The Victory Lab (24 page)

To Checchi, Penn and Shrum were brand names, and once he expressed an interest in hiring them “the gold rush was on,” as one campaign aide told the
Washington Post
. Shrum negotiated a contract to produce one hundred television ads, only forty of which ever aired. The remainder flunked shopping mall screenings Penn conducted before voters, one of his signature techniques for testing the effectiveness of ads and mail. When Penn’s polls showed that the Shrum ads that had aired were not succeeding in changing voters’ minds, the gurus managed to convince Checchi to blame the largely unknown firm that purchased the airtime. Shrum then secured an arrangement whereby his firm would take control of the ad placement, too, earning on each spot a commission that he split with Penn. By the end of the campaign, Checchi’s friends were telling him he had been fleeced, spending $40 million of his own money on a losing primary campaign, $2 million going to Penn and Shrum in fees and commissions. But it was Checchi, the candidate, who was laughed out of politics. Shrum went on to be the lead strategist for the campaigns of Al Gore and John Kerry, Penn for Joe Lieberman and Hillary Clinton.

A lot of the political operatives who entered the field in the 1980s and 1990s did so because they envied the wealth and celebrity that flowed to consultants like Penn and Shrum. Podhorzer entered the field because he aspired to outsmart them, and the only counterweight he knew to the tyranny of the gurus was the incisive power of data. As a child in Brookline, Massachusetts, Podhorzer was a baseball fan who obsessed over the board
game Strat-o-Matic, through which contestants could replicate entire games based on odds derived from ballplayers’ real-life performance. When computers arrived, Podhorzer realized that he could write his own programs testing scenarios in the game, and by the age of twelve he was trekking to local libraries and nearby universities to find time at a Wang terminal. “It at a very early age makes you interested in probabilities and statistical outcomes and so on,” says Podhorzer. “It made the value of thinking about the world in analytical terms—and not taking people’s old saws at face value—ingrained in my personality. That’s the way I’ve pretty much approached everything in politics.”

After graduating from college, Podhorzer went to work for the consumer group Citizen Action, for which he led issue-based campaigns on Capitol Hill. In 1983, when Ronald Reagan announced a proposal to deregulate the natural gas industry, Podhorzer set out to build a bipartisan coalition to beat the White House. But he didn’t buy the ads or hire the lobbying team typical for a high-stakes legislative fight. Instead, Podhorzer identified midwestern Republicans in Congress who weren’t typical allies on progressive issues but had large concentrations of natural gas users in their districts. Then Podhorzer created lists of the Republican members’ supporters and encouraged them to push their representatives to oppose the bill. “We let data inform our pre-strategy,” he says. “One of my first impulses is always thinking ‘what do we know and how can we make that information work for us?’ ”

In 1997, Podhorzer was approached by Steve Rosenthal, who had just become the AFL-CIO’s political director and was tasked with building the biggest turnout operation in labor’s history. Rosenthal had been appointed to his post by the AFL’s newly elected president, John Sweeney, who vowed that organized labor would reassert itself as an election day force. Rosenthal’s job was to mobilize union households who had been dispirited both by a Clinton term they saw as accommodationist and probusiness and by Republican congressional victories in the 1994 midterms. Sweeney helped
to raise $32 million for the campaign called Labor ’96, several times more than any other AFL electoral project in years, but Rosenthal quickly despaired that they were unprepared to spend it effectively.

In his first few weeks on the job, Rosenthal traveled the country to sit in on political training sessions that the AFL hosted for local labor officials. At one, Rosenthal perked up when he heard an AFL staff organizer tell her trainees that “union members read their mail on Fridays”—too busy to read it daily, they let it stack up until the end of the week, when they sort through the piles and open the pieces that interest them. So, the organizer went on, an effective mail program was one that targeted its brochures to arrive on Fridays. “I was just completely blown away,” says Rosenthal. “I had never heard this before, and I couldn’t possibly imagine every union reader had the same habits in reading their mail. I confronted her afterwards, and effectively what she said was ‘That’s how I do my mail, and I’m sure other people are in the same position I’m in.’ ”

Rosenthal thought labor’s tactics deserved a little more rigor. He had followed Podhorzer’s work at Citizen Action and asked his new bosses at AFSCME, the public employees’ union in whose political department Podhorzer now worked, if they could lend him out to the new campaign. “I had a very good gut sense of the message—what union members wanted to hear, what they needed to hear,” says Rosenthal. “Mike had a very good sense of the mechanics, and thinking outside the box about what needed to be done to move people.”

The two men could not have been more different. Rosenthal was a bearish Long Islander with a bushy mustache who made a point of getting out of his Washington office and personally visiting labor’s organizing campaigns around the country, raising his own voice to rally a crowd of workers clad in vinyl jackets. (Sweeney, who had been elected as a consensus candidate during a fractious campaign, was a poor communicator and left an outsized public role to his political director.) Podhorzer looked like he would be far more at home in a faculty lounge than a union hall. “He had the lowest profile of anyone in D.C.,” says Judith Freeman, who worked in the AFL
political department under Podhorzer. “He’s just naturally introverted. He didn’t spend his time out there in the city telling people how smart he was.”

Rosenthal was pleased with the results of Labor ’96, which helped to increase labor’s share of the total popular vote to 23 percent (from its dismal 13 percent in 1994). But as he drew up plans for a 1998 sequel he wasn’t convinced he knew which items in the AFL’s variegated budget actually delivered the votes that had helped to return Clinton to the White House. “I looked at it and said there isn’t an industry in America that would spend $32 million without spending a significant piece of that budget trying to understand what works and what doesn’t,” Rosenthal says. “The problem with almost everything the party committees and campaign did is that people rarely take the time to look back.” So he and Podhorzer began casting about for ways to measure the effect of their voter contact programs. They organized one program where they paid union members in battleground states one hundred dollars each to collect all the political mail they received—marking each leaflet with the date of arrival before placing it in a pre-addressed envelope to Washington—and even asked some for permission to install cameras in their homes to tape them as they received mail and opened it.

Along the way, the AFL continued to spend hundreds and thousands of dollars each year on polling and focus groups, much of it to gauge what persuaded or motivated voters. But neither of those more traditional methods revealed much. A good poll can present an indication of how the electorate, and significant demographic groups within it, is moving over time, but cannot isolate the effect of any individual message—and certainly not a single mailed leaflet, one of organized labor’s favorite tools. Campaign pollsters often tested new messages by first asking voters whom they supported, bombarding them with a set of new facts about a candidate—his views on issues, an aspect of his record, a fact about his personal life—and then repeating the initial question. “You’re going to base your whole campaign on who within a minute changes their mind?” asks Democratic strategist Kevin Looper. “There’s a good reason campaigns treat people like idiots.”

When pollsters wanted a more nuanced look at how voters made assessments, they would convene a focus group. For presidential campaigns, this required identifying a swing state city and inviting around a dozen people, embodying a range of ages, genders, and races roughly representative of the electorate. Their opinions were valued, the invitees were told, and they would be compensated for a few hours of their time with all the sandwiches they could eat. A moderator might show an ad, pass around a brochure, or read a snippet from a possible speech, and then call on people around the table to see what they thought of it. Under the odd conventions of the consulting business, the same public opinion firms that did polling typically did focus groups, so often the person charged with coaxing latent opinions out of strangers was there only because he had shown a deftness with numbers.

Focus groups could give a rich impression of how a collection of voters responded to an individual leaflet, but they would only get the instant reaction of someone being paid one hundred dollars for his or her attention. A focus group could not reveal much about whether a typical voter would even notice the brochure if it showed up in the mail wedged between a birthday card and a water bill. Consultants who sat behind the one-way glass noticed that voters being paid in cash and sandwiches to weigh in on political communication are often overcome by self-consciousness. In focus groups, people say they don’t like negative ads, but when the ads come over their airwaves they seem to succeed in worsening voters’ impression of their target. “You get this spiral of negativity about an ad. People want to compete with each other to be more cynical,” says Mark Mellman, whose firm ran both polls and focus groups. “They’re noticing things just to notice them, because you’ve asked them to fill up the two hours.”

Podhorzer was desperate for a method that would cleanly measure cause and effect in the real world, and he found one after encountering James McGreevey. In the summer of 1997, McGreevey, a young state senator and mayor running for governor in New Jersey, came to Washington
to introduce himself to political directors of the national unions, a typical stop for statewide candidates seeking support. It did not go well. “He came in and gave what was probably the worst presentation any of us had ever seen by a candidate,” Rosenthal recalls. McGreevey’s delivery was flat and his themes too general, showing little interest in workers’ issues. (For years, Rosenthal and his peers would call it “pulling a McGreevey” when a rookie candidate came to Washington and bombed.) Rosenthal concluded after the meeting that McGreevey’s challenge to Republican incumbent Christie Todd Whitman was hopeless, unworthy of the AFL’s investment—but a useful ground for research that could help more promising candidates in 1998. Podhorzer designed a test in which he let some of the state’s 3,300 union locals continue with their plans to deliver pro-McGreevey mail and phone calls to members while others stayed on the sidelines. Afterward Podhorzer was able to compare turnout by union members across the state and attribute disparities to the different levels of communication by locals. Podhorzer’s experimental design (which did not clinically randomize voters) may have failed to meet academic standards, but he was onto the same approach Gerber and Green would refine the next year. Yet after the election, the AFL kept the details of their test secret: Whitman ended up winning by barely twenty-five thousand votes, and Rosenthal worried that his decision to allow locals to sit on their hands could get blamed for failing to put McGreevey over the top. “The running joke,” he says, “was that if we had just run the full program to everyone we could have helped bring home a good chunk of those votes.”

After the 2002 elections, Rosenthal left the AFL. New campaign finance laws taking effect would dramatically restrict the way candidates and parties could raise money, and Rosenthal intended to launch an independent organization that would assume some of the campaigns’ core functions, like registering and turning out voters, under looser fund-raising rules. (Rosenthal’s group became America Coming Together.) Podhorzer remained at the AFL, but with a diminishing faith in the traditional infrastructure of political expertise. The debates over which institution would
be responsible for turning out voters seemed to ignore the fact that none of them really knew how to do so.

Podhorzer came to believe that his perch at the AFL offered a unique opportunity to reinvent the way campaigns work. Such permanence was a rare trait in the political world. Candidate campaigns were short-horizon projects, rarely interested in developing unproven techniques. The national parties might have looked more durable, but they had turnover, too, and it usually came at the worst time imaginable—in the middle of a presidential election season when the party’s nominee was granted effective control of the enterprise. Only institutions like labor unions—and other allies in the Democratic coalition, like the League of Conservation Voters and the women’s group EMILY’s List—would always have a stake in improving their tactics for the next election. “Until you get into a more rigorous approach, you are essentially left with what we had,” says Podhorzer, “which is that everything you did in a winning campaign was a good idea, and everything that you did in a losing campaign was a bad idea.”

Podhorzer had read the first Gerber-Green paper when it came out, and he began implementing their conclusions about the relative value of campaign tools. Robocalls, proven ineffective at turnout, were banished from the AFL’s GOTV repertoire. Along with Malchow, one of three AFL mail vendors, Podhorzer began regularly speaking to the Yale professors about how to adapt experimental methods to the particular challenges faced by organized labor. “Most of their studies have focused on simply why people vote, not who they vote for, which is of obvious importance to an organization that’s involved in politics,” Podhorzer says.

In the run-up to the 2004 elections, the AFL would be sending almost monthly mailings to targeted members, and Podhorzer thought randomized-control trials could offer the basis for a “continuous feedback loop.” The twenty-thousand-person Ohio survey sample would be converted into a massive focus group, but without any of the respondents knowing they were under examination. Voters would encounter the AFL’s messages the way they normally did, fetching them from a mailbox, instead
of having them handed by a moderator. Podhorzer could test different messages with small subsets of voters, while leaving a control group untouched. If voters never looked at a piece of mail, or picked it up but never paid attention to it, or read it and seconds later forgot what it said, that was worth measuring. If they studied the brochure but their opinions didn’t shift, that was worth measuring, too. Polling across all the groups would make it possible to isolate the specific impact of the mailer and send the most influential ones to a broader array of members.