The Victory Lab (37 page)

ON JANUARY 2
, hours before the deadline to file for the primary ballot, Carole Keeton
Strayhorn stood at an Austin high school and declared she was leaving the Republican Party and would run as an independent in the fall. The announcement shocked Perry’s headquarters, which had begun the new year ready for an eight-week sprint until the March 7 election. Instead, their show of strength had earned them a free trip to the general election, where they would likely be part of a four-way race. (Democrats were far from selecting their nominee, and country singer-songwriter and novelist Kinky Friedman was preparing his own independent candidacy.)

But the news unsettled the eggheads, who had designed their most important experiment—staggering Perry’s broadcast advertisements in random waves across the state—to test the effects on the governor’s standing in a two-way primary that would no longer take place. Carney huddled with the eggheads and considered their options. Green argued that, if anything, Strayhorn’s withdrawal was good news for their experiment and that they should continue as planned. Before her move, the eggheads were unsure of how they would deal with the noise that would come from a lively primary—such as the possibility that Strayhorn could focus her ads in specific markets—noise that could distort their ability to isolate the effects of
Perry’s randomized buys. All they would do to adjust is open up the polling to all voters and ask them whether they viewed Perry favorably, then test his performance in a four-way general election. Carney decided to strike the last of the four weeks he had allocated for the experimental buy, to stockpile some money for the fall, but gave the eggheads approval to book ad time.

The morning of January 5, Texans awoke to triumphant coverage of the University of Texas Longhorns’ dramatic upset victory in the Rose Bowl’s closing seconds to claim the national championship and seal an undefeated season. Television viewers in five media markets across the state were given another new reason to feel good about their state: Perry’s first ad of the campaign had slipped into the regular rotation on their local news. “I’ve never been more proud to call myself a Texan,” said the tall, strong-featured man standing before the state capitol, where he kept his office, and fighting for screen time against an immodest parade of Lone Star clichés: waving bluebonnets, the Alamo, cattle patrolling a central business district. The governor recounted his policy successes in education, tort reform, and economic development. “Our people are compassionate. Our vision, bold. Our values, strong. The best is yet to come,” he went on. “I’m proud of Texas. How ’bout you?”

The thirty-second ad was standard introductory fare, but the scattered schedule on which it aired looked nothing like the order that David Weeks, who scripted and shot it, would have put in to his media buyer under normal circumstances. Political consultants calculated their buys in terms of gross rating points, or GRPs, the basic unit of measurement for TV exposure; the rule of thumb was that with 1,000 GRPs the average viewer saw each spot ten times in a week. Like most consultants, Weeks—who had begun his career as a campaign spokesman and press secretary—had a set of rules about the size of his ad buys, all grounded in a mix of tradition and instinct. “An ad has to have a certain amount of weight behind it,” he says. An attack ad, Weeks believed, needed between 800 and 1,000 GRPs to have an impact; “you can get by” with 500 to 600 points when pushing a positive message. And Weeks adhered to a widely accepted maxim about
the importance of keeping a continuous, steady presence on the air for his candidate: “When you go up, you stay up.”

The eggheads had designed the experiment to deliver its treatment in three doses: 1,000, 500, and 250 GRPs. Markets were assigned their volume randomly, each one rolled out in a staggered fashion so that at any point different markets would be getting Perry’s ads in different volumes, some receiving none at all. (To make the study as unobtrusive to campaign strategy as possible, the eggheads agreed to remove the state’s two largest cities, Houston and Dallas–Fort Worth, from the experiment entirely, leaving eighteen media markets. These varied widely in terms of size and demographics and so the academics matched similar ones for balance, reducing the likelihood that the randomizer could end up commanding Perry’s media buyer to shrink its presence in all of conservative West Texas at once, for example, or effectively give up on speaking to Latino voters statewide.) Weeks had never bought only 250 points in a market for Perry, as the campaign found itself doing in Odessa-Midland. At such a low volume, the “Proud” ad found its way into the early-morning local news,
Live with Regis and Kelly
, and a pair of early-evening game shows—all prime hours to find older viewers who make up the most loyal voters. In Victoria, which received a 1,000-point buy starting the same day, Perry’s declaration of state pride cut a much wider swath through the programming grid, including
Oprah, Law & Order, Extreme Makeover: Home Edition
, and an NFL playoff game.

Green would look at the nightly results of one thousand short polling calls being made statewide to see if the ads were having any effect. During the first week of the experiment, Perry’s statewide approval rating was just under 45 percent, and he led the four-way matchup (against Strayhorn, Friedman, and an unnamed Democrat) with 34 percent of the vote. By January 26, every market in the study but one had received Perry ads, although in different volumes, durations, and rhythms. (Abilene had been randomly assigned zero points in each of the three weeks.) Victoria received 1,000 points the whole time, Odessa three weeks at 250, and San
Antonio got nothing the first week but two weeks at 500. At the end of the third week, in which the ad appeared in seventeen markets, the polls registered a significant improvement for Perry: his statewide approval rating had moved up to 47 percent, and a much larger share of these registered as “strongly positive” than they had at the start of January, although he fell slightly in the four-way horse race. When the eggheads took apart these numbers to reflect the broadcasts’ distribution and timing, they found that the ads had “sizable effects” on the way voters perceived Perry. His support increased nearly five percentage points per 1,000 GRPs in weeks when his ads aired.

At the end of the third week, the ads stopped. Although Perry did other things befitting a statewide candidate, he ran no more broadcast ads before the primary on March 7. For the final few days before the primary, Green ordered more polling calls. Perry was significantly more popular than he had been just after the new year, with an approval rating over 55 percent. But the eggheads’ regression analysis gave the ads no credit for that movement. Even in Victoria, which received the maximum dose of 1,000 GRPs and saw Perry’s standing jump by six percentage points over those three weeks, the ads’ power had worn off by March. The ads may have delivered sizable effects on the weeks in which they ran, the eggheads concluded, but they decayed rapidly. Much of Weeks’s folklore was right: if your goal was to move public opinion, it made sense to wait to go on TV until you would be able to sustain the buy.

Shaw thought that was an important conclusion, but he believed that to reach it Green had made a major trade-off. By staggering the buys for a nuanced understanding of how opinions move over time, the experimenters had decided to be in fewer markets on any given day. As a result, Shaw thought, they had sacrificed the power of the data to illuminate how much advertising moved public opinion at all. Shaw had argued that the campaign should run two different ads as part of the experiment, to be able to measure their relative impact, and lost out to Green. “The real weakness of the study is it’s a single ad, and you’re extrapolating from that,” says
Shaw. “You could make an argument that okay, there’s a diminution of effects, but that’s just because it’s not a very memorable ad.”

Much of Shaw’s personal research addressed one of the most vexing questions in the study of elections: do campaigns matter? The first postwar studies, like the Michigan scholars who pioneered the National Election Studies, argued that the factors that drove voters’ choices—their social networks or party identification—moved more slowly than any individual election season. It suggested that the quadrennial pageant of candidates crisscrossing the country was all but incidental to an outcome that was effectively preordained. In the 1970s,
political scientists believed they could explain all presidential elections with the same formulae. The most important inputs were presidential approval ratings and national economic indicators, reduced by statistical models to a neat set of rules: if the economy was growing faster than 1.2 percent annually, the incumbent president would win. It didn’t matter how much money he raised, if his ads were any good, or whether his opponent looked presidential. The winner had basically been determined by June.

Shaw’s own experience gave him good reason to be skeptical of this thesis that campaigns had “minimal effects.” By its standards, Shaw’s boss in the 2000 campaign should have been a certain loser. Al Gore entered the nearly two-year election cycle as the tribune of a party with a popular incumbent in the White House, presiding over a very healthy economy.
An average of seven academic election-forecasting models (using these criteria and others) presented at the American Political Science Association conference that August predicted that Gore should have carried 56 percent of the popular vote against Bush in a two-way contest. But by November, the Republican had turned the election into nearly a tie. (Gore won 50.2 percent of the two-party vote, leading Bush by a little more than 500,000 votes.) Some of those in the minimal-effects school
responded by adjusting the economic indicators that went into the models or arguing that Gore’s inability to focus on the economy mitigated the importance of those conditions to voters. But Shaw thought that basically conceded
the point. Bush had run a better race than Gore. What they did mattered. Campaigns had effects.

As a graduate student, Shaw had been unsatisfied with previous works that had attempted to answer this question by relying only on survey data or laboratory experiments attempting to measure what types of political communication could shift voters’ preferences. “
While experiments allow researchers to isolate the influence of particular effects, critics question whether such a controlled environment is comparable to the hurly-burly of an actual campaign,” Shaw wrote. Instead, he reconstructed candidates’ electoral-college strategies, relying on professional contacts where necessary to understand how past campaigns had selected the states in which they chose to compete and how they allocated resources among them. In the 1980s, presidential candidates largely stopped buying national ads from the three networks and started going to local affiliates in their targeted states. As a result, campaigns began to rigorously divide the country into battleground and nonbattleground states—devoting time and resources to the former and completely avoiding the latter. It created, in effect, a natural (if nonrandomized) experiment that Shaw could test, and he was able to determine that the amount of money a candidate spent in a state did affect the share of the vote he won there, albeit in a minor way.

But TV ads were only one of the “most obvious and visible manifestations of the campaign,” as Shaw wrote in his dissertation. The other was the candidate’s own travel—the whistle-stop tours and bus trips and swelling rallies that provided the country’s dominant political iconography and drove daily news coverage within an election year. What was more of a campaign effect than a visit from the candidate? Major campaigns, especially those covering enough turf that transit time was a consideration, negotiated the candidate’s itinerary with more attention than anything else. “At the end of the day there are two finite resources in every campaign: money and time,” says Deirdre Delisi, “and the most valuable time is the principal’s time.” It was possible to put a price on an hour of the candidate’s day, whatever he or she could bring making fund-raising calls over
that time. But no one knew how to weigh that against the political value of shaking hands in a diner or giving a speech to a neighborhood association.

American history may have been filled with anecdotes of campaigns swayed by the schedule, like how in the last days before the 1960 election, Richard Nixon, holding to a pledge to visit every state, flew to Alaska instead of trying to sway voters in deadlocked Illinois. But in the current era of ubiquitous television and Internet coverage, wondered Shaw, did a politician’s physical presence have any impact?

ON JANUARY 10
, Rick Perry strode into the Frazier Alumni Pavilion, a shed covered with Spanish-style clay roof tiles and attached to Texas Tech’s sixty-thousand-seat football stadium. On autumn Saturdays the pavilion was popular for fancy pregame parties, but on a Tuesday morning in January it housed one hundred supporters, many of them wearing Perry 2006 stickers. “There’s been seven times I’ve stood before the people in this state and asked them to place their trust in me,” Perry said, as university flags fluttered from wooden rafters overhead, “and I have never been defeated.”