Social Science and the Promise of Polling
In 1935, the New York Times observed that “professors who revel in a study of the unknown have the listening habits of the unseen radio audience as something ... new to analyze.”89 The occasion for the article was the publication of a new book, The Psychology of Radio, written by social psychologists Hadley Cantril and Gordon Allport. As the authors described in the preface, their goal was “to map out . the new mental world created by radio.”90 In particular, Cantril and Allport believed that radio “interposes a serious psychological barrier between the broadcaster and his audience,” resulting in “a marked distance between the listener and the speaker.”91 This new distance created challenges for those who wished to communicate with the public, but it also presented opportunities. Someone like Father Coughlin or Huey Long, Cantril and Allport noted, uses “less bombast and more artistry, less brute force and more cunning. He [directs] his attention to the invisible audience and [makes] each listener feel welcome as a member of the circle.”92 These radio spellbinders developed new ways to communicate, and understanding their appeal to the public required new ways of knowing. “Radio is a novel means of communication ... requiring novel methods of investigation,” Cantril and Allport insisted.93 Combining surveys, census data, and the results of psychological experiments, Cantril and Allport offered, as the Times put it, key insights into “the psychology of the invisible multitude.”94
Cantril’s interest in the effects of mass communication drew him to the study of survey research just as George Gallup and Elmo Roper began popularizing their poll results in syndicated columns and in Fortune magazine. As Cantril later recalled, the work of Gallup and Roper convinced him that “here was a new instrument the social scientist ... had better look into.”95 Cantril got his chance in 1936 when the New York Times asked him to write about the use of polling to predict the upcoming presidential election.96 Cantril traveled to Princeton, New Jersey, where he met with George Gallup. Recounting the meeting, Cantril wrote that Gallup “was delighted to have a social scientist take his work seriously, and offered his facilities at cost for any research I might want to do.”97 The meeting would begin a long friendship. For Cantril, the opportunity to make use of Gallup’s facilities was so attractive—the pollster employed a staff of thirty and nearly a thousand part-time interviewers across the country—that Cantril accepted a position in the Princeton psychology department the following year.98 For Gallup, on the other hand, forging ties with an academic social scientist boosted his status as a public opinion professional.99
At Princeton, Cantril advanced the academic study of public opinion even as he solidified his ties with a budding polling industry.100 In 1937, Cantril joined Princeton political scientist Harwood Childs in launching Public Opinion Quarterly. In its initial incarnation, the journal served as a forum for both the academic and the commercial side of survey research. As the editors described their mission in the first issue published in January 1937, “The understanding of what public opinion is, how it generates, and how it acts becomes a vital need touching both public and private interest.”101 The Public Opinion Quarterly was a big tent indeed. Its first issue included contributions from editorial board members Harold Lasswell and Pendleton Herring as well as pollster Archibald Crossly and public relations counsel Edward Bernays. In its first year of publication, in fact, a third of all articles in the journal were written by authors with a business rather than academic affiliation.102
Cantril was also a central figure in a scholarly network formed under the auspices of the Rockefeller Foundation that examined the effects of radio on public opinion. In 1937, a Rockefeller grant established the Princeton Radio Research Project. Its goal was to apply commercial survey research methods to the study of radio and its psychological effects.103 One of Cantril’s first efforts for the project used Gallup surveys to study the panic that ensued following the radio broadcast of “War of the Worlds” by Orson Welles.104 In 1939, Cantril secured another grant from Rockefeller, this time to establish the Princeton Office of Public Opinion Research, the first university institute of its kind solely devoted to survey research. Its purpose was to examine “the development of opinion, the changes which opinion undergoes under varying conditions, and the reasons for change.”105 Specifically, Cantril proposed taking interview data collected by Gallup so he could study the influence of radio, print, and film on public attitudes toward the war in Europe.
Although public opinion research would eventually enter the social science mainstream, survey methods raised eyebrows in the 1930s and 1940s. To its critics, polling was a commercial tool better suited to marketing studies or crude election forecasts than to meaningful investigation into public thinking on important matters.106 In fact, many of the scholars who first embraced the study of public opinion were closer to commercial pollsters than their academic colleagues in the social sciences. Lacking the basic infrastructure to conduct their own surveys, scholars like Cantril depended on commercial pollsters like Gallup and Roper, who provided access to their national network of interviewers scattered across the country. More fundamentally, academic survey research rested on a set of assumptions about human behavior that differed little from their commercial counterparts. Just as advertisers wanted to know why consumers bought certain products, students of political behavior wanted to know why voters supported certain candidates. As Cantril noted in a 1940 article, all forms of strategic communication employed emotive symbols such as “justice, beauty, [and] liberty” whether it was intended to promote “cigarettes, political campaigns, [or] appeals to join the army.”107 The purpose of a survey, in other words, was to understand the architecture of individual choice. As the sociologist Paul Lazarsfeld explained, surveys assumed “the methodological equivalence of ... voting and buying soap.”108
Lazarsfeld’s career illustrates this overlap, in both substance and method, between consumer research and political behavior. As a young scholar in Vienna, Lazarsfeld became interested in social stratification, its psychological effects on individuals, and, above all, the kinds of methods one could employ to study them. It was at this time that Lazarsfeld first came across marketing studies that employed survey methods, and he quickly incorporated them into his own research on social psychology. In 1934, Lazarsfeld arrived in the United States on a fellowship from the Rockefeller Foundation, and in 1937 he became the director of the Rockefeller-funded Princeton Radio Research Project. In 1940, Lazarsfeld moved to Columbia University, where he began work on his landmark study of voting behavior, The People’s Choice.109 Lazarsfeld’s interest in voting was part of his broader research on mass communication, using the effects of campaign messages to study “how and why people decided to vote as they did.”110 Focusing on Erie County, Ohio, Lazarsfeld used a sample of 600 respondents who were interviewed seven times over the course of the 1940 presidential campaign. Thus, the election panel study was born. In fact, the panel method was already a common tool of market research used to save time and money by reinterviewing the same people. Lazarsfeld’s insight was to use the panel method to study the “psychology of choice” and how the decision-making process unfolded over time.111 Writing several years earlier in the Harvard Business Review, Lazarsfeld described consumer behavior as a sequence of steps that culminated in the purchase of goods. Survey research disaggregated the “structure of the purchase” and rendered it intelligible by revealing the psychological influences behind the decision of what to buy.112 Lazarsfeld saw voting much the same way. Moreover, elections offered methodological advantages over consumer research because the sequence leading up to the “purchase” culminated on the same day for each individual and under a relatively controlled environment of media and other stimuli thought to influence the vote. Elections, in other words, offered a chance to study “a large-scale experiment in political propaganda and public opinion.”113
As it turned out, the results of the Erie study provided very little evidence that radio or other forms of political communication influenced the vote. Most of those surveyed made up their minds early in the campaign, and only very few, 8 percent, switched their support from one party to the other. Lazarsfeld and his colleagues also found that socioeconomic status, religion, and rural or urban residence explained a great deal more of the variance in voting behavior than exposure to radio or other political messages. In fact, many respondents paid relatively little attention to the campaign. Of those who were exposed to political propaganda through the newspaper or radio, most had already made up their minds. Voters whose support vacillated or remained undecided until the end—so-called independent voters—were the least politically engaged of those in the study. At best, Lazarsfeld and his coauthors concluded, the campaign reinforced political predispositions or activated latent partisan sentiments. Summarizing his findings in a 1944 article in Public Opinion Quarterly, Lazarsfeld wrote, “Modern Presidential campaigns are over before they begin.”114
Lazarsfeld’s conclusion about campaign effects had important consequences, especially for the polling industry. The mistaken prediction of a Dewey victory in 1948 had a number of causes; however, a critical source of pollster error was the acceptance of Lazarsfeld’s main findings about minimal campaign effects. When Republican candidate Thomas Dewey opened a sizable lead over President Harry Truman after the conventions, Gallup and Roper effectively pronounced the election to be over. Writing in his nationally syndicated column on September 9, Roper, who partially funded Lazarsfeld’s research, explained that “past elections ... have shown us that normally there is little change in the final standings between early September and Election Day.” Convinced that “Thomas E. Dewey is almost as good as elected,” Roper decided to cease all polling on the presidential contest, adding that he would rather “devote my time and effort to other things” than act “like a sports announcer who feels he must pretend he is witnessing a neck and neck race.”115 Although their own recent experiences of presidential elections certainly influenced the decision, Gallup and Roper adopted the position of leading social scientists in their decision to cease polling and incorrectly call the race for Dewey.
Indeed, the debacle of the 1948 election showed just how much the commercial polling industry and academic survey research shared a common fate. Writing in Public Opinion Quarterly in 1949, Robert K. Merton—a close friend and colleague of Lazarsfeld—discussed the “traumatic November episode” and whether it might adversely affect the “public images of social science.”116 That same year, the International Journal of Opinion and Attitude Research published the proceedings of an extended symposium in 1949 on “The Opinion Polls and the 1948 U.S. Presidential Election.”117 Meanwhile, the Social Science Research Council (SSRC) conducted its own investigation, establishing the Committee on the Analysis of Pre-election Polls and Forecasts eight days after the election to discover what went wrong. In its report issued just weeks later, SSRC President Pendleton Herring noted that “extended controversy regarding the pre-election polls among lay and professional groups might have extensive and unjustified repercussion upon all types of opinion and attitude studies and perhaps upon social science research generally.” The report concluded that the decision by commercial pollsters to end operations weeks before the election was a considerable source of error; however, the committee was quick to emphasize that “the public should draw no inferences from pre-election forecasts that would disparage the accuracy or usefulness of properly conducted sampling surveys.”118 If anything, the errors of 1948 illustrated the need for more polling, not less. With the status of commercial survey research basically intact, the crisis proved to be a fleeting one. Roper himself noted approvingly a year after the election that his business had barely registered a blip after the unfortunate events of the previous November.119
As it developed in the 1930s and 1940s, polling combined market research with academic social science to produce a new and powerful instrument of opinion measurement. Such a tool was particularly valuable at a time when the tremendous potential of radio seemed limitless, yet a precise way to reach (and influence) such a vast audience remained elusive. Whether in marketing, commercial polling, or the academy, it was the search for an invisible or unseen public that sparked innovation. Unified by a common goal, academic and commercial concerns overlapped in a diverse network of actors who worked together to perfect their methods and popularize their techniques.
Indeed, with backgrounds that spanned advertising, marketing, and the social sciences, survey researchers held a particular view of politics and the public. The measurement of mass opinion often assumed a rough equivalence in the motivations guiding behavior, political or economic. Scholars such as Lazarsfeld approached the question of how a political campaign shaped the decisions of voters the same way as market researchers studied the effects of an advertising campaign. Attachments to particular products or brands were the methodological equivalents of attachments to particular candidates or parties. As Lazarsfeld and his coauthors wrote in an important summary of the literature in 1954, “The decisions that a modern Western man makes every few years in the political arena are similar to those he makes every day as a consumer of goods and services.”120 Building on these analogies with market share and product loyalties, pollsters brought both the language and the tools of business into the conduct of politics.
More fundamentally, public opinion polls offered a way of “selling” candidates and issues to the public, trimming political messages so that they aligned with the preferences of a particular audience. This was an important step in the commercialization of political advice. In time, polling became inseparable from media as those who mastered the technical demands of survey research would join the experts in mass communication to occupy a privileged role in the conduct of campaigns.
This union did not occur right away. Pollsters and media experts would have to work hard to establish a niche for themselves in a market for political intelligence still dominated by journalists and party workers. In the meantime, the marriage of media and polling found a home in a presidency increasingly dependent on the mobilization of the public to achieve its policy goals. The use of poll-tested messages by the Roosevelt administration demonstrated the promise of combining modern survey methods with crafted communication, as well as the opportunities for a new kind of practitioner to establish a position of influence in political affairs.