad, generational battle. They had faced down what had become an unjustified obligation for presidential candidates to mindlessly spend money on television; they had unmasked the medium’s inefficiency with new technology and analytics. Only by discrediting establishment tactics, Moffo thought, could politicians truly liberate themselves from establishment money and institutions. “If you can figure out how to do it better, you don’t need progressive interest groups or organized labor,” he says. “We all had this sense on the Obama campaign that we’re changing how politics is done.”

UNLIKE SOME CAMPAIGNS
, which held their public-opinion research closely among top advisers, the Obama leadership distributed all of the polls they conducted to state-level staff, a degree of openness that became a point of pride for the entire operation. There was even greater
satisfaction that not a single number from any of these internal surveys ever leaked, a testament to the “no drama” camaraderie that kept staffers all focused on their common goal rather than pushing personal agendas through the media. So it was a bit out of institutional character that the campaign leadership signed off on a proposal to share all of its private polling with Nate Silver, the amateur statistician whose startlingly accurate electoral forecasts on the blog
Daily Kos
had won him acclaim during the primaries. When Silver launched his own blog,
FiveThirtyEight
, his sympathies for Obama were no secret to readers—even as he claimed his partisan loyalties had no effect on his statistical forecasts—but it was his data-centric approach and skepticism about the existing polling-media complex that won him an obsessive following in Chicago. (When one of Silver’s correspondents reported from an Obama field office in Big Stone Gap, Virginia, that 92 percent of the state’s neighborhood team-leader slots had been filled,
Bird wrote in to say that in Ohio they had reached 93 percent.) Silver had developed a forecasting model that mixed state-level demographic information and available public polls weighted according to their prior reliability, and Obama’s targeters wanted to see what would happen if their internal polls were subjected to his formula, as well. After securing a confidentiality agreement from Silver, the Obama campaign gave him access to hundreds of polls it had conducted. “We wanted a little external validation that what we were seeing is what was actually going on,” says Simon.

The Obama campaign was blessed with good fortune that demonstrated itself so quietly that its once improbable cause now looks like a matter of historical inevitability. There were a few down moments—a surprise loss in New Hampshire, the emergence of Obama’s provocative pastor, the Reverend Jeremiah Wright—but the bigger indicators, like polls and fund-raising figures, almost always pointed up. The campaign’s obsessive reliance on data and metrics usually had a calming effect: reality was demonstrable, and not up for debate. Once the system that created the numbers earned people’s loyalty, their faith in what the numbers represented naturally followed. But that almost monomaniacal confidence in
the ability to accurately measure the vagaries of a presidential campaign tugged a persistent worry in its wake. If one thing was wrong in the ever-complex math undergirding the Obama enterprise—a shoddy source of information entered into the databases, a misplaced coefficient in the algorithm working through them, or an intellectually faulty assumption about how they fit together—an entire chain of tactical and strategic calculations could be flawed.

Such unease dogged the campaign leadership throughout. On the eve of the Iowa caucuses, Plouffe pulled Strasma aside in Des Moines and asked, “So are we going to win?”

“Yes,” Strasma replied. “Unless everyone’s lying to us.”

There was one obvious area where people could be lying. There had never before been a credible black candidate for the presidency. But past opportunities for white voters to cross the color line had produced evidence that they were bashful about acknowledging their hesitation to do so. Throughout the 1980s, as the first black candidates sought major city and state offices, observers noted that polls showing them with large leads seemed to dissolve into lean victories or losses when election day arrived. This became known as “the Bradley effect,” after Los Angeles mayor Tom Bradley, a black Democrat who led his white opponent, George Deukmejian, in polls by as many as 15 points in the 1982 California gubernatorial race, only to be defeated by a small margin.
One political scientist’s later analysis showed that African-Americans running statewide in the late 1980s and early 1990s performed 2.7 percentage points worse than pre-election polls had predicted they would, and white candidates running against blacks began to count on that buffer of latent support to carry them through. “
It’s just a fact of life,” said Bill Roberts, Deukmejian’s campaign manager. “If people are going to vote that way, they are certainly not going to announce it for a survey taker.”

Early on, there was no suggestion that voters were misleading either Obama’s canvassers or public pollsters. “The white Iowa pig farmers were perfectly willing to vote for the African-American,” says Strasma. So were
their peers in Idaho, Maine, and Wyoming; in fact, Obama ran up some of his most formidable margins in the country’s whitest states. But those wins had come in caucuses, which drew from populations of committed party activists, who could be rightfully expected to be more liberal than the general electorate, including on issues of race. Perhaps more crucially, Iowa caucus-goers announced their support in public, which meant that the disconnect of the Bradley effect—between what someone is willing to say aloud to a stranger and what he or she does behind the voting curtain—wouldn’t apply.

But in later primary states with greater assimilation, such as Pennsylvania, Strasma sensed that racial attitudes might be pulling Obama down. Unusual patterns “would just bubble up in the models,” says Strasma, allowing him to spot specific variables that were exerting an undue influence in predicting candidate preference. Strasma had noticed that high-income white voters in largely black neighborhoods had very high Obama-support scores, while low-income whites in comparable neighborhoods were far less likely to support the black candidate. “If you’re high-income and racist, and your neighborhood changes, you move,” Strasma theorizes. “If you’re low-income and racist, the neighborhood changes and you can’t move, you get more resentful and are less likely to vote for Obama.”