The Numerati
Page 8
This is a tall order. How will Gotbaum, with his modest funding, dig down deep within you and me and masses of other voters, and scope out our philosophical moorings? Even if he talks to 1,000 people, or 10,000, how can he extend what he learns to chart the political currents of an entire nation? And if most of us don’t feel like thinking or talking about politics, what kind of data will express our unspoken (and often unknown) political convictions?
Gotbaum smiles and takes me back to late 2005. “It was pretty clear,” he says, “that Republicans had been spending money to understand the independent voters. I got back from Hawaii, and I thought, ‘Corporations spend a zillion dollars on market research. Why can’t that research be made available to Democrats?’ That was my notion.” So early in 2006 he called in two market researchers and a pair of political pollsters, and he dispatched them to survey the values of Americans.
The raw material for this type of research—the details that make up our lives—is collected and sold by a fast-growing group of data companies. ChoicePoint, headquartered in the Atlanta suburb of Alpharetta, Georgia, is a prime example. The company quietly amasses court rulings, tax and real estate transactions, birth and death notices. Many of these records have existed for centuries in file cabinets and courthouse ledgers. But ChoicePoint employs an army of data collectors to harvest those facts, sometimes writing them down by hand. Then they put them into digital files.
Files that used to exist on different pieces of paper in different buildings can now be brought together. Our profiles start to take shape, and they can glide around the world on networks. Human resources managers dig up our Choice-Point files to see if we’re lying on our résumés or have neglected to mention that year of hard time on Rikers Island. While ChoicePoint sticks to identity data, plenty of other data companies, as Robert O’Harrow Jr. writes in his authoritative book No Place to Hide, add flesh and blood to those bare bones. One of the largest, Acxiom of Conway, Arkansas, keeps shopping and lifestyle data on some 200 million Americans—nearly every adult in the country. Acxiom knows how much we paid for our house, what magazines we subscribe to, and which book we bought two days before taking that trip to Club Med in the Alps. The company buys just about every bit of data about us that is sold, and then they sell selections of it to anyone out to target us in a campaign.
Those companies provided data, but Gotbaum needed someone to turn it into something useful—a tool to find pockets of swing voters. Enter J. Walker Smith, president of Yankelovich. Since the early 1970s, this research company, based in Chapel Hill, North Carolina, has been surveying the shifting attitudes of the American public. It compiles these trends in a report called Monitor, whose motto says it all: “What in the world are they thinking?” It was this type of research, Gotbaum believed, that would enable him to create a new type of political profile—to locate millions of potential Democratic voters.
At the heart of Monitor are three sets of questions, Smith tells me, from his offices in Atlanta. First, what does the future hold? Does it look promising? Scary? Exciting? Bleak? The second question centers on how people will fare in that future. How do they define success? A healthy family life? Making piles of money? Grappling ever upward in the business world? Gaining respect as a good neighbor, an upstanding member of the community? The third question focuses on the skills and assets people believe they will need to achieve success in the coming years and decades. These questions, vague as they sound, get to most of the issues that have us gnawing on our fingernails by day and tossing and turning at night: What’s ahead, both good and bad? Where do we want our lives to take us? How can we make this happen? Marketers in the auto and travel industries find these studies useful in deciding which features to promote in their products. If the public mood in the Monitor takes a skittish turn, for example, a carmaker might emphasize antilock brakes and a chassis that can withstand a bruising encounter with an oak tree.
“We suggested to Josh and his team at Spotlight that these values could be used to profile voters,” Smith says. The idea was not just to identify voters in a key state or congressional district but to group every single voter in the country according to a set of values. Gotbaum was on board from the start. But others on the team feared that a consumer survey full of vague “values” questions would lead to a political dead end. Smith himself couldn’t guarantee it would work. “We wouldn’t know until we tried it,” he says.
So they went ahead. In a marathon session in early 2006, they drew up a survey that mixed the traditional Yankelovich battery with lots of other questions. They asked, for example, if “taking care of our country’s children” should be the number one priority, if the country should do “whatever it takes” to protect the environment, if we need to “rebuild and strengthen our shared vision” on what it means to be the American community. They also planned to ask voters which party they had voted for in recent elections. In the end, they figured it would take an average voter about 35 minutes to answer the 140 questions on the survey. This was a lot to ask, but Gotbaum trusted that even people who hang up on political pollsters and other phone spammers would take time to respond to questions about their personal values. “People like to talk about themselves,” he says.
The next step was to choose 3,000 people to participate in the survey. These would become the test group, the gold standard. Interviewing these 3,000 would be the painstaking work of humans. Computers would later build upon the patterns culled from those 3,000 to classify, at warp speed, the rest of us. Naturally, this group had to reflect a cross section of voters. Gotbaum made sure that each one of them was already represented on Yankelovich’s database of American consumers. That way, his team would have fat dossiers on every person surveyed. They would be able to slice and dice the group every which way. If those most concerned about taking care of the country’s children appeared different in any way—if they lived in certain types of neighborhoods, went to church more often than others, even if they drove Pontiacs or ate sushi in disproportionate numbers—Gotbaum wanted to know about it. There would be no blank slates in this sample. His pollsters would start with telephone interviews and then reach as many of those people as possible for follow-up questions via the Internet. (This idea also generated heated arguments, Gotbaum says, because it’s hard to find a representative population on the Net. Those responding to Internet surveys are likely to be slightly richer, younger, and better educated than the voting population as a whole. But small start-ups like Spotlight take such shortcuts to save money and later adjust the results for the bias.)
By summertime, the surveyors had conducted their interviews. The Spotlight team was sitting on reams of new data. Gotbaum asked Yankelovich and one more corporate research firm, StrataMark, to analyze the answers about values. “I told them to ignore politics,” he says. “Just tell us, if you were going to segment this population on attitudes and values, what would the segments look like?”
Both researchers returned a month later with parallel results. Each of them concluded that voters fell into six clear values categories. They tossed out the smallest one, the 29 voters who appeared to be either angry or alienated from the political process. No sense in building campaigns for them. That left five groups. As Smith describes them, they were focused around these values:
1. Extending opportunity to others
2. Working within a community
3. Achieving independence
4. Focusing on family
5. Defending righteousness
It doesn’t take an advanced degree in political science to guess which way a couple of these groups would vote. “Extending opportunity” sounds like a code word for bleeding-heart Democrat. And that righteous fifth group appears unmistakably conservative.
Gotbaum saw as much himself. He hardly bothered with the extremes. They were firmly in the Democratic and Republican camps. Interestingly, the core Democratic group, with 37 percent of the total, was more than twice the size of its Republican counterpart, which only
had 16 percent. This meant that Republicans were managing to attract many more of the middle segments, the swing voters. The three swing sectors, grouped around community, independence, and family, accounted for 47 percent of the electorate. These were the voters who guided the country from right to left and back again, from Reagan to Clinton to Bush, to . . . well, whichever candidate figured out how to connect with them. Some of these people saw government as a potential ally; others viewed it with suspicion. They differed greatly on religion. And yet lots of voters in each of these groups were clearly open to persuasion. What moved them? To win them over, Gotbaum believed, Democrats had to appeal to their core values: community, independence, and family. If he and his team could just pinpoint thousands of these voters in a state or congressional race, the Democrats could craft custom-made messages for each group. The way Gotbaum planned it, the Spotlight system would locate perhaps 20,000 or 30,000 members of a certain value group in a tight congressional race. If research showed that many in this group listened to a certain gospel radio station or perhaps watched a cooking show on cable TV, the campaign would reach them with a targeted ad. A more precise approach would be to bombard each of the voters with carefully calibrated letters and brochures. Gotbaum didn’t yet know exactly what those ads would focus on. For that, he needed to slice the data more thinly.
The Spotlight team studied each voter’s views to see how deeply they cared about these issues of politics and values. Some in each group were heavily committed, hard-core, while others appeared to focus more on other aspects of life. Still working with their original test group, Spotlight split each of the five segments in two, one group that was more committed, one that was less. The more committed voters might be harder to swing. But they were also more likely to pay attention, and to vote. Moreover, the commitment, especially in the middle groups, didn’t necessarily signal unswerving loyalty to one party or the other but instead to particular values. Following the traditions of consumer marketers, Spotlight gave each of these ten groups a descriptive name, such as “Barn Raisers,” “Hearth Keepers,” or “Inner Compass.” As Gotbaum tells it, these are tribes. All of us are members of these values tribes (and countless others) without even knowing it. These tribes of ours have no logos, no history, no home turf, no particular cuisine or religious leaning. They span all races and ethnic groups. In that sense, they’re a bit like those buckets of broccoli eaters or Mars Bars buyers in the supermarkets.
I stop and try to think of anyone I know who might be a “Hearth Keeper.” Such people, according to the Spotlight blurb, focus on family and faith. But they resent attempts to politicize these values. They’re the less committed of the independence segment, not quite as rambunctious as the Barn Raisers. Looking at a color-coded chart, I see that Barn Raisers tend to be more entrepreneurial and mingle more within their communities, while Hearth Keepers, as their name would indicate, focus more on family satisfaction. Both groups adhere to “faith-based living” and lean Republican. But Hearth Keepers are more likely to keep it to themselves, and they resist “marketing intrusions into their private lives.” (They don’t sound like the most promising group for one of Gotbaum’s microtargeted campaign calls. “I know you don’t like this kind of call,” the volunteer’s prompt might read. “But this one is different . . .”)
I look around the coffeehouse where I’m writing. Any Hearth Keepers sipping cappuccinos around here? The college student to my right—shaggy hair, flannel shirt, anthro books on the table—has his feet stretched out on a chair. He’s wearing black low-cut Keds. If this is a Hearth Keeper, he’s hiding it. Next to him, a middle-aged man with horn-rimmed glasses is just getting settled. He wears a neat V-neck blue sweater vest over a freshly ironed powder-blue shirt. Could be, I think. It’s Sunday, he may have come straight from church . . . But shouldn’t he be at home, with his family? Now he opens an Apple laptop (exactly like mine), and I’m quickly revising my judgment. I could be in the presence of a techno-friendly Right Click, another Republican-leaning tribe. (These people flocked in 1992 to the independent candidacy of H. Ross Perot, a former high-tech entrepreneur. Their quieter cousins, the Civic Sentries, feel less economically alienated than the Right Clicks but worry more about safety and economic security.) Considering where we’re sitting, in the liberal New Jersey enclave of Montclair, this man could conceivably be a moderate Inner Compass—someone who insists on fitness, both in body and balance sheet. He looks pretty buff, although with sweaters like that you never can tell . . . In the same swing group are those who care more about career satisfaction and material success: the Crossing Guards. Could it be? In the end, I can speculate until my coffee goes cold. If this were a fashion or technology study, my observations would inform me. I could document my conclusions. But to divine the political views of others—even those pecking away at laptops a mere eight feet away—is impossible. That is Gotbaum’s refrain. Voters don’t wear uniforms. Unless I go over to his table and lead him through the Spotlight questionnaire, there’s no way of knowing which group he belongs to.
One blustery afternoon in New York, I ask Gotbaum if I can take the test. He’s a little uneasy about it, worried that my knowledge of the process will skew the results. But he finally relents, and the next day he e-mails me a questionnaire. This isn’t the 30-minute version, I’m disappointed to see, but a boiled-down sampler. In the course of their research, Gotbaum explains, the Spotlight team found 17 questions that predict, with 92 percent accuracy, how people would answer the others. So they cut to the chase. It takes me about five minutes to answer the multiple-choice questions about religion, schools, and my feelings about my neighborhood and country. When I get back the results, a day later, I learn that I’m a “Still Water.” This is the less committed wing of the core Democratic group. It’s the largest of the ten tribes and represents 19 percent of the voting population. Although 87 percent of the Still Waters identify with Democrats, they’re “independent-minded.” I read on. They see “an affirmative role for government” but stand back from the “political vanguard.” The firebrand allies of the Still Waters are the Resourcefuls. They have less esteem than Still Waters for entrepreneurs. Neither group shows much interest in faith-based living.
I tell Gotbaum that he’s not wildly off the mark. But as I point out, I went to him and volunteered to be analyzed. Most voters wouldn’t dream of doing this. How can he determine which tribes the rest of the electorate fits into? How can he find, say, 10,000 Barn Raisers in Tuscaloosa or Duluth? Until he organizes the entire population into tribes, his political clients won’t know where to start.
This is where consumer files came in handy, Gotbaum says. To place all of us in our groupings (without having to knock on millions of doors), the Spotlight team dredged Yankelovich’s vast database of 175 million consumers. (That’s 33 million more than voted in the 2004 elections.) Throughout the summer of 2006, statisticians and data miners searched for patterns within the files that would help slot practically the entire voting population of the United States (along with millions of nonvoters) into the ten political tribes.
To build our political profiles and sort us into tribes, they harvested loads of tidbits with only the most tenuous relation to politics. These are known as proxies, or stand-ins. They’re what statisticians rely on when they don’t have the answers. Here’s an example. Imagine that you’re catering a wedding party and—horrors!—you’ve lost all the forms the guests filled out with their dinner selections. You’re too proud or scared to take the simple route and ask people whether they prefer the sausages or the vegetarian fricassee. So you study the people and look for signs that might link each of them with one dish or the other. One man is telling loud jokes with what sounds like a strong Wisconsin accent. Wisconsinites, with their German heritage, eat loads of bratwurst, so you give him the sausage. You give the vegetarian dish to the skinny women, men with ponytails, a guy with a save the whales button on his lapel. Anyone who’s overweight or drinking beer instead of wine gets th
e sausage. Judging people by these proxies is clumsy. Yet it’s how most of us think. We analyze the patterns we know (or believe) to draw conclusions about others. The crudest of these form the foundations of prejudice and racism. Often they’re wrong, or unfair. At this wedding party, quite a few of the guests will inevitably get the wrong dish. However, if you’re smart in picking out these indicators, you’re right more often than you’re wrong. And if, like Josh Gotbaum, you have hundreds of data points on each person and a team of statisticians sifting through them, you can begin to place millions of us into tribes. (Usually we end up being grouped with others in our household, since most of our details, from the size of the mortgage to a Field & Stream subscription, are shared.)
So which nuggets of my data will expose me (and my wife) as Still Waters? Spotlight’s statisticians can delve into all kinds of details. Cat owners, Smith tells me, are more likely to be Democrats. (We have two cats.) Republicans trend toward dogs. (We’re down to zero, though I’m not happy about it.) Still Waters are more likely than their more ardent cousins, the Resourcefuls, to be college educated and married, with children living at home. (Yes, yes, yes.) We Still Waters show more interest in cooking than other groups do. A subscription to a gourmet magazine, Smith tells me, would help to identify me. (No such luck.) These details about pets, kids, cooking, and college education are proxies. They may have statistical relevance, but they’re a far cry from evidence. On the other side of the scale, party registration is a clear political statement. (Like many independent-minded Still Waters, I don’t belong to either party.) An even more compelling signal is the record of a financial donation to a political candidate. (My wife gives. If her donations are linked to me, I’m implicated, just as my father was by my mother’s bumper sticker.) Reports of personal donations provide hard data. They leave behind the nebulous world of proxies. It’s as if the wedding caterer saw the loud midwesterner wolfing down a plate of frankfurter hors d’oeuvres. Once the diner has established himself as a sausage eater, his accent and other statistical proxies are beside the point (though it’s still possible that he ordered the vegetable dish).