The Victory Lab (10 page)

Few Democrats thrived in the contentious aftermath of the 1994 Republican takeover of Congress as much as Malchow, who had won the Democratic National Committee’s contract for fund-raising mail. “Nothing is better for direct mail than this sort of discord,” says Malchow, who eagerly cited Newt Gingrich’s latest outrage in his letters. So many rank-and-file Democrats had become energized by Gingrich’s agenda that Malchow thought he could dramatically expand his pool of fund-raising targets. At the time, there were about 2 million established liberal donors who would reliably write checks to campaigns and lefty causes like the ACLU and the Sierra Club. But there were 40 million committed Democratic voters, and Malchow thought it might be time to start soliciting them, too.

Malchow had pushed himself to master CHAID, taking his statistics textbook on vacation and reading it on the beach, to his wife’s complaints. Now he thought he had the perfect application for it in politics: he would use CHAID to break down the party’s ranks of existing donors to isolate
traits they had in common geographically—using donors’ nine-digit ZIP codes to
group Census data in twenty-five categories like education, income, race—and then assemble those into a “model” of what a potential Democratic donor ought to look like. In the end, Malchow could never quite master the computer models, but he became fixated on the idea that CHAID could be used to find not only people worth begging for cash but also those whose votes were up for grabs, too. “I’m hungry to do this,” Malchow recalls thinking. “Looking for the first chance.”

RON WYDEN WAS
a former college basketball player with an idiosyncratic streak on health policy. Gordon Smith was the former CEO of his family’s frozen food business and—despite his Mormon faith—had a liberal instinct on social matters. But within days of winning Oregon’s Democratic and Republican primaries, respectively, in December 1995, these two thoughtful men reduced each other to partisan caricature.
Smith dismissed his Democratic opponent as a “tax-and-spend liberal,” while Wyden accused the Republican of “extremism.” It was the most generic version of a mid-1990s election imaginable, a proxy battle in the Washington war between Bill Clinton and Newt Gingrich over the size and role of government.

Wyden and Smith were joined in this conflict only through unusual circumstances. Three months earlier, Senator Bob Packwood had resigned from the seat he had occupied for nearly three decades, shortly after the Senate Ethics Committee voted to expel him over charges of sexual harassment. Since 1969, no one but Packwood and Mark Hatfield had represented Oregon in the U.S. Senate, and the two moderate Republicans had used their longevity to build the state’s clout in Washington and check their party’s rightward movement. Their permanence had also made it difficult for other Oregon politicians of either party to rise, so when Packwood stepped aside it offered a rare opening. Wyden, an eight-term congressman,
and Smith, who had served in the state senate for only three years but was already the body’s president, rushed into the race.

The special election was scheduled for January 30, 1996, effectively inaugurating a campaign year that would culminate in Clinton’s reelection effort that November. But the Oregon race would be controlled by a distinctive set of factors. Because the campaign would bridge the holidays, when it is ever harder to recruit volunteers and reach voters at home, the campaigns had to expect that Oregonians would check out of politics for one of the race’s final weeks. There would not be time for the patient introduction usually necessary for a first-time statewide candidate: the biographical television ads, the policy speeches that allow a candidate to knit policy positions into a coherent ideological worldview. The calendar offered little chance for candidates to visit with voters individually or parry with local journalists and editorial boards that could help filter the politicians’ backgrounds into rich portraits of the two men. In addition, because of a change in Oregon law, the primary and general elections would be the first ever conducted entirely by mail ballots anywhere in the United States, and no one knew what that would mean for voter turnout. It was nearly impossible for strategists to forecast how many votes were likely to be cast and where they would come from, crucial information for allocating scarce campaign resources.

In December, Wyden had just emerged from a hard-hearted primary against a fellow congressman, Peter DeFazio. The fault lines among Democrats were delicate—the
populist DeFazio had attacked Wyden as a “corporate Congressman”—but there was scarce time for patient coalition building or diligent canvassing to discern exactly where they ran through the electorate. The six-point primary victory was best interpreted with a map:
Wyden carried the Portland metro area that housed his congressional district but lost southwestern Oregon, where DeFazio’s district hugged the Pacific coastline, by more than 2 to 1. Wyden’s strategists hoped that portraying Smith as a doctrinaire right-winger would help them consolidate
Democratic support, but they were also aware that independents, who made up nearly one-quarter of the state’s voters, might not respond as readily to such partisan messages. “The fundamental strategic challenge was how to break out of that trench warfare, just big bombs back and forth,” says Mark Mellman, who presided over one of Washington’s top Democratic opinion research firms and had consulted on Wyden’s campaigns for years.

Mellman’s polls were great at showing where the campaign stood, and in their ability to run a sequence of questions that gave a deeper understanding of what people knew and what they believed, where there was room to give them new information and change their views. Pollsters loved to use their queries to identify the conflicted public mind: Did voters like the incumbent personally but show disappointment in his work on a particular issue? If so, should a challenger focus on that specific policy difference instead of a broader leadership critique? Then the pollster could break down the numbers into demographic blocs, showing how opinions differed by gender, race, party, age, or region of the state. Going back into the field with the same methods each week would allow the pollsters to see which type of voters were moving, and start to make educated guesses about why.

Hal Malchow had worked with Mellman since Al Gore’s first Senate race, in 1984, when Malchow served as campaign manager and Mellman conducted polls from the New Haven, Connecticut, apartment where he had lived while finishing his graduate studies at Yale. Malchow thought highly of Mellman’s statistical acumen, strategic perspective, and willingness to experiment with new methods made possible by changes in technology. But in his capacity as Wyden’s direct-mail consultant, Malchow found that Mellman’s insights on the race had limited utility. Each of Malchow’s brochures was destined for a specific mailbox, so he was less interested in understanding the nuances of public attitudes than he was in identifying which individual voters held which views. He wanted to send persuasion mail to people whose minds were not yet made up but who looked like they could be swayed by pro-Wyden appeals. Paying for persuasion mail to one of Wyden’s firm supporters would be a waste, while sending
it to one of Smith’s risked inflaming the opposition with a reminder of an election that could otherwise be easily ignored. For Malchow’s purposes, knowing that 20 percent of seniors or suburban blacks were undecided was not terribly helpful. A mail program that aimed at voting blocs defined in such crude terms would ensure that 80 percent of the printing and postage costs went to waste.

More and more types of individual data were showing up on Malchow’s computers, and he thought he could use them to cut the electorate into much finer slices. In 1990, the U.S. Census Bureau started releasing information on individual city blocks, which typically include eight hundred households, in three dozen demographic categories—and each was attached to a nine-digit ZIP code. While a pollster could break down a six-hundred-person survey into key constituencies, as he subdivided those into the smaller categories they called crosstabs (for “crosstabulation”) they lost their statistical power. Once you started looking at what, for instance, young, college-educated Republican women thought of your candidate, the margin of error on the poll numbers rose to double digits, which made it effectively worthless. Malchow wanted to invert the quality of the data: more information on the people, less on their views. Instead of contacting a small sample with a long battery of questions, he wondered, why couldn’t the campaign ask just one or two questions each of a lot of people?

Campaigns were already making calls to identify potential supporters: one of the first places a field organizer would stick an eager volunteer was a phone bank, with a list of voters and a few simple scripted questions written to discern whether the person on the other end had picked a candidate and how strong his or her support was. This effort helped campaigns decide which voters needed to be badgered in the run-up to election day, and which ones would be a priority to rouse to the polls in the closing get-out-the-vote operations. While some strategists asked for daily summaries of phone bank interviews as a rough indicator of where voters stood, the calls were far from random—lists usually started with party registrants because it made little sense to waste time identifying the other
side’s loyalists—and had no statistical worth as representative of public opinion. But Malchow didn’t care where the body politic was headed; he wanted to pull out its parts and dissect them so he could see how every cell inside pulsed differently and assign each a slightly different prescription.

With CHAID, Malchow could get beyond the limits of precinct targets and aim for actual people. If it was possible to address mail to an individual, why were campaigns still profiling them based on the neighborhood in which they lived? He would need a large-scale poll, as much as twenty times bigger than the standard opinion survey conducted by media organizations to identify a front-runner or a campaign pollster to measure the dynamics of a race. In this case, Malchow thought it would take about ten thousand phone calls to get the detail he wanted, costing about a dollar each from a professional call center. But it would not require a sophisticated polling operation, which would have been far more expensive. With just a handful of questions there would be little worry about the order in which they were asked—which with opinion polls can bias the results—and since the responses wouldn’t be assembled to assess public opinion, there was less of a need for a representative sample.

In Wyden’s campaign, Malchow knew he had found the perfect opportunity to experiment. He was not, however, well positioned to propose a radical departure from traditional campaign tactics. Malchow was already deep in battle with Amy Chapman, Wyden’s campaign manager, and the only reason he thought he had succeeded in holding on to the contract was that Chapman’s brother worked at his firm. There was no way that Chapman would write a check for a ten-thousand-dollar poll on top of the existing budget for voter lists and mail, so Malchow decided he would pay for it himself.

The Wyden campaign reached 9,051 voters and asked them three questions to identify whom they supported and how committed they were. The first surprise was one that should have been picked up by a standard poll, although the massive sample meant Malchow could put great confidence
in the finding: the 30 percent of voters undecided about their choice were not necessarily clumped in the ideological middle. While 32 percent of independents said they hadn’t settled on a candidate, 30 percent of Republicans and 28 percent of Democrats hadn’t, either.

It was the last group that most interested Malchow. There was no reason to spend forty cents to print and send another piece of persuasion mail to the 72 percent of Democrats who were already with Wyden. But if Malchow could pick out those 28 percent, one by one, and speak to only their mailboxes, he could make better use of his resources and communicate with them more often. CHAID revealed a chunk of them were Democrats who lived in the Eugene media market, several hours south of Portland. This was logical: the area was home to Peter DeFazio, the man Wyden had just beaten in a contentious primary.

Another group of Democrats in the 28 percent were those who lived in neighborhoods with the lowest education levels, an indication that they could be the type of so-called low-information voters who do not follow politics closely and tune in to races just before election day, if at all. According to Malchow’s poll, more than one in three of the voters in this group were undecided: not only were these Democrats still making up their minds, but the numbers showed that they presented a more target-rich environment than the general category of independents. One hundred leaflets sent to this group would reach more actually undecided voters than merely pulling the names of registered independents.

Then Malchow looked at the 32 percent of undecided independents and started sprouting branches off the tree to see if he could find clusters that would be more fruitful. When he clicked on independents to break down by age, he got three more boxes: 18- to 44-year-olds, 45 to 64, and seniors. The youngest group was 37 percent undecided, compared with 26 percent and 34 percent of the older ones, respectively, so Malchow clicked on that category. When he divided those young voters by the education level of their Census blocks, he saw a meaningful split: those in neighborhoods with the fewest graduate degrees were undecided at a much higher
rate—42 percent to 33 percent—than those in the more educated neighborhoods. Malchow broke down that group by family size, and hit gold: those in neighborhoods with the most children were undecided at a rate of 58 percent. He thought these young independents in neighborhoods with lots of kids but few graduate degrees resembled a politically familiar bloc: the culturally conservative, working-class voters now known as Reagan Democrats. Only 67 of the original 9,051 people surveyed fit the category, but Malchow’s file of all Oregon voters revealed a much larger pool of targets likely to share a frame of mind: around 8,000 likely voters. Malchow expected they would be nearly twice as likely to be undecided as other Oregon independents. A dollar of Wyden’s campaign budget would go twice as far with them.