Victory Lab

Anatomy of a Narrow Victory

Five years in the making, Romney’s Iowa win took a lot more than money.

Mitt Romney in Iowa.
What went into Romney’s Iowa win?

Photograph by Chip Somodevilla/Getty Images.

DES MOINES—On Monday night, 24 hours before Iowans would participate in their state’s caucuses, around 25 volunteers sat in an old Blockbuster Video and placed calls from their personal cellphones on behalf of Mitt Romney. They had the trappings of a Romney crowd—oxford shirts, small talk concerning fruit salad—and the names of the voters on their list were of Romney people, too. “The reason I’m calling is because I have you down as a supporter,” the callers chimed, reading off a get-out-the-vote script.

This fluid caucus season has birthed and vanquished new front-runners—Gallup calls it the “most volatile” nominating contest ever polled—but there has been one constant in Iowa: a list of more than 30,000 supporters that Romney’s team believed were unlikely to vote for any of his rivals. The process that produced this list had taken nearly five years to complete. Romney’s previous Iowa campaign allowed him to stockpile voter data and develop sophisticated systems for interpreting it. It was that data and those interpretations that supported one of the riskiest strategic moves of the campaign thus far: Romney’s seemingly late decision to fight aggressively for his first-place finish in Iowa.

Even as his campaign leadership claimed into the fall that they were keeping their options open here, Romney’s targeters were quietly maintaining a continuous tally of their supporters in Iowa, a list that proved unexpectedly stable even as other candidates rose and fell in the polls. It had become a stock observation to note that Mitt Romney just couldn’t move from 25 percent in Iowa—his support was both resistant to growth and impervious to decay. But what was more important for Romney’s team was not just that his total share of the vote remained steady but that the individual voters who comprised it didn’t move either, making it easy to keep track of who they were and to mobilize them personally.

It was the ability to pinpoint and track supporters that settled Romney’s decision to publicly commit to winning Iowa late this fall. Romney’s campaign made a big show of converting the former video store into a headquarters, while spending millions on local television ads and dispatching the candidate to travel the state more aggressively than he had. But a ruthless yet largely invisible strategy had already been in place for much of the year, tracking both Romney’s supporters and his opponents. Only when Romney’s count appeared to exceed any rival’s did advisers unveil the trappings of a traditional caucus campaign.

“We had a fair picture of the people who were sticking with us,” says David Kochel, Romney’s top Iowa strategist. “It is a significant enough share to give us good confidence that we would have a respectable base.”

In the end, Romney was able to carry the caucuses without assembling much of the heavy apparatus the media typically validate as a “ground game” or “organization”—the congeries of campaign offices, phone banks, field staffers, and phalanxes of volunteers—whose primary objective for much of the pre-caucus season is to maintain a dynamic census of potential caucus-goers. Romney maintained that census through statistical work and new technology, all built on having survived a dry run four years ago in which a profligate Romney spent $10 million on an embarrassing second-place finish.

Confident in what his data was telling him about his chances in the caucuses, Romney spent his final days in the state this year quoting from patriotic songs while rarely drawing any contrasts with his opponents.

“We weren’t as concerned with persuasion as we were last time,” says Michael Meyers, a data analyst who with his partner Alexander Gage has advised Romney for the past decade. “We spent a lot more time ensuring we could turn our people out.”

Meyers and Gage first worked with Romney on his 2002 campaign for governor of Massachusetts. The two Michigan Republican operatives had grown frustrated with campaign targeting that was restricted by the limited individual information available on the electoral rolls (party registration and vote history) and historical tallies at the precinct level, where actual results are available. Gage noticed that commercial marketers and credit-scorers had begun to organize reams of demographic and consumer data—from information on education levels to who had a hunting license—that made it possible to profile an individual across hundreds of variables at once. A former pollster, he designed large-scale surveys that would allow him to tether those individual profiles to topical political attitudes to reveal patterns in electoral behavior. Advances in computing power had made it possible to manipulate tens of millions of those records at once, and Meyers and Gage let algorithms find relationships between them. Their approach allowed them to analyze voters with far more nuance than had previously been possible. When Gage visited Romney’s Cambridge campaign headquarters with a PowerPoint presentation describing his untested method, a former venture capitalist serving as Romney’s deputy campaign manager spoke up. “You mean you don’t do this in politics?”

Romney won that campaign, aided by Gage’s ability to pick out Massachusetts independents and Democrats who would be receptive to the candidate’s positions on specific issues like taxes and education. That success, along with similar projects for Republican tickets in two other states, helped Meyers and Gage win a lucrative contract to perform such “microtargeting”—as Gage successfully branded the technique—for George W. Bush’s re-election campaign. Bush’s advisers wanted to identify segments of nontraditional Republican voters, like Latino women or church-going African-Americans, who could be pulled over to Bush with targeted appeals. In 2007, Gage’s firm, TargetPoint Consulting, signed on with Romney’s first presidential campaign, and set to work finding friendly caucus-goers in Iowa.

Predicting candidate support in primaries is incredibly difficult. In general elections, party registration and a precinct’s historical voting behavior offer a reliable guide for individual support, but primaries have little such structure. The candidates are ideologically similar, few self-evident demographic distinctions separate their coalitions, and historical precedents are usually too messy to be helpful. Particularly in Iowa, where Republican caucus turnout had always been under 100,000—and the state party will sell a list of those past attendees to candidates for $35,000—campaigns have traditionally set out to identify the entire universe of likely caucus-goers and ask them individually whom they support and what they care about.

In the spring of 2007, Meyers and Gage commissioned a large-sample survey of Iowa Republicans to match voter attributes with topical political opinions so that algorithms could look for unexpected relationships between them. At the time, Romney wasn’t well-known, so the campaign saw plenty of room to grow the candidate’s support. To find promising targets among the uncommitted, voters in the polling sample were asked for the issue that most concerned them. Romney’s advisers then cherry-picked a few issues that spoke either to one of their strengths or to an opponent’s weakness—jobs and the economy on one hand, gay marriage or guns on the other—to create a statistical profile of the voters who shared those concerns. Then algorithms searched an Iowa database TargetPoint had built to pick out other voters in the state who shared a common profile and appeared targetable on those same issues.

With the idea that a win in Iowa would propel him on to victory in New Hampshire, Romney had built a formidable machine in the state to count voters. The campaign hired consultants and 10 regional field directors, who identified their most active volunteers and bestowed upon them official campaign titles and stipends of $500 per month. “People in Iowa were starting to get paid for things they were used to doing for free,” says David Roederer, a state official who chaired John McCain’s Iowa campaign in 2008 and remained neutral this time.

Romney’s Iowa staff triaged the electorate based on their micro-targeting research. Republicans who showed no interest in caucusing—or looked unlikely to back their candidate—were pushed off Romney’s lists for good. Those who appeared persuadable received flights of mail and phone calls on the issues they were predicted to care about. Those who declared themselves Romney supporters were offered a trip to the Ames straw poll in August.

Romney’s staff presided over an aggressive effort to both monitor the supporters and keep nudging the persuadables. Field organizers managed phone banks where volunteers would place calls to identify voters’ leanings (known as ID calls), registering the results onto bubble sheets by hand and then scanning them digitally. Volunteers would patrol the doors of Romney’s events to ensure attendees filled out cards identifying them as supporters. All of that information was fed back into TargetPoint’s databases to refine the quality of its predictions. When Meyers and Gage conducted another round of survey calls in the fall, they found a growing base of Romney supporters. The microtargeting algorithms picked out other Iowans who looked like those supporters, generating new lists of promising targets for volunteers to call.

By caucus day four years ago, Romney’s team had a count of more than 50,000 voters, compiled through either individual contacts or through microtargeting models. Campaign strategists didn’t think they’d ever get that many votes: Some of those voters had been last reached months earlier, and might have moved on to other candidates. But Romney’s advisers dramatically underestimated turnout, which hit a record of nearly 120,000 in large part to a wave of irregular caucus-goers pulled in by the winner, Mike Huckabee. “We got the number we thought we were going to get. The other guy got more,” says Meyers, who concedes an underestimation of the ability of a candidate to change the electorate as a major tactical error in the 2008 caucuses. “There were more people we could have gotten out that we didn’t.”

***

Last May, Romney visited Iowa for a day of meetings and events, his first trip to the state during the 2012 campaign. Media coverage fixated on the idea that Romney had retreated from a state where he had sunk millions on a tragic second-place finish in 2007, and had yet to decide whether he would ever muster the wherewithal to try again. “This year, there are no commercials, no bulging payroll, and no headquarters at all,” The New York Times reported. “And he has yet to signal whether he will treat Iowa with deference (as he did in 2008) or indifference (as some advisers have urged him to do).”

But a crowd showed up at a Cedar Rapids barn for an evening picnic nonetheless. “We all looked at each other and said, ‘Oh, you’re still on Team Romney?’ ” says Renee Schulte, a mental-health therapist who had been one of Romney’s four paid volunteers in Linn County in 2008. She has since been elected to the state legislature and named one of Romney’s state co-chairs for the second race. As she reunited with her fellow Romney boosters, she realized that many had been lured there because of a targeted activity invisible to media coverage: a robocall to the thousands of people the campaign had identified as supporters in Linn and Johnson counties in 2008, offering an invitation to reconnect with the erstwhile candidate.

Schulte began reassembling her team of local volunteers. This time there were no titles or stipends, and there was little need for phone banks or offices. A two-person campaign staff supplied names and numbers of supporters who had been identified or modeled as supporters in 2008. And this time, volunteers didn’t have to leave their homes to call them and log their responses: A web-based tool automatically registered voters’ responses with no need for scanners or staffers to manage a paper flow. They called former Romney supporters, and realized—despite a reordered set of national priorities, a different field of opponents, and a change in posture on Romney’s part from culture warrior to corporate technocrat—voter sympathies had not much shifted. “As we started ID’ing people,” says Schulte, “we realized there weren’t a lot of people who had moved.”

Meyers and Gage had noticed something similar from their office in Virginia, along with a small team that had been convened at Romney’s Boston headquarters to bring some of the targeting and analytics work in-house. They were launching regular flights of thousands of automated ID calls, recordings that invited recipients to signal their support for a candidate, likelihood of voting, and issue concerns. The automated ID calls fed back into the microtargeting database, and gave the algorithms updated information about where voters stood. The variables that were playing a crucial role in predicting who would support Romney—higher income and education levels, living in the suburbs, subscribing to magazines that suggested an interest in culture and history—were pointing volunteer callers to the same type of people they had targeted last time. Romney people were still Romney people.

The campaign was so confident that its 2008 predictions, updated for 2012, successfully tracked Romney’s supporters that it didn’t even spend $35,000 for the state party’s caucus list, usually the standard guide to finding likely voters. Instead, as they looked for ways to expand their pool of targets, Romney’s advisers turned to a precinct analysis of the 2010 Iowa gubernatorial primary. Former governor Terry Branstad had defeated religious-conservative activist Bob Vander Plaats, a race whose structure —a pro-business establishment moderate against an insurgency fueled by religious fervor and Tea Party enthusiasms—resembled the one Romney’s advisers anticipated two years later. Romney’s team identified strong Branstad areas and made lists of Republicans who turned out for the primary but had skipped the 2008 caucuses. Those voters became Romney persuasion targets, too, and the microtargeting algorithms started working on figuring out what issues would most drive them.

Romney advisers looked on uneasily as opponents rose and fell to challenge their candidate’s place at the front of the caucus. Their microtargeting calls were also building models of Iowans who looked likely to support each of Romney’s opponents. When Romney was the second choice of another candidate’s soft supporters—as was often the case with backers of Gingrich and Perry—his campaign gingerly added them as persuasion targets, too.

There were two candidates, however, whose supporters Romney’s team would not be able to clearly measure. The targeters grouped together Michele Bachmann and Rick Santorum because “there really wasn’t any data to back up which if either of them would be players at the end,” as a Romney adviser put it. Besides, their voters were coming from the same demographic pool, and seemed likely to move only from one to the other. “We try to be pretty cognizant of who we might be able to peel off,” says Meyers. “We weren’t too interested in turning a Bachmann voter into a Santorum voter or vice versa.”

Instead of trying to win over potential Romney voters with broadcast or online ads, conspicuous direct mail, or cultivating media coverage, the campaign used a new tool to narrowly target potential 2012 voters. So-called tele-town halls would ring an individual voter’s phone with a recorded message inviting him or her to participate in a conference call with the candidate. When a voter chose to participate, an automated prompt would ask for the same information that would be solicited by volunteer ID calls: who a voter supported, how likely they were to caucus. Romney’s team was able to put together different universes for each. Supporters would be invited to “friends and allies” calls (in one, Romney assuaged those made uneasy about attacks on his health care record). Persuasion targets who were modeled to care about the economy were invited to hear Romney introduce his jobs plan; those concerned about immigration were alerted to a call with Arizona sheriff Paul Babeu, a Romney surrogate.

The tele-town halls proved popular—often tens of thousands of voters would listen in—and, at only pennies each, fused a persuasive medium like a radio ad or a candidate visit with the ability of automated survey calls to measure response. Most of the Republican caucus campaigns used these tele-town halls to inexpensively reach voters spread out geographically, but they had a particular value to Romney as he tried to add voters who hadn’t been with him in 2008. The tele-town halls allowed him to make his case to targeted groups of Iowans on specific issues without raising media alerts that he was aggressively contesting the state.

On caucus day, Romney’s staff was confident in their vote count, although aware that they had dramatically underestimated turnout four years earlier in a way that made their counts irrelevant to projecting the outcome. But even this time, Romney had an advantage over rivals who hadn’t competed before in Iowa: The campaign had ranked the state’s 1,774 precincts into five tiers based on which ones would offer the best return-on-investment to Romney. The highest tiers were the ones that had produced the most votes for Romney and Branstad, and it was there the campaign first worked to enlist precinct leaders: A volunteer representative who speaks publicly on the campaign’s behalf and is authorized, under caucus rules, to observe the vote-counting. Romney’s campaign set up a phone center in Des Moines for those precinct leaders to call in with the local returns—so he could maintain his own statewide count in case of irregularities or a close race.

Romney’s team got that close race, and after the caucus results were counted his advisers appeared a bit shocked that Santorum had emerged in the last two weeks to come within eight votes of their total. But that might not have been the most revealing margin of the night. Romney closed out his Iowa campaign with 30,015 votes, six fewer than the number that resigned him to second place in 2008. Four years later, he was still stuck at 25 percent, but he left the state a winner.