THE TWO PERCENT SOLUTION

In the early afternoon of November 7, 2000, Karl Rove began to panic. The first wave of national exit polls had been released at midday, and as the numbers arrived at Governor George W. Bush’s headquarters in Austin, Texas, they foretold catastrophe. Rove, the campaign’s chief strategist, was worried. He summoned his regional political directors one by one into his office to ask what they were hearing from each of their states.

Matthew Dowd, who oversaw Bush’s polling operation, was not as shocked as Rove. Dowd had prepared for a tough election day. Typically campaigns cease their polling on the Thursday prior to a vote, just before voters become hard to reach on their home phones over the weekend. Even then, there is little room for strategic adjustments; a campaign manager who learns on Friday that the dynamics of a race have shifted has few tools left to use. It is usually impossible to find a block of advertising time still for sale, and the closing spots have already been cut and delivered to stations. At best, the manager can reroute the candidate’s weekend itinerary, order a final burst of new robocalls, or try to shuffle GOTV resources. But the headlines would force Dowd to break from standard practice. On Thursday, Dowd learned that a Maine television station had unearthed a driving-under-the-influence charge that had been lodged by local authorities against Bush twenty-four years earlier. Unbeknownst to Rove, Dowd ordered up two more nights of polling in three crucial states.

What had been a slight advantage for Bush at the end of the previous week disappeared. In Michigan, Bush had trailed Al Gore by one to two percentage points on Friday; when Dowd awoke to new numbers on Tuesday his candidate lagged by five points. Bush’s three-point lead in Florida on Friday was gone; the state had become dead even. Maine moved from a toss-up to out of Bush’s reach. As Dowd looked closely into the polls, he saw why margins had closed: social conservatives, who had always been a bit skeptical of Bush, had lost enthusiasm for him after the drunk-driving revelation.

As election day wore on, the updates Rove was getting from the regional political directors grew more dire. From around the country came increasingly fearsome reports of Democratic field operations outmatching their Republican counterparts. “We found out there were state parties that lacked sufficient volunteers to get the job done,” says Rove. “We recognized we were having a problem. It just became a bigger problem than we had anticipated because of the impact of the DUI.”

When the polls started to close on the East Coast, the consequences of this tactical gap grew evident to Dowd. The county-by-county returns in battleground states delivered the dénouement in a narrative that had been foreshadowed by the weekend polls. While Bush had suffered slightly depressed turnout among voters who should be the Republican base, Democrats—though never particularly enthused about Gore’s candidacy—still showed up in force. Even before the first planes of lawyers set off from Austin to Florida, a consensus developed at Bush’s headquarters that Democrats had turned the election into an effective tie because of their mobilizing prowess.

A few Republican strategists were prepared for this. Some had even quietly suspected that the opposition’s surprising strength in the 1998 congressional elections could be attributed to what appeared to be a newly toned turnout muscle of Democrats and their labor allies. But Republicans readily accepted this disparity as a fact of life, just as Democrats cursed their opponents’ advantage in financing their campaigns. Now the outcome of a presidential election hung perilously in the balance as a result. “That was a little bit of a shock to the Republican system,” says Curt Anderson, a longtime Republican consultant who worked on Bush’s campaign. “They were better equipped to communicate with their voters and get them to the polls. We’d been doing the same thing over and over again for twenty years.”

OVER THOSE DECADES, the Democrats started to approach field operations with a new seriousness. The reason had a lot to do with Paul Tully, a chain-smoking former Yale offensive lineman who had made his name in Democratic circles organizing the Iowa caucuses for Ted Kennedy in 1980. Four years later he strategized Walter Mondale’s path to the party’s nomination, and he became a mentor to dozens of young operatives who were inspired by the way that he brought a quantitative rigor to Matt Reese–era advances in electioneering. “Paul was always poring over numbers and looking for an interesting hook,” says Doc Sweitzer, who was Tully’s partner in a Philadelphia-based consulting firm in the early 1980s. “So many consultants aren’t good at this because they never took a math class. That made him different from guys who come from the political-hack side.”

Tully was one of Ron Brown’s first hires when Brown became the Democratic National Committee’s chairman in early 1989. Democrats had suffered three straight presidential losses and were seen as hopelessly divided. Compared with the regimented Republicans, the Democratic Party of the 1980s resembled a loose alliance of identity-based groups and organized labor—too undisciplined and varied ideologically to rally around broad, thematic election-year messages. Brown, who was close to Kennedy and had worked on Jesse Jackson’s campaign in 1988, thought that only the DNC could rise above the party’s fractiousness. He relegated the ethnic and racial caucuses, imagining the committee as the permanent presidential campaign organization. (Tully made clear he didn’t think the DNC’s agenda included electing Democrats to Congress, a branch of government to which he referred collectively as “the midgets.”) Committee operatives, he decided, would spend the four-year cycle between votes compiling binders stuffed with opposition research and state-specific strategy plans, ready for handoff to whichever candidate Democrats nominated at their 1992 convention. The fact that Brown had hired Tully as his political director—and Tully had accepted the job, which many thought beneath him—was an immediate signal to the Washington political community of Brown’s seriousness.

“Up until that point many people had thought of the DNC as a party in some European sense,” said Mark Steitz, who had worked with Brown on Jackson’s campaign and joined the committee as its research director. “He made clear to us that every day our job was to come in and figure out what we could do to win the presidential election in 1992. Anything else—improving the image of the Democratic party, making the coalition of people in the Democratic party happy with one another—we could spend time on, but only if it was in the purpose of electing a president in 1992.”

Late in the day, Tully would often stop at the National Committee for an Effective Congress’s office on the other side of the National Mall, and lose himself in the maps and files overnight. When he returned to the DNC in late morning, he would spend hours on the phone with people around the country, gathering information on what was happening in their states. “I’m in the information business!” he would pip, by way of explanation. Precinct-level data was getting richer and richer, and when Tully was dispatched to Little Rock, Arkansas, in 1992 to coordinate Bill Clinton’s campaign he had a new trove of numbers at his disposal. After its 1990 count, the U.S. Census for the first time introduced data on block groups—previously tabulated only for cities—for every county nationwide. That gave analysts a granular unit, around four thousand people, for measuring education and income, often complementary measurements of socioeconomic status. Fresh Census numbers matter, especially in dynamic areas like the Sun Belt, where populations churned so much that data became stale by the end of a ten-year cycle. An early user of computers for campaign work, Tully was obsessed with finding new sources of information—especially economic data that could pinpoint political terrain that had become more competitive during a recession—helpful to a Clinton war room working to discern where its voters were.

Six weeks before election day, the forty-eight-year-old Tully, whose perpetual energy had been sustained by voluminous amounts of coffee, cigarettes, and pizza, died of a heart attack in his Little Rock hotel room. “He had worked for four years on this—he had every map, every target, he probably knew the name of every swing voter in the country,” Clinton strategist James Carville told the New York Times for its obituary. On election night, the laminated necklace credentials granting access to the campaign’s boiler room said “270 for Tully,” a reference to the number of electoral votes necessary to win the presidency. (Clinton ultimately won 370.) “The Republicans’ orientation towards data at the time had been more advanced,” says Celinda Lake, a Clinton pollster. “Tully was the first person who really tried to drive innovation on data on the Democratic side.”

Tully had left the DNC a new apparatus designed to specialize in turnout. The national and state parties, legislative campaign committees, and statewide and congressional candidates were all required to chip into a single fund that would run unified GOTV operations for the entire Democratic ticket. These so-called coordinated campaigns handled all of the voter contacts and IDs (and the heavy election-day machinery of phone banks and vans) so that instead of duplicating the same contacts Democrats could widen their universe of targets. As Bill Clinton antagonized liberal interest groups in his first term with his support of welfare reform and the North American Free Trade Agreement, party leaders saw a well-organized turnout operation as key to rousing the “base vote,” referring largely to urban minorities, before his 1996 reelection. Party efforts were boosted by a new “Labor ’96” program, led by the AFL-CIO and its affiliated unions, which made a conscious effort to reassert themselves in Democratic politics by turning out voters directly rather than merely writing checks to favored candidates.

A sense was settling in that the parties simply approached election day differently. Democrats, whose voters tended to be packed into dense settings where door-to-door contact was highly efficient, equipped their canvassers with clipboards and maps outlining friendly precincts that targeters at the National Committee for an Effective Congress had flagged for “blind pulls,” indiscriminately yanking voters from their home. Republicans imperiously ran voter contact operations from afar, their typically better-funded campaigns blasting messages over television and relying on a well-networked coalition of satellite groups—gun owners, churches, farm bureaus—to talk to their members through voter guides. One side practiced politics as though it were a series of Great War infantry battles, the other as though directing the Kosovo air war from a command center hundreds of miles away, disinclined to muddy a single boot on the ground.

The efforts Tully had inspired were solidified in legend only one election later. The most revered Democratic operative of 2000 was not Gore speechwriter and strategist Bob Shrum or campaign manager Donna Brazile, but Michael Whouley, an elusive Bostonian whose job description was often given as “field general.” His cult extended from the party’s most junior field organizers to the political reporters to whom he refused to grant interviews all the way to Gore himself, who fanned his aide’s aura of mystery. Newspaper profiles were filled with accounts of his tricks, notably the rush-hour traffic jam Whouley had been accused of instigating on the day of the New Hampshire primary to keep supporters of Gore’s primary rival, Bill Bradley, from the polls. The Whouley mythology courted a component of the party’s often smug self-conception: Democrats won their elections through hard work and street smarts. It was, in essence, a labor theory of value in politics, favorably contrasted with the top-down capital-intensive efforts of the right. “It’s just a cultural difference,” says Tom Bonier of NCEC, who spent the fall of 2000 working on Democratic campaign efforts in Michigan. “They just don’t do this the way we do.”

Bush may have carried the election, but Democrats had won the ground game. Republicans had shown themselves highly disciplined about every part of politics but the voting. Next time, many Republican strategists worried, they might not be so lucky, or able to count on lawyers and judges to carry them through. Now they had one of their own in the White House, and if they wanted to keep him there they had four years to catch up in the crucial game of turning out votes.

EVEN BEFORE GEORGE W. BUSH moved into the White House, Matthew Dowd installed himself in an office at the Republican National Committee in Washington. He had spent the previous two decades working exclusively on Democratic campaigns, and even though he had been detailed to the RNC by George W. Bush’s closest adviser, Karl Rove, the party’s Capitol Hill headquarters still felt a bit like enemy territory. Dowd was an Austinite whose entry into Republican politics had come on local terms: not as the result of any ideological shift but from a confidence in the charismatic power of Bush. Dowd had been assured that his tour of duty at the RNC would be a brief one, offering him a chance to resettle in Austin before being called up for Bush’s reelection. “Dowd was in the enviable position of being on the outside with a lot of time to think about stuff,” says Sara Taylor, who became a regional political director at the White House. “He was the guy sitting in the room with no dirt on his hands.”

While the Bush campaign leadership litigated the outcome of the 2000 election, Dowd looked ahead to 2004. He saw a country more riven than ever before along partisan lines. The phenomenon of ticket-splitting was effectively dead. In 1984, one-quarter of voters had cast a ballot that included both Democrats and Republicans; in 2000, only 7 percent had. The next campaign would be a “motivation election,” as Dowd put it in the first strategy memo of Bush’s reelection. Swing voters had entranced presidential strategists for a generation, Dowd thought, but those mercurial centrists were shrinking in number. A new premium should be placed on finding and mobilizing those who already identified with Republicans. It would be easier to grow Bush’s market share by expanding his base than by chasing new votes in the middle.

Bush’s reelection campaign was about to inherit two of the most valuable advantages in American politics: an incumbent president’s unchallenged control of a national party committee and the ability to raise hundreds of millions of dollars. At the White House, an exultant Rove envisioned using the four-year term to plot a partisan realignment of the electorate just as his hero, Mark Hanna, had engineered one on behalf of William McKinley a century earlier. Rove believed Bush’s “compassionate conservatism,” with its focus on education and immigration reform, could sand off the hard edges of post-Reagan Republicanism and create a long-lasting home within the party for Latinos and moderate suburbanites, two of the country’s fastest-growing demographics.

Dowd had an agenda that was at once more historically modest and practically ambitious. His long, ovaloid head was perennially racked with worry, a condition that friends attributed to having been raised as one of eleven children in a Michigan Catholic family. Rove dubbed him “our dour Irishman,” and now Dowd’s almost primal sense of fatalism focused on the political moment. Perhaps the president’s leadership will trigger a permanent shift, thought Dowd, but it would be foolish to plan on it. In fact, Dowd’s analysis glumly foretold an indefinite era of ideological deadlock. “We would go into the election assuming it would be as close as 2000,” says Dowd. “What are things we could do differently that might affect a point or two?”

The RNC’s chief of staff, Jack Oliver, enlisted his deputy, Blaise Hazelwood, to help assemble an “Election Day Operations Task Force” to investigate an answer. Hazelwood, a twenty-nine-year-old veteran of presidential, statewide, and legislative campaigns, was known as one of the most disciplined and discreet operatives at party headquarters. Hazelwood had started working for the party in 1994, just out of college. After November, she was supposed to be laid off as part of a routine downsizing between elections. But when Curt Anderson, the party’s political director, would arrive at headquarters at 6 a.m., he would find the coffee made and Hazelwood hard at work. “She was on the list to be fired but she would not let herself be fired,” says Anderson. “What I found is you can’t ever fire her.” Hazelwood had long, dark, wavy hair and a tentative manner on the few occasions she spoke to large crowds, as though nervous that saying too much could lead her to give away valuable secrets. But her obsessive attention to detail, and her perpetual willingness to work eighteen-hour days without complaint, made her a perfect candidate for a project Dowd realized would require epic feats of organization and indefatigability.

They named the project the 72-Hour Task Force, a reference to the closing three days of a campaign during which Dowd believed Republicans were being repeatedly outmuscled by Democrats in the quest to make sure supporters voted. By happenstance the task force’s first full meeting, in mid-March, took place in room 2000 of the party headquarters. The reminder of the previous November had its uses, Dowd thought. Many of the party’s most influential consultants had worked on Bush’s presidential campaign, and Dowd worried they were insufficiently chastened by the outcome. Bush had lost the popular vote nationwide by half a million votes, had carried six states by a margin of five percentage points or fewer, and his election had been secured only by thirty-six days of epic litigation and petty political maneuvering in Florida. “We basically got through on the skin of our teeth,” Dowd said.

The materials spread out before select members of the RNC, congresspeople, and other party grandees who attended were meant as a sobering tonic after months of intoxicating triumphalism. Hazelwood had asked the RNC’s research department to scour newspapers for accounts of what liberal allies had done on Gore’s behalf the previous fall. Their six-page compendium—bearing the headline “What the Bad Guys Did?”—documented the way Democrats had learned to mobilize like an infantry, from the forty field staffers that the NAACP dispatched to battleground states to a Service Employees International Union phone bank with thirty-six lines, computer-assisted dialing, and the ability to make one thousand calls an hour, all within the trailer of a purple eighteen-wheel truck that volunteers called “Barney.” What most awed Republicans was the turnout strategy from Hillary Clinton’s successful run for Senate the previous year. A copy of the document found its way into Hazelwood’s hands, and she studied its careful assignment of activity by paid and volunteer staff, plotted over a six-month period down to the block level. “We were obsessed with her plan,” Hazelwood says.

Hazelwood was astonished at how different this looked from the way that Republicans prepared for election day. Reagan’s new right showed the movement’s core to be an assemblage of enthusiastic constituencies—religious true believers, free-market enthusiasts, military loyalists committed to a hawkish foreign policy. Even though conservatives scorned Democrats as beholden to unions and other identity organizations, Republican strategists started thinking about their party as a coalition of distinct interest groups, all invited to march in lockstep at election time. In the late 1970s, Richard Viguerie, the direct-mail fund-raising innovator, had acquired lists from ideologically sympathetic institutions—religious book clubs, righty magazines, and gun clubs—and mailed their members on behalf of conservative causes, with dire warnings about the liberal threat. In the 1980s, thanks in large part to Viguerie’s success cultivating small donors, the Republican National Committee had built up a permanent financial advantage over the Democrats. Party leaders made heavy investments in burnishing the party’s brand through television ads and pushing issues over talk radio. Yet when it came to directly getting their voters to the polls, Republican strategists hewed largely to Viguerie’s approach of subfranchising the work to coalition allies, trusting outside groups to know their followers.

In 2000, Hazelwood had served as the RNC’s coalitions director, overseeing three dozen networks, including sportsmen and Hispanics. Meanwhile, the Bush campaign focused on seven major coalitions, including agriculture and social conservatives. The campaign collected membership directories from the farm bureaus and megachurches whose ranks were packed with active Republicans, and then developed a strategy for communicating with each set of voters on the issues they could be expected to care about (agriculture subsidies, prayer in school). But the coalitions rarely had significant manpower of their own to offer, and certainly not on the scale of what Democrats got from labor unions, so when Republicans wanted to put out pre-election reminders by phone they had to use paid call centers. When it came to their party’s turnout strategy, Republicans would chuckle that GOTV stood for “Get on Tele Vision”—a joke about the lack of an available workforce and the diffusion of their supporters. “When Democrats walk a precinct they’re going to hit more votes than we are. It’s just the nature of it,” says Adrian Gray, who worked as a regional political director for Bush in 2000 before joining the White House staff. “We had to take geography out of the equation.”

One of the biggest problems, as Dowd laid it out in room 2000, was the voters Republicans took most for granted. Christian conservatives were the Republican counterpart of the Democrats’ labor base, but task force members griped that they lacked the “intensity of organization at the grassroots level” that the unions had. Dowd presented exit poll data showing that after Republicans took back Congress in 1994, the share of voters who identified with the religious right slipped. (One of Dowd’s slides showed that while unions and religious conservatives each represented about 20 percent of the voting-age population, 26 percent of those who voted in 2000 were union members and only 14 percent were religious conservatives.) If capturing both houses of Congress diminished the conservative base’s appetite for power, task force members wondered, what messages could mobilize it when the movement was represented in all three branches of government, as it was now?

Dowd was interested in a debate over issue message and strategy, but he thought there were more basic matters to address. How could Republicans tailor narrow messages to their base if they did not know, in individual terms, who the members of their base were? And how would they deliver those messages if the party had no protocols for voter contact? Republicans, he decided, had to figure out what made people vote. “I had heard enough of these stories, but where’s the fucking data? Everybody always bullshits all the time about Oh, this won the campaign!” says Dowd. “We just knew that a systematic approach was the only way we could really judge it, and rate differences as opposed to stories, anecdotes.”

Dowd had quit graduate school at the University of Texas in the 1980s to work on campaigns, but the former political science student prided himself on paying more fealty to academic research than did many of his consulting peers. At the end of 2000, as Dowd tried to catch up on reading that had piled up during the campaign thrum, he came across the article in American Political Science Review in which Don Green and Alan Gerber recounted their New Haven experiment. The findings that in-person contact worked better than mail and phones confirmed Dowd’s instincts. The article also convinced him that the academics’ randomized-trial method was one the party could adopt to run its own experiments and develop a new set of best practices for turnout. “One of the incentives for doing tests was to change the culture of allocating resources,” says Hazelwood. “To convince our campaign managers and general consultants to take money out of their media buys and put it into the ground, we needed real results and real tests.”

Dowd knew a serious experimental regime would threaten many of the party’s longtime consultants and vendors, especially media specialists whose overfed campaign budgets had starved Republican field operations. He didn’t much care. “Having come from the Democratic side, I wasn’t aware of all these people,” he says. “I didn’t have a stake in it.” But Hazelwood, who had worked in Republican campaigns her whole career and was married to one of their leading direct-mail vendors, did. It would be her job to insulate the party’s culture from the jolt Dowd was about to deliver. To reduce the likelihood of internecine conflicts that could follow among consultants whose interests were newly pitted against one another, Hazelwood decided it would be important that their research agenda look like a party-wide effort, at once centralized and collaborative. Hazelwood’s e-mailed invitations to meetings of the 72-Hour Task Force did not need to spell out a warning that the 2004 campaign budget would be shaped by the task force’s decisions, and that it was better for any consultant to be inside the room than outside it. “It was Karl Rove’s baby and they wanted to be part of that,” says Hazelwood.

Every week or two throughout the spring of 2001, around forty Republican consultants, summoned by Hazelwood’s e-mail, would arrive in Washington, traveling from across the country at their own expense. The group slowly compiled a list with dozens of possible research queries, many skeptically addressing “things we all thought we knew,” according to Hazelwood, and others with questions no one had ever thought to ask. They would test the value of a message sent by phone as opposed to mail, obviously, but what about two phone calls versus one, or two phone calls versus two mailed brochures? Did a piece of mail followed by a phone call have a different impact than the opposite sequence? Did it matter whether the caller or door-knocker was a volunteer or a paid contractor? What could the party do to increase new registrations and take advantage of opportunities to cast an absentee ballot or participate in early-vote programs being introduced in several states?

Unlike Dowd, who had worked in media and polling, Hazelwood saw a particular value in the experiments beyond the findings themselves. Because Republicans lacked the tradition of in-person canvassing that ran deep in Democratic culture, Hazelwood worried that field organizers would have trouble convincing volunteers to take on unfamiliar, and often charmless, duties. A scientific study with measurable effects could help show field-workers that their tasks were crucial to the broader effort. “They were crying out for this data, all of the time asking: ‘Why are we doing this?’ ” she says, recalling her interactions with grassroots activists she tried to enlist in 2000 as part of her coalitions work. The next time those questions came up, “we could say: ‘I know it sucks in Florida walking in the summer, but look at these tests—it improved turnout.’ ”

Hazelwood mapped out a rough schedule for the next three and a half years. She wanted to complete the tests quickly so that Republicans could start to rebuild their electioneering operation in 2002, when one-third of Senate seats and a majority of the nation’s governorships would be in play, and perfect them in time for 2004. For those seeking to shift the balance of power in the United States, 2001 offered a meager roster of elections. But for those seeking to learn how campaigns work, it amounted to fertile experimental terrain.

THE FIRST WEEK of October 2001, eight RNC staffers drove from headquarters seventy miles to the northwest of Washington, D.C., and came to a halt, as though looking to settle just beyond the periphery of the capital’s influence. They had journeyed through dense inner-ring Virginia suburbs, and then past the blossom of office parks and new housing developments that marked the quickly emerging communities demographers like to call “exurbs” and that White House strategists saw as friendly turf nationwide for Bush’s moderate, family-oriented conservatism. The caravan settled in Winchester, the redbrick seat of bucolic Frederick County, in the Shenandoah valley. The RNC staffers found an empty storefront between a local Republican headquarters and an Irish pub and began to turn it into a field office in a place where the national party had never before had a direct presence.

As in much of rural Virginia, Republicans had taken Frederick County’s votes for granted ever since they broke for Richard Nixon in 1968. With its network of politically involved churches and active gun culture, there was little chance of the area going blue. But it was also a place where the Republicans’ laissez-faire approach to base turnout may have limited their vote totals at the expense of statewide candidates. A few months earlier, Hazelwood had approached Timmy Teepell, who had succeeded her as coalitions director, to propose that he design and oversee the most ambitious experiment on the task force’s docket. Hazelwood had prepared to make a large investment to test how a new approach to coalition organizing could benefit Republican performance. What would happen if the RNC overwhelmed one of its strongholds with both paid contact and volunteer operations?

Virginia had one of the few gubernatorial elections on the calendar, a race whose stakes would matter to social conservatives. For comparison, Teepell selected two counties whose size and demographics were similar, and where coalition allies had a large footprint: Frederick, in the northwest corner of the state, and Roanoke, to its south. (In 2000, Bush had carried both counties with more than 60 percent of the vote.) In one, the RNC would try to maximize its presence; in the other, they would stay away and let the state party proceed with its normal efforts on behalf of its gubernatorial nominee, who was Attorney General Mark Earley, and down-ballot candidates. “Frederick was closer to my home, so we used that as the test and Roanoke as the control,” says Teepell, who commuted to Washington from suburban Loudoun County. “I had three kids and wanted to sleep in my own bed.”

Teepell’s eight-person team, comprising the RNC’s entire coalitions department, arrived in Winchester in time to spend the last five weeks cultivating the conservative base there. Teepell was given the standard binder of research that the party had prepared on Democratic gubernatorial nominee Mark Warner, a former telecom entrepreneur who in an earlier Senate race had built a statewide reputation as a probusiness moderate with the potential to win crossover votes. The material that polls showed was most likely to provoke social conservatives was an unearthed clip from a Warner speech to a Democratic convention seven years earlier when Warner, the party chairman, attacked the state’s conservative activists as extreme. In Frederick County, Warner’s quote became grist for four targeted mail pieces (“Mark Warner described the views of people of faith and homeschoolers as ‘threatening’ to America”), four paid phone calls, and two ads on local Christian radio stations.

Because he came out of the world of coalitions organizing, Teepell tended to think of politics not as an activity conducted in well-bounded geographic spaces but as one that pulsed through networks of people linked by common interests. Teepell had his team make a list of the local communities he called “social precincts,” and when it became time to recruit volunteers he had his staff fan out to gun shops and church events. “You may have a neighborhood that tends to vote a certain way,” he says. “You can also have a homeschool cooperative who think similarly about the same issues, but because they are across the region you can talk to them when they come together.” Teepell arrived in Frederick with a particular interest in marshaling those homeschool families, many of them Christian conservatives opting out of secular schools. His own political career had begun as a homeschooled fifteen-year-old going door-to-door for Republican candidates in his native Baton Rouge, Louisiana. Now he looked at families like his—who tend to be exceptionally well networked because parents organize extracurricular-style activities to bring homeschooled children together—as an ideal volunteer constituency. A coalition-based volunteer strategy, more broadly, was based on a similar logic: finding people who are used to being active together and assigning them a new political mission.

As his staffers recruited new volunteers and oversaw phone bank shifts and printed out walk lists for them, Teepell obsessively tracked their work to report back to Washington. Over the course of the month, volunteers spent a combined 407 hours at the door identifying voters and 459 hours persuading them there, leaving behind doorhangers with a picture of a crawling baby under the slogan “Protect Virginia’s Values.” From a phone bank, volunteers added 305 hours on IDs and 314 on persuasion. On election day, 75 volunteers were assigned to rouse voters at their doorstep while 53 did it by phone.

Warner beat Earley statewide, in one of two governor’s races the Republicans lost that night. But there was a glimmer of good news for those monitoring results in the RNC’s political department. Earley ran five points stronger in Frederick than he did in Roanoke, four points of which their later analysis credited specifically to the coalition exercises they had run to motivate social conservatives. “That was like, ‘Oh my God!’ ” says Hazelwood. “That was pretty shocking to me, that we could increase turnout that much.”

In Hazelwood’s mind, 2 percentage points had been something of a benchmark for turnout improvements. Any intervention that exceeded that was promising and worthy of future attention; one that fell short probably was not worth the trouble. The numbers from Frederick suggested Hazelwood had been dramatically underestimating the influence a rigorous campaign could have. When RNC pollsters called social conservatives in the two counties, those in Frederick were 16 percent more likely to say they had been contacted about Earley’s campaign, which suggested that not only had the additional resources and manpower expanded the party’s reach but also that the quality of volunteer interaction had been high enough to make an impression.

By the end of 2001, Hazelwood had overseen fifty experiments and had developed a formidable body of knowledge about what worked in voter contact. In South Carolina, calls from volunteers had turned out voters at a rate five points higher than paid call centers. Sending new Republican registrants in suburban Philadelphia a piece of literature with local polling place information increased their turnout by six points. Dispatching “ground troops” to flush voters to the polls on the Monday and Tuesday of the election improved Republican turnout by an average of three points. Assigning a full-time precinct worker, as the party did in both Pennsylvania and Virginia, added three. At one time, such margins might have been shrugged off as negligible. “One good thing about the closeness of the last presidential election is that it erases the need to convince people how important two percent can be,” Hazelwood, recently promoted to RNC political director, told the party’s winter meeting in Austin the next January.

Hazelwood devoted a considerable share of her PowerPoint presentation at that gathering to explaining the Frederick County coalition test, as an example of what Republicans could do when they flooded an area already home to a developed network of coalition allies. “We aren’t here to tell you about your state, but rather to give you an idea of what can be accomplished within a target group if you really make a commitment to do it,” she said to the gathering of party officials at an Austin hotel ballroom.

Another set of experiments offered lessons that might prove more portable to other parts of the country where Republicans lacked an existing infrastructure. In two precincts in different states, a new-style targeting of phone and mail messages had increased Republican vote share by almost exactly the same amount: a 3.6-point gain by Earley in the Virginia governor’s race, and 3.4 points for congressional candidate Joe Wilson in South Carolina. In these cases, the targeting was based on the standard individual-level data available on the party’s national voter file, which mostly designated voters by straightforward demographic categories. Richer data, Hazelwood explained to her crowd, would only make the targeting sharper. “Knowing where a voter lives, how old they are, what gender they are, and all those things are very important. But nothing is as important as understanding what they really care about,” she said. “To accomplish this, we need better information on more voters. This can be done, but it will take a team effort and a lot of people willing to give some time.”

Hazelwood concluded her presentation by encouraging state and county party officials to schedule a daylong turnout seminar in which the RNC staff would train local activists to better use volunteer door-knocking, paid phone banks, and coalition lists to identify individual voters for the 2002 midterm elections. Already Hazelwood had hosted four regional meetings that included party leaders from forty-seven states. “The other three states have been kicked out of the party,” she joked.

Dowd was sitting on a panel to discuss the findings. Though he kept it to himself at the time, he had recently learned of a method that promised to render entirely moot the arduous and costly process that Hazelwood had just outlined. This new tool would offer a powerful boost to whichever party first mastered it: a way to divine what issues every single voter in America cared about without having to track them down and ask them individually.

A FEW MONTHS EARLIER, Matthew Dowd had wandered the long, colonnaded porch of the Grand Hotel on Michigan’s Mackinac Island, aware that the political terrain he had so delicately mapped with 2004 in mind was already shifting beneath his feet. It was one week after Al Qaeda’s assault on New York and Washington. Immediately after the attacks, staffers at RNC headquarters had been instructed not to place any phone calls or send e-mails outside the building, for fear they would be seen as playing politics at a time of national tragedy. Pundits speculated that partisanship could be indefinitely pushed aside in favor of a greater sense of national purpose. But the point of the Mackinac Republican Leadership Conference was partisanship, and the island’s autumnal camouflage offered a low-profile refuge for swing state operatives whose competitiveness had not been dulled by the terrorist attacks.

Alex Gage arrived on the island with his mind fully on 2004. That summer, Gage, a longtime Republican pollster based near Detroit, had attended one of Hazelwood’s many presentations of the 72-Hour Task Force’s work. Two of the eight items on her “Prescription for Victory” checklist particularly drew his attention. One was the need for “sharper targeting” of messages by phone and mail; the other was “voter identification,” where Hazelwood said Republicans demanded more efficient methods. Gage decided to treat these “almost like a Harvard Business School problem,” as he put it. When he traveled a few months later to northern Michigan, he was eager to tell Dowd that he might have figured out a way to meet both challenges at once.