The Development of Commercial Polling
By applying the tools of market research to a host of social and political questions, George Gallup built a successful enterprise measuring the public pulse. Although polling on social and political questions had commercial potential, as Sarah Igo reminds us, “opinion entrepreneurs” like Gallup “were first and foremost market researchers, devoted to the science of improving corporate profitability through carefully crafted advertising campaigns and public relations strata- gems.”57 As Daniel Robinson has written, “Opinion polling developed conceptually and methodologically largely as an adjunct of consumer surveying itself.”58
In fact, Gallup and other polling pioneers like Elmo Roper began their careers in advertising and market research before establishing commercial firms of their own. Gallup, for instance, first became interested in polling while helping to conduct a house-to-house survey of readers’ likes and dislikes for the St. Louis Post-Dispatch. The experience informed Gallup’s 1928 PhD dissertation in psychology, entitled “An Objective Method for Determining Reader Interest in the Content of a Newspaper.” Using a quota sample of 1,000 Iowans, Gallup discovered that most people read less than 15 percent of the paper, focusing more on cartoons and photographs than on front-page news. Gallup’s findings attracted attention, particularly in the field of advertising, landing him a position in New York with the Young & Rubicam advertising agency.59
It was while working for Young & Rubicam that Gallup perfected his methods. Fielding polls in his spare time, Gallup made his first political poll in 1932 on behalf of his mother-in-law, Ola Babcock Miller, who won a narrow victory as Iowa secretary of state. Looking back, Gallup credited Miller’s candidacy with helping him become “interested in the whole possibilities of polling” and their potential use across a wide range of areas.60 In 1935, Gallup established his own private firm, the American Institute of Public Opinion (AIPO), with offices in Princeton across the street from the main entrance to the university—a name and location that lent Gallup an air of academic credibility. Gallup began producing a syndicated feature, America Speaks, which reported polling results on various issues of the day. By the end of its first year in circulation, AIPO polls appeared in sixty major metropolitan newspapers across the country.61
The 1936 election set the cornerstone for what became a polling empire. In July, less than one month after the nominating conventions, Gallup reported the results of a nationwide survey showing a slight lead for Roosevelt over Kansas Republican Alf Landon. Gallup’s prediction ran counter to the famous Literary Digest straw poll, which forecast a Landon victory in November. Straw polls had been widely used to gauge public sentiment since the nineteenth century, and the Digest poll was the largest and most accurate of its kind. In 1932, the magazine sent out nearly 20 million ballots to its readers and came within three-quarters of 1 percent of predicting Roosevelt’s margin of victory.62 However, the
Digest mailed ballots chiefly to households listed in telephone directories or with automobile registrations, a method that drew responses overwhelmingly from upper-income voters. Gallup understood that this flawed methodology overrepresented the extent of Landon support in the electorate, and he believed his quota sampling method based on a representative cross section of the voting public would yield a more accurate prediction of the race. In a rather daring gambit, Gallup backed up his claims with the offer of a money-back guarantee to the more than seventy newspapers subscribing to his column if his prediction turned out to be incorrect.63
As the campaign unfolded, Gallup issued a steady stream of poll results that tracked the overall state of the race as well as the constituent parts of each party’s coalition.64 The AIPO polls gave a new look to the “horse race” of the campaign by emphasizing the battleground states where the Electoral College would be won or lost. Tracking polls followed FDR’s support over time, offering what Gallup boasted was “the first graphic picture of the ups and downs in a President’s majority ever available in history.”65
Gallup also explained the procedures he used in an effort to convince the reader of the accuracy of his methods.66 More was at stake for Gallup in 1936 than the outcome of the race. As the Washington Post observed, “When American voters give their verdict on Election Day, it will not be political questions alone that are decided. At least one scientific question will be decided too. ... ‘How far can science go in predicting the outcome of an election?’ ”67 On November 1, in his final column before Election Day, Gallup wrote that “Tuesday’s election will show which procedures for polling public opinion can be relied up [sic] with greatest faith in the future—the mass-balloting method or the scientific sampling method.”68 Although Gallup’s prediction of a Roosevelt victory with 54 percent of the popular vote and 477 electoral votes was off by a fairly wide margin, it was far more accurate than the Digest poll, which called the election for Landon. The following week, the Post reported that the election result “vindicates the scientific sampling technique.”69 Meanwhile, Gallup triumphantly announced an expansion of AIPO polling in the coming year on a wider range of subjects, adding that “if public opinion is to rule in a democracy, there must exist the mechanics for reporting it.”70
The 1936 election offered a very public demonstration of the reliability of scientific sampling, as well as the commercial potential of polling. However, Gallup was not the only one to recognize these opportunities. Elmo Roper also built a successful polling business out of a career in market research, although he began somewhat by chance. Crisscrossing the country as a traveling salesman for a jewelry manufacturer, Roper concluded that the product line he sold fell between two market segments: too fancy for farmers, yet not fancy enough for an urban clientele. Roper later described his analysis to Richardson Wood, a friend and advertising executive at J. Walter Thompson, who encouraged Roper to consider a career in market research. When Roper moved to New York a short time later, Wood introduced him to Paul Cherington, who had recently left Thompson to work as an independent consultant. In 1934, the three established their own market research firm, Cherington, Roper, and Wood (CRW). The firm’s promotional materials offered “to supply to management significant facts ... from a representative sampling of consumers.”71
Much of the early work of CRW used the tools of market research to inform corporate public relations. With the rising activism of the New Deal, the threat of government regulation made selling the corporation to the public just as important from a business standpoint as selling the product itself.72 Power companies in particular sought out CRW’s services as they battled against proposals for public ownership of utilities. For instance, facing a referendum on municipal control of the water supply, one utility company hired CRW to study its public image. When the survey revealed that most consumers were happy with their service, the utility company launched a campaign touting the benefits of private ownership and eventually defeated the referendum. Roper described the survey as “our first piece of political research,” and it illustrates how the tools of market research could be used to measure public sentiments on a range of political and social matters their corporate clients deemed important.73
Seeing this potential, in 1934 Wood secured a major new client when Fortune magazine, the country’s leading business weekly, commissioned CRW to produce a quarterly report on the beliefs, preferences, and opinions of the American public.74 Introducing the first column in the July 1935 issue, Fortune proudly announced “a new technique in journalism,” one that used “the technique of the commercial survey” to provide “a sampling of public opinion” on various issues.75 The Fortune survey quickly became one of the most popular features of the magazine, reporting on a wide range of subjects, from vital issues of the day to the seemingly trivial. The January 1936 report, for example, described the extent of anti-Semitism in various regions of the country and included a fairly sophisticated discussion about the challenges of question wording on matters such as “racial antagonism” that most people were reluctant to discuss openly with an interviewer. Yet, the same issue of Fortune (on the very same page) reported the results of a survey that asked people whether they preferred their beer in bottles or cans.76 Undoubtedly, both kinds of questions were of concern to Fortune readers. By 1938, the survey had become a monthly feature, and Roper, who had formed his own firm after parting ways with Wood and Cherington, was a recognized authority on public opinion.77
Yet unlike Gallup, Roper was reluctant to delve too deeply into the business of election forecasting. In 1936, the Fortune survey reported public sentiments toward Roosevelt on the eve of the election but did not venture an outright prediction.78 When Gallup privately suggested four years later that “those of us who are interested in this job of measuring public opinion ... set up rules by which we could stake our claims for the championship,” Roper politely declined.79 However, the commercial pull was too great, forcing Roper to make predictions, albeit reluctantly. Although motivated by a desire “to protect from harm this infant science of marketing and public opinion research,” Roper wrote to his friend Jay Darling in 1944, “I think the prediction of elections is a socially useless function. ... But apparently those of us who have stuck our necks out in election predictions are expected to keep on doing so.”80 Roper feared that one miscalled election would do irreparable harm to the polling business, a reality he and others would confront four years later in the fiasco of 1948 when the major polling companies incorrectly predicted that Dewey would defeat Truman in the race for president.81 The difficulty, as Roper explained, was that election forecasting required knowledge not only of voter preferences but also of the composition of the electorate on Election Day. Differential turnout rates among various groups made it difficult to draw an accurate sample of likely voters—a challenge pollsters still face today.
However, polling on political questions faced more than just technical hurdles. For commercial pollsters like Roper, elections offered little in the way of financial rewards, particularly when compared with the corporate clients that were the mainstay of the polling business. Roper did accept political clients from time to time. In 1946, he conducted polls in his home state of Connecticut on behalf of Chester Bowles, who was eyeing a run for governor.82 However, political work paid poorly—the fees Roper received from his political clients was less than 1 percent of his income in 1946—when it paid at all.83 Trying to secure payment from a candidate in the 1942 Louisville mayoral race, Roper pleaded, “please, pretty please, and for god sake to send us some money.”84 In fact, Roper was still chasing up $500 from Chester Bowles one year after his election as governor in 1948.85 In other instances, Roper simply offered his services for free. A lifelong Democrat, Roper worked for the Democratic Senatorial Campaign Committee in 1957 as an unpaid adviser, helping raise $100,000 to underwrite polling efforts in every state with a competitive Democratic candidate.86
For these early pollsters, politics was something of a sideshow. Election forecasts could publicly demonstrate the validity of survey methods, as they did for Gallup, or allow one to privately pursue a cause, as was the case for Roper. However, the vast majority of income from polling still came from corporate clients. Despite this fact, commercial pollsters did lay the groundwork for the wide use of polling as a political tool. As Sarah Igo has shown, Gallup and Roper made polls a part of popular culture; the AIPO and Fortune surveys offered a new way to portray the public to itself. In time, survey results would supplant older measures of public sentiments such as letters from constituents, editorials in newspapers, or even the reports of precinct captains and ward heelers as the voice of the people.87 Pollsters, Igo concludes, became, “crucial middlemen in deciphering popular views and determining what public policy stances candidates and organizations should take.”88 However, this way of looking at politics and the public did not go unchallenged. Instead, polling met with skepticism, fueled by a general misunderstanding about the nature of statistical sampling techniques. Many wondered how anyone could divine meaningful information from a survey of only a few hundred people. Pollsters would have to overcome these suspicions before they could establish their supremacy as interpreters of public sentiment.
Social science would play an important part in this development. For an enterprising group of academic researchers, commercial polling offered a way to study mass publics. In particular, academic interest in the effects of radio on political behavior brought scholars in contact with commercial pollsters skilled in survey methods. During the 1930s and 1940s, George Gallup and Elmo Roper offered technical support and advice to leading scholars of public opinion such as Hadley Cantril and Paul Lazarsfeld. United by their belief in the benefits of these techniques, a partnership between social scientists and commercial pollsters simultaneously furthered the academic study of public opinion even as it enhanced the scientific credibility of commercial polling. And when the pollsters failed to predict the outcome of the 1948 election, it was the academy that came to the rescue.