The Victory Lab (5 page)

In the spring of 1923, Merriam approached Gosnell with an idea for a joint research project that would merge the older man’s activist agenda with the younger’s interest in modern research methods. Chicago voters had just booted Big Bill Thompson for a Democratic judge named William E. Dever in a mayoral election that seemed to dominate citywide attention. Indeed, the conversation in political circles focused on the fact that turnout was surprisingly high. But Merriam, fueled by intellectual curiosity and residual bitterness over his own loss to Thompson, turned his attention to those who never cast a ballot. During his own campaigns, Merriam had worked on expanding the electorate by recruiting new voters, especially among new immigrant arrivals who had yet to fall under the machine’s spell. Now the city had about 1.4 million adults, but only 900,000 were on the electoral rolls; among them, 723,000 cast a vote for either Thompson or Dever. It galled Merriam, as it had during his own campaigns, that barely half of the city’s eligible voters had been involved in picking their leader; the apathy of the rest helped keep the machines in power. He suggested to Gosnell that they investigate the reasons the nearly 700,000 nonvoters had for opting out—and what might be done to lure them into the process.

Gosnell prepared a survey to ask them, relying on U.S. Census data to guide him in each of the city’s fifty wards toward a representative mix of respondents. Gosnell and Merriam decided to use a hybrid survey, which would have multiple-choice questions but leave room for free answers, an approach they believed should yield a healthy batch of data but also qualitative responses with richer texture. All interviews would be done face-to-face at
people’s homes, so Gosnell had to train graduate students to navigate the city’s racially and ethnically complex neighborhoods, where he worried they might not find a warm welcome for outside researchers asking nosy questions. Gosnell dispatched a Swedish-speaking student to a heavily Swedish neighborhood, and hired a Polish interpreter elsewhere. (Of the sixteen doors the professor knocked on himself as part of the project, one happened to belong to writer Ben Hecht.) The researchers’ forms were then coded, and the data moved onto punch cards so they could be tallied by machine. The university did not have the proper equipment, so Gosnell went to city hall and found a clerk in the comptroller’s office willing to run the cards on his own time for one dollar an hour. When Gosnell looked over the six thousand answers his students had gathered, he was pleased by one particular sign of their diligence. Those assigned to the city’s so-called Black Belt, where Gosnell had feared that the response rate would founder, had been so aggressive that African-Americans now overwhelmed the sample. Gosnell removed some of them to maintain the delicate demographic equilibrium essential to the project’s credibility.

Gosnell’s findings, with edits by Merriam, were assembled under both men’s names in
Non-Voting: Causes and Methods of Control
. Published by the University of Chicago Press in August 1924, the book—released just months before a presidential election—received national attention for trying to explain the fact that women’s suffrage had not dramatically increased voter participation. (Gosnell found
twice as many women as men who didn’t vote.) But Gosnell’s conclusions, that “
general indifference” led people to stay home, made less of an impact than his technique. It was the
first major political science study to rely on random sampling in a way that broke down the sample by different demographic attributes.

“
If scientific methods seem hitherto to have found too little favor with American politicians, political scientists must admit that they themselves are largely to blame,” Harvard professor A. N. Holcombe wrote in a short but enthusiastic article in the
American Political Science Review
. “But on the basis of this first experiment at Chicago it ought to be possible,” Holcombe
suggested, to draw conclusions about elections “with all the assurance of a chemist proving the quality of a new paint-remover or a biologist testing a germicide.”

Gosnell was already thinking in those terms. He had begun meeting with social psychologists who recommended tools that would allow him to find out what, if anything, could change nonvoters’ behavior. The psychologists explained the rudiments of a field experiment: Gosnell could introduce what they called “controlled stimuli,” in this case reminders of a coming election, and then measure their effect. By setting up a control group, whose members did not receive the treatment, and comparing their vote performance against the rest, Gosnell would be able to measure whether various appeals could turn people into voters.

This was the scientific method at work, and despite their title no political scientist appeared to have ever tried such an approach. Gosnell had observed nonvoters and had theories about their behavior based on why they said they did not participate. Now he became convinced that only a randomized-control experiment would allow him to see if anything could change that. This ambitious agenda was making Gosnell’s research a lot more expensive than the typical office work practiced by traditional politics scholars. The 1923 poll had cost five thousand dollars. Even in Judson’s absence, getting the administration of Robert Maynard Hutchins, one of his successors, to back research forays into the messy world of urban politics was not easy at a school Hutchins was elevating into a global citadel of canonical study. “We were hopeful that democracy could be made to work,” Gosnell wrote. “But President Hutchins thought otherwise. All worthwhile ideas were to be found in the Great Books. Social science research in a metropolis was trivial.”

But the Chief had his own sources of money and so started shaking the trees for his protégé, approaching his former campaign donors and business leaders. The most lucrative avenue was the Rockefeller family, whom Merriam reached through Dr. Beardsley Ruml, a psychologist, PR man, and Macy’s department store official who as a Roosevelt administration
official later helped design the country’s first withholding system for federal income taxes. Gosnell was dazzled by Ruml, privately sketching caricatures showing the “financial genius as bargain basement statue of Buddha,” able to bring Rockefeller cash into Merriam’s account to support further research.

Gosnell conceived his experiment as a two-stage study: the first would measure whether citizens who were not registered could be pressured to sign up, and the second would test what could be done to get already registered voters to turn out at a higher rate. Gosnell identified six thousand adult citizens scattered across twelve Chicago zones, and arbitrarily divided them into two groups, checking to ensure that they looked demographically similar. One group would be his treatment sample and the other his control. “The study was aimed to give an answer to the question whether the non-voter is such by a deliberate act of will or whether he is a non-voter from ignorance but not a deficiency of public spirit or alienation,” Gosnell later wrote.

In the fall of 1924, Gosnell sent postcards emphasizing the importance of registering to vote before the presidential election that November. (In addition to English, Gosnell drafted versions in Polish, Czech, and Italian.) The postcards had their intended effect: people who received them were nine percentage points more likely to register. Then Gosnell prepared another set of two postcards for the 1,700 voters who were unmoved by the first appeal, one with another nonpartisan message about the urgency of registration and the other with a cartoon picturing nonvoters as “slackers who fail their country when needed,” according to the caption. Both pushed people to register at a higher rate than the original control group. In the end, 75 percent of those who received at least one of Gosnell’s cards ended up registering, while only 65 percent of nonrecipients did. He had put about three hundred new voters on the city’s rolls who wouldn’t have been there otherwise.

The next February, Chicago would elect aldermen, as the fifty members of its city council were known, and Gosnell set his sights on the nearly
2,200 new voters who had registered after receiving one of his notices. (Most, he knew, would have registered without his intervention.) Gosnell drew up another cartoon, this time depicting “the honest but apathetic citizen as the friend of the corrupt politician.” Again Gosnell left his mark on the election: 57 percent of those who got the cartoon turned out to vote for alderman, compared with 48 percent of those who didn’t.

Gosnell did his calculations by hand, and as he looked more closely at these numbers, he realized that his mailings were most persuasive among new residents, who Gosnell concluded had few other sources of information on how to vote, and in districts where party organization was weakest. In demographic terms, they had the most impact on “the native-born colored women and the women born in Italy,” wrote Gosnell. The reason, he found, was that the League of Women Voters was directing most of its attention toward native-born white women and little toward minorities. Gosnell’s conclusions were obvious—mobilization efforts can have the biggest impact in places where little else is pushing voters to the polls—but no one had ever before quantified them.

It was
likely the first field experiment ever conducted in the social sciences outside psychology, and it was well received when published in book form, as
Getting Out the Vote
, in 1927. Political scientist George Catlin wrote that Gosnell’s study “
has the high merit of being precisely a scientific social experiment.” This time Gosnell’s innovation jumped from scholarly journals into the news pages. “
This study is not only a model of careful method in a virgin area of political exploration,” Phillips Bradley wrote in a
New York Herald Tribune
review, “but offers some pretty plain evidence that what has here been done privately in the case of a few thousand voters should become a regular part of our official election procedure.”

But Gosnell never ran another experiment. In the 1930s, he turned his attention to pioneering studies in black politics and machine organizations, goaded on by Merriam’s continued bitterness about the forces he believed had unfairly denied him his place at city hall. “
Perhaps Mencken is right,” Gosnell consoled his mentor. “The people usually vote for crooks.”
Despite the enthusiasm that greeted Gosnell’s method for studying campaigns, no one tried to copy him, replicate his study, or build upon it. After printing Gosnell’s article, the
American Political Science Review
did not publish another finding from a randomized field experiment for a half century. During that time, political science grew into a major discipline obsessed with studying voters and elections, but to do so it returned to the library and stayed off the street.

THE FEW EFFORTS
by political scientists to revive Gosnell’s experimental technique proved evanescent. In 1954, University of Michigan professor Samuel Eldersveld
used new statistical methods to dispatch mail, phone calls, and in-person canvassing visits across eight hundred Ann Arbor residents according to a random-assignment procedure, and then measured their relative effectiveness on turnout. Eldersveld’s experiment had more impact on local politics—three years later, he succeeded where Merriam had failed and was elected mayor of the college town—than on the academy. Afterward,
entire decades would pass without a single randomized field study about political behavior being published in a scholarly journal.

Political scientists didn’t take to experiments in part because they knew that they would never control the laboratory. The party machines that dominated most American political activity lacked the self-examining impulse, and were unlikely to welcome ivory-tower visitors into their clubhouses. Meanwhile, campaign finance laws and the universities’ nonprofit tax status made it hard for them to do anything on their own that, even inadvertently, advanced the interests of a specific party or candidate.

Political scientists instead happily flapped about in deep pools of new data generated by a postwar revolution in research methods. The ubiquity of household telephones made large-scale survey-taking possible, and increased computing power permitted complex statistical regressions.
Specialists in the new field of polling developed protocols for assembling interview samples that would reflect the broader population, and for scripting survey questions to make sure they elicited meaningful responses. Everyone started doing polls, but quality was inconsistent. In 1948, most pollsters flubbed their electoral predictions—leading to the
Chicago Tribune
’s morning-after “Dewey Defeats Truman” front page—because they stopped talking to voters in the race’s closing weeks, therefore failing to pick up on a late movement toward the incumbent.

One of the pollsters who did not make that error was Angus Campbell, a social psychologist who had spent the war years in a research office of the Department of Agriculture, modeling how consumers would react to the conflict’s end so that policymakers could anticipate what they were likely to do with their war bonds. In 1946, Campbell and several colleagues decamped to Ann Arbor, where the University of Michigan built a new Survey Research Center around them. After the 1948 election, Campbell ran a post-election survey to make better sense of Truman’s comeback. As 1952 approached, Campbell mapped an ambitious plan to track the attitudes and opinions of the electorate, unfurling a series of lengthy questionnaires that would be used to interview voters nationwide throughout the election season.