J. Michael Hogan

George Gallup and the rhetoric of scientific democracy

Профессор Майкл Хоган (Hogan,J.M.) – выпускник The University of Wisconsin-Madison, там же он получил Ph.D. в области риторки в 1983 году. 

Его научные интересы сосредоточены на анализе характера и качества обсуждения в обществе различных социальных проблем. В частности им разрабатываются темы, связанные ходом политических кампаний, развитием социальных движений, дебатами по вопросам внешней политики, общественному мнению и опросам.

Его основные публикации: The Panama Canal in American Politics (1986) и The Nuclear Freeze Campaign (1994), под его редакцией вышла книга Rhetoric and Community. Studies in Unity and Fragmentation (1998), объединяющая статьи ведущих специалистов по риторике и дискурсу, идет работа над томом Imperialism and Reform in the Progressive Era.

В The Pennsylvania State University профессор М. Хоган читает курсы:Political Campaigns in the Age of Television, Contemporary American Political Rhetoric, Political Communication and the Media,  ведет семинар Historical Criticism: The Rhetoric of Social Movements и др.  

Предлагаемая ниже статья – является одной из наиболее цитируемых современых работ по методологии исследований общественого мнения, политической риторике и научного наследия Джорджа Гэллапа. Профессор М. Хоган любезно предоставил нам право ее публикации в сети.

Борис Докторов

Throughout his career, George Gallup, the `father" of modern polling, crusaded tirelessly to establish polling's scientific and cultural legitimacy. In public speeches, several books, and more than a hundred articles in journals and popular magazines, Gallup mythologized polling's history of `progress, "deflected doubts about the polls ' accuracy and technical procedures with a rhetoric of scientific mystification, and celebrated the collective wisdom of "the people." Gallup's "rhetoric of scientific democracy" sustained polling's cultural legitimacy, yet it also diverted attention from its most perplexing sources of error and stifled debate over its deleterious effects on the democratic process.

"Everybody believes in public opinion polls-everyone from the man on the street all the way up to President Thomas E. Dewey."

Goodman Ace (as cited in Field, 1990, p. 37)

Polling has come a long way since 1936. In that fateful year, George Horace Gallup, the father of "scientific" polling, predicted the spectacular failure of the Literary Digest's presidential "straw poll." Over the next half century, Gallup's name became virtually synonymous with polling, and today his legacy is an industry of more than 2000 organizations with annual revenues in excess of $4 billion (Walden, 1990, p. xiii). Promoting polling as "the pulse of democracy," George Gallup sold America on a new scientific "instrument" that he promised might "bridge the gap between the people and those who are responsible for making decisions in their name" (Gallup & Rae, 1940, pp. 14-15).

Throughout his life, Gallup preached about polling with "evangelical devotion" (Wheeler 1976, p. 87). In public speeches, several books, and more than a hundred articles in trade publications, academic journals, and popular magazines,[i] Gallup served as polling's unofficial historian as he mythologized the critical incidents in polling's story of "progress." In addition, he campaigned hard to establish polling's claim to science by statistically documenting its record of accuracy and developing lay explanations of its technical procedures. Above all, Gallup promised that polling could make "democracy work more effectively" (Gallup & Rae, 1940, p. 11). Celebrating the wisdom of the "common man," Gallup dreamed of a final stage in the evolution of the American democratic experiment-an age of scientific democracy, in which "the will of the people" would provide continuous direction to policy makers "on all he major issues of the day (Gallup & Rae, 1940, p. 125). For Gallup, polling represented more than an imperfect social science or a profit-making venture. To criticize polling was to criticize "progress," "science," and democracy itself.

Gallup died in 1984, but his American Institute for Public Opinion Research remains an industry leader. Today, the Gallup Poll is syndicated to hundreds of newspapers and conducts "exclusive" polls for CNN, USA Today, Newsweek, and other news media. Meanwhile, Gallup's rhetoric of scientific democracy continues to sustain polling's cultural legitimacy by deflecting attention from its most perplexing sources of error and stifling debate over its deleterious effects on the democratic process. With leaders refusing to lead out of deference to the polls, and with references to the polls often supplanting deliberation over the merits of policies, the polls increasingly shape both the agenda and the outcomes of public debate. George Gallup alone did not create our cultural obsession with polls, nor would he have approved of how politicians and journalists often invoke the polls as a substitute for policy analysis. Yet echoes of Gallup still can be heard in the contemporary defense of polling, and he pioneered many of the polling practices that, ironically, now threaten his own dream of a more inclusive and efficient democracy.

Properly conducted and interpreted, polls might well serve democracy as Gallup envisioned by providing at least one (albeit imperfect) measure of public opinion. In recent years, however, the polls have become, if anything, even less reliable, and the mission of polling has been changed radically by journalistic imperatives. Instead of guiding policy makers, polls have become "news events" in themselves that not only substitute for substantive information, but also fuel so-called "horse race" journalism. In one sense, Gallup's youthful dream has become a reality. Polls are ubiquitous in American public life. Unfortunately, they rarely serve democracy as Gallup envisioned, and what Michael Wheeler (1967) said about the polls nearly 30 years ago is even truer today: "for every good poll there are dozens, perhaps hundreds, of bad ones" (p. 301).

The Historical Folklore

As a young entrepreneur in 1935, George Horace Gallup made a remarkable sales pitch to the newspapers of America: a money-back guarantee that he could predict the outcome of the 1936 presidential election more accurately than the famed Literary Digest poll. As David Moore (1992) has written, this was "no small gamble." If he failed, Gallup faced financial ruin, and the Literary Digest's record had been impressive. In 1928, it accurately predicted Herbert Hoover's landslide, and four years later it came within a percentage point in predicting FDR's victory over Hoover. Despite confidence in his new "scientific" methods, Gallup remained anxious. He had faith in his figures, but as he confided to friends, there was "always a reasonable doubt" (pp. 31-33).

The story had a happy ending, of course. Gallup proved right, the Digest proved wrong, and polling thereafter emulated Gallup's use of "scientific" methods. In large measure, however, the story of 1936 was the product of Gallup's own public relations and ignores one important fact: Gallup also missed the mark badly in 1936. Gallup correctly predicted the winner, but he too significantly underestimated FDR's vote-by nearly seven percentage points.[ii]

Telling the story in his coauthored 1940 book, The Pulse of Democracy, Gallup actually defended the Digest against critics who suspected a Republican conspiracy. The Digest poll had not been "rigged" to favor the Republicans, Gallup insisted; the "sincerity and honesty" of the poll's sponsors were "beyond question." So why, then, did the Digest "go wrong?" The answer lay "not in a lack of honesty. . ., but rather in the fact that in this business of polling public opinion, sincerity and honesty . . . are not in themselves guarantees against errors and inaccuracies." The "disaster" of the Literary Digest poll resulted from its "sampling methods," not from "the morals of its organizers," according to Gallup. Thus emerged the case for his new, more "scientific" methods (Gallup & Rae, 1940, p. 44).

Gallup's "Lessons of 1936" were cleverly misleading. In discussing his own performance (Gallup & Rae, 1940, pp. 44-55), he began by emphasizing more accurate figures that he had released earlier in the 1936 campaign. When he finally admitted that his final prediction missed the mark by seven percent, he claimed to have "erred on the side of caution" and, of course, he still did better than the Digest poll. Gallup also discussed his own poll in the context of other "scientific" polls that did better: the Crossley poll (less than 6 percent error) and the Fortune poll (which came "amazingly close" in predicting the outcome within a percentage point). In mentioning these other "new polls," whose methods also were "based on scientific principles," Gallup took at least some of the credit for those who did better:

The surprising accuracy of the results obtained by these new polls in 1936, bearing in mind that it was the first real test of scientific sampling in a national election, bears witness to the cogency of the criticisms of the Digestpoll, and to the underlying soundness of the alternative methods which were put to the test. (Gallup & Rae, 1940, p. 49)

 Gallup thus designated 1936 as the starting point of the modern era in polling and dis inguished it from a dark age of straw polling dating back to 1824. Even though some of the new polls "weren't quite as accurate as their sponsors hoped," all proved to be far more accurate than the Literary Digest poll, despite their reliance on "only a small fraction of the gigantic sendout employed by The Digest" (Gallup,1940b, p. 23). Gallup marveled at how "in our own day" polling had "developed from a glorified kind of fortunetelling into a practical way of learning what the nation thinks." Describing this "firs  national-election test" as "by no means a final statement of the accuracy of the results which could be obtained by the sampling method," Gallup promised still better things to come (Gallup & Rae, 1940, pp. 5, 55).

Gallup appeared to deliver on that promise over the next decade as he wrote a new chapter in polling's story of "progress" with each new election. Correctly predicting FDR's victories in both 1940 and 1944, Gallup's became America's best-known and most trusted pollster, a man celebrated in a Time magazine cover story as the "Babe Ruth of the polling profession," a "big, friendly, teddybear of a man with a passion for facts and figures" (The black and white beans, 1948, p. 21). A few critics questioned his methods, and some even alleged political biases (see, e.g., Ginzburg, 1944, pp. 737-739), but Gallup (1944) simply dismissed such allegations as "unintelligent, even fantastic." Confident that such an inquiry would put an end to the "misrepresentations and downright falsehoods" and help Americans "become familiar with the technique of polls and their contribution to this government" (p. 795), he even endorsed calls for a congressional investigation of polling.

Then came the "scientific pollsters"' own great disaster: the prediction that Thomas Dewey would defeat Harry Truman by anywhere from five to fifteen percentage points in 1948. When Truman actually won by more than four percentage points, the entire industry came under attack, with Gallup personally taking much of the heat. Initially, an obviously distraught Gallup suggested that bribes and rigged ballot boxes must have altered the outcome and argued, in effect, that the election, not the polls, had turned out wrong (Wheeler, 1967). Soon, however, Gallup joined with a number of social scientists, journalists, and politicians in investigating what went wrong and, in the process, rhetorically transformed the disaster of 1948 into a "blessing in disguise."

The Social Science Research Council (SSRC) sponsored the first of two major investigations (Mosteller et al., 1949). Concerned that the pollsters' miscall might have "extensive and unjustified repercussions," not only on polling but "social science research generally," the SSRC appointed a committee to study "the technical procedures and methods of interpretation" that had led the pollsters astray (p. vii). Completing its investigation in just five weeks, the SSRC committee considered possible errors at every step in the polling process, including sample and questionnaire design, interviewing, problems in dealing with "likely" and "undecided" voters, data processing, and the interpretation and presentation of results. In its final report, however, the committee identified two major "causes of error" in 1948: (a) errors of sampling and interviewing and (b) errors involving the pollsters"' "failure to assess the future behavior of undecided votes and to detect shifts near the end of the campaign" (pp. 299-300).

The first problem, the SSRC argued, stemmed from the "quota method" then used by all the major pollsters in their national forecasts. In that method, quotas were established for respondents with certain attributes, but interviewers selected specific individuals to be questioned. Aside from the problem of determining relevant "attributes," as the SSRC observed, this method caused two major difficulties: "the composition of the population may not be accurately known for the determination of quotas, and the interviewers may operate . . . as 'biased' selecting devices" (p. 84). The SSRC concluded that "probability" sampling, which the pollsters had considered too costly, would have eliminated at least the problem of interviewer bias in selecting respondents.

The second source of error, according to the SSRC, involved a "central problem" in all social science research: how to predict human behavior. With no good way to assess the future behavior of undecided voters or to detect late trends, the pollsters simply ignored "undecided" voters or allocated them arbitrarily to candidates, which created an error that was significant, but "not precisely measurable." The committee predicted that "the gap between an expression of intent and actual behavior" would continue to be a major source of error in election forecasts and that this had to be "recognized as a baffling and unsolved problem imposing serious limitations on opinion poll predictions" (p. 302).

While cooperating fully with the SSRC committee, Gallup challenged its basic conclusions at a second postmortem held a few weeks later at his alma mater, the University of Iowa (Meier & Saunders, 1949). Suggesting that the SSRC committee "felt that it had to point an accusing finger at the pollsters," Gallup declared: "Hindsight is always twenty-twenty!" Having since read the committee's report Gallup still had not seen "any evidence to support the contention that our sampling and interviewing methods accounted for any large part of our error." Gallup had a very different view of what had gone wrong: the pollsters, in effect, had been victimized by their own remarkable record, by the public's belief that polling had "reached a stage of absolute perfection" (pp. 178, 203-204).

Gallup himself preferred to view 1948 "against the background of poll performance recorded in the years since 1935," since the "average error in these forecasts-including those made in the presidential election of 1948"-remained "approximately four percentage points." That, he argued at the Iowa Conference, fell within the expected range of error for polling research "at this stage of its development." Even counting 1948, Gallup insisted that the pollsters had established a remarkable record: "I, for one, marvel that elections in this country can be forecast with an average error of only four percentage points" (pp. 177-178, 182).

To the extent that they erred in 1948, Gallup insisted that pollsters had made "mistakes in judgment and not in basic procedures." At the Iowa conference, he identified two such "mistakes": the failure to conduct "a last-minute poll" and the  ssumption that "undecided" voters "could be ignored and eliminated from the sample." As it turned out, "many voters" apparently had been "lured back into the Democratic fold" in the final days of the campaign, and "many" of the "undecided" voters had, in fact, voted for Truman (pp. 177, 180-181).

Gallup thus turned 1948 into a useful learning experience and, in effect, an argument for more extensive polling. Promising "improvements in every single department" of his operation, Gallup concluded his remarks at the Iowa Conference by again predi ting better things to come: "You may think that I am merely a chronic optimist, but I can't help feeling that this bitter experience of November, 1948, has made a lot of people examine the whole business of polling more critically, and I think that is a real gain" (pp. 204-205). In later years, Gallup routinely referred to 1948 as a "blessing in disguise" (see, e.g., Gallup, 1955-56, p. 237; Gallup, 1972, p. 200). Because it inspired an "agonizing reappraisal," 1948 offered valuable lessons about how to deal with undecided voters, how to screen out nonvoters, how to take account of "intensity and prestige factors," and how to poll "within a couple of days of the election" (Gallup, 1972, pp. 200-01). Gallup never admitted that his sampling methods had failed in 1948, but over the next several elections he quietly expanded his use of "pin-point" or "precinct" probability sampling. By the 1956 election, he had aba doned "quota" sampling altogether and had developed new "machinery" for measuring "trends to the very end of a campaign" (see Gallup, 1957-58, pp. 25-26). By the early 1960s, Gallup could boast that he had bested even his most optimistic predictions of the pre-1948 era by achieving an "average error" of only about "one to one-and-a-half percentage points." "There are still problems," Gallup told an interviewer in 1962, "but . . . I doubt if we can improve our accuracy very much . . . obviously we can't do much better than that" (as cited in McDonald, 1962, p. 22).

Today, the Gallup Poll Monthly features a "Gallup Poll Accuracy Record" that begins with 1936 and calculates the "average error" of the Gallup polls separately for the pre- and post-1948 periods: 3.6 percentage points between 1936 and 1950, compared with only 1.6 percent between 1952 and 1992 (see Gallup oll Monthly, 1992). The elections of 1936 and 1948 thus serve as defining moments in the history of polling by providing benchmarks for statistical demonstrations of improved accuracy and anecdotal evidence of key developments in polling's story of "progress."

Unfortunately, polling's historical folklore has diverted attention from more fundamental questions about polling's claim to "science." While pollsters undeniably have made technical advances in sampling and statistical procedures, they have barely scratched the surface of more basic methodological problems-problems related to what the pollsters euphemistically call "nonsampling error." For more than forty years, George Gallup downplayed these more mysterious and nonquantifiable sources of error by boasting of the industry's record in predicting national elections. In the final analysis, however, Gallup not only misled the public but encouraged his own industry to ignore its most perplexing methodological problems.

A Rhetoric of Science

While pollsters today generally disclaim the predictive value of polls (see Crespi, 1988), Gallup portrayed election forecasts as important scientific "tests." In The Pulse of Democracy (1940), he and Rae wrote: "Pre-election tests are important because they enable surveyors to put this new method of measuring public opi ion through its most exacting test" (p. 81). Several years before the 1948 miscall, Gallup told the House Campaign Expenditures Committee (U.S. House, 1945) that although the polls had not achieved "absolute accuracy," their record did "represent a degree of accuracy found in few fields outside the exact or physical sciences" (p. 1254). Even after the 1948 debacle, Gallup called election predictions the "acid test for polling techniques" and boasted of "a record of which we can all be proud." In "no other field," he insisted at the Iowa Conference, "has human behavior been predicted with such a high degree of accuracy" (Meier & Saunders, 1949, pp. 177, 286).

Gallup conceded that election predictions have "no great social value" (Meier & Sa nders, 1949, p. 286). They were, Gallup (1972) wrote, "the least socially useful function that the polls perform" (p. 123). So why predict elections? Why risk standing "naked before the whole world" the day after an election? For Gallup, the answer lay not so much in the realm of science as public relations: "It was the demonstrated accuracy of election polls, based upon scientific sampling procedures, that convinced the public and office holders that public opinion on social, political, and economic issues of the day could be gauged" (p. 124).

For half a century, Gallup campaigned tirelessly to establish that polling had indeed passed the "test," a conclusion that in turn rested on two distinct claims: polling's record in predicting elections had, in fact, been good, and this record established the reliability of all other polls. Neither of these claims holds up under scrutiny. Both rested on a misleading-indeed, a deceptive and disingenuousrhetoric of science.

Gallup invariably employed the same statistical measure to document the accuracy of his election forecasts: the "average deviation" of those forecasts from actual election results, in "percentage points," on an "all-party basis." Yet not only did this measure ignore the impact of the electoral college,[iii] it also put he best face on a record that, by most other statistical measures, was far from impressive. For example, Gallup boasted at the Iowa Conference of the "surprising" fact that, on an "all-party basis," even his 1948 miscall fell within the expected "margin of error" of four percentage points (Meier & Saunders, 1949, p. 178). Yet calculating the "average deviation" on an "all-party basis" in 1948 meant summing the error in his predictions for each of the two major and the two minor-party candidates, then dividing by four. This procedure produced a respectable "average deviation" of only 2.9 "percentage points." Gallup's error in predicting the vote for just the two major candidates was far less impressive: 4.7 percentage points.[iv]

Looking at 1948 through yet another statistical lens, an even more disma  performance is revealed. In terms of raw percentages, Gallup overestimated Dewey's vote by about 10 percent and underestimated Truman's vote by approximately the same amount. Similarly, in predicting that third-party candidate Henry Wallace would get 4 percent of the vote, Gallup missed Wallace's actual vote of 2.4 percent by only 1.6 "percentage points." Yet  n percentage terms, Gallup missed the mark badly; he overestimated Wallace's vote by 40 percent (Rogers, 1949, pp. 117-118, 124).

Even by Gallup's preferred statistical measure, the pollsters' overall record in predicting presidential elections has been, to say the least, spotty. Between 1952 and 1972, as Wheeler (1976) pointed out, the pollsters predicted only the obvious "landslides," and one could have done just as well simply by picking incumbents to win every election (pp. 39-40). Since Wheeler's study, the pollsters have done even worse. In 1980, they completely missed the Reagan landslide, with most major polls calling the election a toss-up and some even predicting that Carter would win (Shelley & Hwang, 1991, p. 61). In 1984, the pollsters disagreed widely over the magnitude of Reagan's reelection, with figures ranging from Roper's underestimate of 55%-45% to Gordon Black's overestimate of 60%-35% (5% undecided) for USA Today (Crespi, 1988, p. 3). Despite further attempts to tighten their procedures in 1988, only one of the five best-known polls predicted the correct vote for both Bush and Dukakis within three percentage points (Shelley & Hwang, 1991, pp. 67-68). And in1992, the pollsters had their worst year since 1936, as the Gallup Poll overestimated Clinton's vote by 5.8 percentage points. Despite this substantial error the Gallup Poll Monthly's "Accuracy Record" continues to boast of an "average deviation for 28 national elections" of 2.2 "percentage points." Yet, curiously, the "Record" no longer appears in every issue, and it now bears the indignity of an asterisk referring readers to an apologia for the 1992 miscall.[v]

The pollsters' election forecasts have been off so frequently in recent years that even some pollsters now question their own industry's claim to "science." Reacting to the polls' poor performance in 1984, for example, Burns Roper lamented: "I'm very concerned. This raises real questions of whether this business is anywhere near a science." Pollster Irwin Harrison went even further: "There's no science in the questions. There's no science in the interviews. And there's more art than science in the results" (as cited in Crespi, 1988, p. 2).

All this would be reason enough to question whether election forecasting has established the reliability of polling. Yet there is another, more important reason for questioning Gallup's "test": presidential election polls bear very little resemblance to the vast majority of polls. Recognizing that "their reputations, and as a consequence their profits, depend on how close they can come to the actual outcome," the pollsters historically have employed much larger samples in their final election polls,[vi] and they have tightened other procedures as well to avoid being "proven wrong" by the voters (Wheeler, 1976, pp. 38-39). As a result, final election forecasts are not so much a "test" of their general procedures, but instead unique, special events. Indeed, election polls are not only methodologically different from all other polls; they are, as Wheeler (1976) has pointed out, "the best things that the pollsters do" (p. 43)!

Far more suspect are the hundreds, even thousands of issue polls conducted each year-the polls Gallup considered most socially useful. Throughout his career, Gallup (1940b) insisted that, technically, issue polls were "relatively simple" compared to election forecasts, since the latter could "go awry" because of any number of factors affecting turnout (p. 25). Yet the notion that one need not discriminate among respondents in issue polls implies that variations in knowledge and salience have no substantive significance. It suggests no need to distinguish uninformed and perhaps unstable opinions from carefully considered positions, nor the passionately committed from the largely indifferent.

More importantly, issue polls have far more potential than election polls for what the pollsters call "nonsampling error." These sources of error, which range from interviewing problems to flawed interpretive theories, have received far less attention than sampling and statistical techniques in the literature of polling. Yet not only are the effects of such errors often "five or ten times greater" than those of sampling error, as Brady and Orren (1992) have pointed out, but they also are "harder to understand and measure" (p. 68).

A comprehensive review of the various sources of "nonsampling error" is beyond the scope of this essay. Suffice it to say that a growing body of research has documented enormous "response effects" from such "nonsampling" factors as the race, gender, or class of interviewers, the context and timing of surveys, and variations in the order and phrasing of both questions and response options. In addition, researchers recently have called attention to the need for pollsters to develop explicit and more sophisticated interpretive theories. Yet none of these concerns has undermined discernibly the apparent cultural legitimacy of public opinion polling. As may be illustrated by just one example, the problem of question-wording, George Gallup helped shroud such problems in a fog of scientific mystification that only now is beginning to lift.

The "Science" of Question-Wording

Concerned with complex subjective phenomena-beliefs, attitudes, and opinions - rather than a simple choice among candidates, issue polls demand far more attention to question-wording than election polls. Not only must the questions in such polls tap appropriate, relevant opinions and attitudes, they also must mean the same thing to different respondents and inject no bias into the measurement process. In contemporary polling, batteries of questions are often employed to assess not only opinions, but also the awareness and salience of issues, the level and character of existing knowledge, the strength of opinions, and even how opinions might be changed by exposure to information or arguments about an issue. Unfortunately, this only introduces more potential for error into the polling process. Not only does error result from the wording of individual questions, but also from the order in which questions are asked and other factors that affect the context within which respondents interpret particular questions.

Throughout his career, Gallup (1972) paid lip service to problems of questionwording by occasionally conceding that question-wording posed "difficult" and "important" problems for polling (p. 77). At the same time, however, he insisted from the outset that variations in question-wording madelittle difference and that "science" already had solved virtually all problems of question-wording. Devoting barely a dozen pages to question-wording in the 290-page Pulse of Democracy (Gallup & Rae, 1940), Gallup declared that his Institute for Public Opinion Research was hard at work perfecting a "neutral vocabulary-a public-opinion glossary-within the comprehension of the mass of people" (p. 106). More importantly, he insisted from very early in his career that scientific experiments had shown that variations in question-wordingwere of "relatively minor importance." Touting his own "splitballot" technique for testing question-wording effects, Gallup noted that his firm often employed two different versions of a question to assess what, "if any," difference question-wording might make. In more than 200 such experiments, Gallup (1940b) reported that "greater-than-expected" differences had been discovered in "only a small fraction" of cases. Where there was "no material alteration in the thought expressed," Gallup concluded, "there will be no material difference in the result, no matter what wording is used" (p. 26).

Despite his protestations, Gallup's own "spit-ballot" data often revealed questionwording effects that seemed quite material indeed. In one of his own experiments in 1939, for example, half of the sample was asked, "Do you thinkthe United States will go into the war before it is over?" The other half was asked, "Do you think the United States will succeed in staying out of the war?"Of those asked the first question, 41 percent said "yes" and 33 percent said "no." Of those asked the second, 44 percent said "yes" and 30 percent said "no." In short, a plurality of respondents answered "yes" to both questions. This represented an eleven-point difference in the percentage of people who felt that the U.S. would go to war. In all likelihood, the explanation lies in what modern researchers call "response acquiescence," or the inexplicable tendency of people to say "yes" when asked about unfamiliar issues. Gallup, however, ingeniously attributed the finding to the instability of opinion on the issue (Moore, 1992,  p. 325-326).

Other researchers employing Gallup's split-ballot technique likewise discovered significant but mysterious effects from seemingly nonsubstantive variations in question-wording. For example, Rugg (1941) analyzed split-ballot data gathered by Elmo Roper. In a survey on attitudes toward free speech, Roper had asked half his sample whether "the United States should forbid speeches against democracy" and the other half whether "the United States should allow" such speeches. The difference proved startling: 46 percent opposed "forbidding" such speeches, but only 25 percent favored "allowing" them.

In the late 1940s, Gallup (1947) announced a major breakthrough in the "science" of question-wording-a new "system of question-design" that he insisted answered all of the most "frequently voiced criticisms of public opinion measurement."Impressively jargonized as the "quintamensional approach," this new "system" allegedly allowed the pollster to "exclude from poll results the answers of any and all persons who are uninformed on any given issue" and to probe in greater depth opinions on "almost any type of issue and at all stages of opinion formation." Employing five categories of questions, including "filter" questions to assess knowledge and follow-up questions to probe the motivations and reasoning behind opinions, the "quintamensional approach," as Gallup described it, allowed a number of different dimensions of public opinion to be "intercorrelated, with a consequent wealth of data by which public opinion on any issue of the day can be described" (pp. 385-393).

Throughout his career, Gallup (1972) touted the "quintamensional approach" as among the most "important developments" in the history of polling, a development that made possible the assessment of public opinion even on the most technical or complex of issues (pp. 90-91). In reality, however, it only created new problems in question-wording. The use of "filter questions," for example, created unique and artificial interpretive contexts for questions that came later in the interview process. The whole approach also encouraged the most empirically and philosophically indefensible practice in modern polling: the practice of "informing" respondents about an issue before asking their opinion. When Gallup (1947) first unveiled the "system," he envisioned this very practice as among its most exciting possibilities, as he suggested that pollsters might use "simple and neutral language" to "explain" issues to those identified by the "filter questions" as "uninformed." This would allow pollsters to compare informed and uninformed opinion and, more generally, "to obtain the views of the maximum number of voters" (p. 389). Only a handful of critics have questioned the possibility of providing genuinely "neutral information" (see, e.g., Robinson & Meadow, 1982, pp. 118-120). So too have researchers only recently begun to ask whether pollsters ought to be in the business of predicting what the public might think if exposed to selected "information" (see, e.g., Hogan & Smith, 1991, p. 547).

At best, the "quintamensional approach" was a diversion and, at worst, a scientific hoax. A classic case of scientific mystification, it did nothing to solve the real problems in question-wording. It provided virtually no guidance for how to word questions; it specified only the types of questions to be asked and their order. It did nothing to answer the most fundamental question about the wording of questions: how does one account for the often enormous impact of seemingly nonsubstantive variations in wording?

Meanwhile, additional studies revealed significant question-wording effects, not only from variations in wording, but also from the number, character, and even order of response options. In 1951, Stanley Payne summarized a number of these studies in The Art of Asking Questions, a book whose very title challenged Gallup's efforts to scientize question-wording. Payne documented a wide variety of questionwording effects that ranged from the false portraits created by obviously "loaded" questions (pp. 177-202) to a more subtle and mysterious "recency" effect-that is, a tendency among respondents to choose the last response option in some but not all questions (pp. 129-134). Payne concluded that "improvements in question wording" could "contribute far more to accuracy than further improvements in sampling methods"-"tens" rather than "tenths" of percents (p. 4).

Over the next twenty years, however, research on question-wording all but disappeared from the polling literature-as if Payne's book, with its "concise list of 100 considerations" in asking questions, "represented the final, and comprehensive, word on the subject" (Moore, 1992, p. 328). Gallup played a major role in this neglect of question-wording by continuing to insist well into the 1970s that "thorough pre-testing" assured the reliability of all questions and that, in any case, "split-ballot" experiments had proven that "even a wide variation in question wordings" did not produce "substantially different results if the basic meaning or substance of the question remained the same" (Gallup, 1972, pp. 78, 83).

Question-wording did not become a serious and sustained focus of research until a new "intellectual movement" emerged among academic survey researchers in the 1980s-a movement that broke the conspiracy of silence about question-wording and finally made research on "response effects" a "mainstream pursuit" (Moore,1992, p. 343). This "movement," led by University of Michigan researcher Howard Schuman and others, began with efforts to summarize and replicate earlier research. The effort to theorize about question-wording was frustrated, however, as explanations for a number of question-wording effects proved elusive. Intrigued by Rugg's 1941 study of attitudes toward free speech, for example, Shuman and Presser (1981) repeated the experiment and found the same effect. Despite the passage of forty years, respondents still were far more likely to agree that the U.S. should "not allow" anti-democratic speeches than to agree that the U.S. should "forbid" such speeches. When they performed the same experiment on the question of "forbidding" or "not allowing" pornographic movies, however, the effect disappeared. In this case, there was no statistically significant difference between those who would "forbid" and those who would "not allow" such movies (pp. 276-283).

In addition to variations of ten, twenty, or more percentage points from seemingly nonsubstantive variations in question wording,[vii] more recent studies also have shown significant response effects even when questions are not reworded at all, but merely reordered. In studying polls on abortion, for example, Schuman and Presser (1981) attempted to account for a 15 percentage point difference in two surveys employing exactly the same question: "Do you think it should be possible for a pregnant woman to obtain a legal abortion if she is married and does not want any more children?" The researchers noted that the survey revealing less support for abortion rights placed this question after another question asking whether it should be possible to get a legal abortion "if there is a strong chance of a serious defect in the baby." Testing the possibility that this question created a different interpretive context for the second question, the researchers conducted a split-ballot experiment in which half the sample was asked both questions, with the "defective fetus" question first, and the other half both questions, but with the order reversed. The difference was substantial, with the "defective fetus" question, when asked first, reducing support for abortion rights in the other question by about 13 percent-the difference between majority support and majority opposition to abortion rights (pp. 36-39).

Expanding into the area of response options, Schuman and Presser (1981) not only confirmed the "recency" effect observed by Payne (1951), but also demonstrated a very substantial effect from offering or not offering "don't know" as a response option. In Schuman and Presser's experiments, explicitly offering the "don't know" option increased the number of people in that category by an average of 22 percent (pp. 113-146). Studies by other researchers revealed that, in the absence of the "don't know" option, many respondents would express opinions even on entirely fictitious policies or programs (see, e.g., Hawkins & Coney, 1981, p. 372).

These and similar findings would seem to confirm what polling pioneer Hadley Cantril observed many years ago in the foreword to Payne's (1951) book: the variables involved in question-wording are "legion"; "they are also subtle and they defy quantification" (p. viii). More recently, Bradey and Orren (1992) went further in concluding that "the more one explores the many sources of . . . error [in questionwording], the more hopeless the prospect seems of ever preparing a poll questionnaire that is not fatally flawed" (p. 77). Reflecting on the limitations of both openand closed-ended questions, Schuman and Scott (1987) have suggested, quite simply, that no survey question, nor even any set of survey questions, should be taken as a meaningful indicator of either absolute levels or relative orderings of public preferences. Not surprisingly, this sort of conclusion has caused considerable consternation among pollsters and survey researchers. Responding to Schuman and Scott, Eleanor Singer (1988), then President of the American Association for Public Opinion Research, complained that such a view threatened to undermine the "intellectual defense" of survey research-that it was only "a short way" from Schuman and Scott's conclusion to the rationale for survey research and polling "being blown out of the water altogether" (p. 578).

The pollsters nevertheless continue to say little publicly about problems of question-wording. They now insist more strongly than ever that a poll's "margin of error" be reported, but this insistence only sustains the illusion of science and wrongly implies "that sampling is the primary, if not the only, source of possible error in a poll" (Cantril, 1991, p. 119). Even pollster Burns Roper has conceded that reporting the "margin of error" is little more than a diversionary tactic. As Roper (1980) observed, pollsters typically present their findings in the media as "within three percentage points" of what the public actually thinks on some issue, "when, in fact, a differently worded question might-and often does-produce a result that is 25 or 30 points from the reported result" (p. 16).

The Illusion of Science

Gallup's great pet peeve was the tendency of some writers to place quotation marks around the word "scientific" as applied to polling. "If our work is not scientific," Gallup (1957-58) complained to the sympathetic readers of Public Opinion Quarterly, "then no one in the field of social science, and few of those in the natural sciences, have a right to use the word." According to Gallup, polling fully qualified as "scientific" even under "the most rigid interpretation" of that word (p. 26). Those who thought otherwise simply did not understand "the nature of the new science of public opinion measurement" (Gallup, 1940b, p. 23). When Professor Lindsay Rogers of Columbia University published a scathing critique of the industry in 1949, for example, Gallup (1949b) characterized him as "the last of the arm-chair philosophers in this field" and declared: "One cannot blame the Professor for his lack of knowledge about polling techniques. He says, quite honestly, that he knows nothing about research procedures-a fact which is abundantly clear in his chapters dealing with these matters" (pp. 179-180). When Congress considered regulating polls in the wake of the 1948 debacle, Gallup (1948c) similarly dismissed critics who raised questions about question-wording as ignorant of "the new procedures that are followed and of the care that is taken to dissect opinion not only with one question but, in the case of complex issues, with a whole battery of questions" (p. 735).

Gallup's claim to "science" also buffered the pollsters against criticism of their interpretive practices. In responding to such criticism, Gallup did not merely claim to interpret his data objectively; indeed, he claimed that, as a scientist, he did not interpret at all! Characterizing polling firms as "fact-finding institutions" whose "responsibility ends with the objective reporting of survey results," Gallup (1960) attributed all complaints about the polls to partisanship or sour grapes. "Obviously," he argued, the pollsters could not be "charged for the manner in which others may interpret or use poll findings," nor could they be "properly attacked for results which happen to be unpalatable to some people." Such attacks, Gallup concluded, were akin to blaming the meterologist for bad weather: "Just as barometers do not make the weather, but merely record it, so do modern polls reflect, but not make public opinion" (pp. ix-x).

The pollsters continue to maintain a scientific facade by placing heavy emphasis on sampling, data collection, and data processing in their public discussions of methodology. In its monthly journal,[viii] for example, the Gallup Poll historically has said little about question-wording and nothing about interpreting data. In the 1980s and 1990s, the Gallup Poll Monthly added a statement about question-wording and other sorts of "nonsampling error" to its regular "Note to Our Readers": "In addition to sampling error, readers should bear in mind that question wording and practical difficulties encountered in conducting surveys can introduce error or bias into the findings of opinion polls." At the same time, however, the Monthly added two new pages on weighting procedures, sampling tolerances, and the "random digit stratified probability design" used in telephone surveys. The Monthly even added tables for calculating the error for subsamples of various sizes (see Gallup Poll Monthly, 1992).

Gallup's rhetoric of science thus lives on, not only within the firm he founded, but in the entire polling industry's claims of "scientific progress." Pointing to continued improvements in sampling techniques and a variety of other "new technologies available through computers," pollsters began boasting in the 1970s that their "science" had "truly come of age" (Roll & Cantril, 1980, p. 13). In this respect, pollsters may differ little from economists, psychologists, sociologists, or anthropologists, whose claims to "science" have been demystified in recent years by students of the "rhetoric of inquiry" (see Nelson et al., 1987). Yet pollsters take such claims one step further by boasting that their science can "greatly strengthen the democratic process by providing `the people' yet another way of making their views known particularly between elections" (Roll & Cantril, 1980, p. 14).

Preaching the Democratic Faith

In classical democratic theory, public opinion did not equate with some numerical majority, however measured. Only ideas tested in public discourse were considered worthy of the label "public opinion" (MacKuen, 1984, pp. 236-238). Gallup (1948a) subscribed to a different view. He embraced James Bryce's definition of "public opinion" as the "aggregate of the views men hold regarding matters that affect or interest the community" (p. 84). Substituting undifferentiated mass opinion for "tested" ideas and thoughts expressed privately and confidentially for genuinely "public" opinions (see Herbst, 1993, pp. 64-65), Gallup, in effect, redefined public opinion as that which the polls measured.

This redefinition, in turn, raised anew the most fundamental philosophical and empirical questions of democratic theory. Should public opinion, however defined or measured, dictate public policy? Is the public, especially as more broadly defined by polls, sufficiently interested and informed to govern itself? Not surprisingly, George Gallup had answers to these questions. Less surprisingly still, those answers rested largely on the results of his own polling. Enthusiastically preaching the democratic faith, Gallup did not merely celebrate America's democratic heritage and affirm the people's right to govern themselves; he flattered the American people by invoking his own polls as proof of their seemingly mystical "collective wisdom."

Gallup did not dwell on the philosophical question of whether public opinion should dictate public policy. His answer seemed obvious from how he phrased the question itself: "shall the common people be free to express their basic needs and purposes, or shall they be dominated by a small ruling clique?" (Gallup & Rae, 1940, p. 6). In Gallup's polarized world, there were only "two points of view" on this question, and between them there could be "no compromise" (Gallup & Rae, 1940, p. 8). One either sided with those "who would place more power in the hands of the people" or those "who are fearful of the people and would limit that power sharply" (Gallup, 1940b, p. 57). In no uncertain terms, Gallup stood with the democrats: "In a democratic society the views of the majority must be regarded as the ultimate tribunal for social and political issues" (Gallup & Rae, 1940, p. 15).

The empirical question remained: were the "common people" capable of governing themselves? Only with the advent of polling, according to Gallup (1940b), had it become possible to answer this question-to ascertain whether the people were "responsible or irresponsible; blindly radical, blindly conservative or somewhere in between; and whether their judgments may or may not make sense when viewed in perspective" (p. 57). Happily, according to Gallup (1948b), "the polls" had shown again and again "a quality of good common sense" in the American people that was "heartening to anyone who believes in the democratic process" (p. 177). In a refrain that echoed throughout his career, Gallup insisted that the people not only proved to be "right" about most issues but typically were "far ahead" of their leaders.

Early in his career, Gallup pointed to attitudes concerning preparedness and World War II as proof of the public's wisdom and foresight. Not only did the public understand far better than their leaders the threat posed by Hitler, but they were willing, even eager, to sacrifice personally to counter that threat. "Instead of balking at increased taxes," Gallup (1940a) wrote, "the American people have been ready to make direct, personal sacrifices in order to build up an army, a navy and an air force-and all this long before Congressmen and other political leaders were ready to recognize the fact" (p. 21). In a radio address, he even described the public as "cheerfully" accepting "the burden of heavy wartime taxation" (Gallup, 1942a, p. 688). Elsewhere, Gallup (1942c) detailed how the American people would be willing to sacrifice 10 percent of each paycheck for defense bonds, to accept "complete government control" over prices and wages, and even to allow the government to dictate "what kind of work" they should do and for "how many hours" (p.16). Thus, the polls not only confirmed "the sound judgment and foresight of the common people" on "virtually all . . . important war-time issues," but also their willingness to sacrifice personally "to help the country win the war." Indeed, the people were so far "ahead of their political leaders," Gallup (1942b) concluded, "as to raise the real question as to whether the leaders are leading the people or whether the people are leading the leaders" (pp. 440-442).

In the 1950s, Gallup changed his tune somewhat as he lamented the public's failure to appreciate fully the threat posed by communism. "Today for the first time," he wrote in the New York Times Magazine in 1951, "I must confess that I am concerned lest lack of information lead the American people to decisions which they will regret" (p. 12). Significantly, however, Gallup blamed a "lack of information," not the people themselves, and he took it upon himself to "inform" both the public and Congress of the need for more vigorous American propaganda to counter the Soviet threat (see U.S. Senate, 1953, pp. 772-793). Gallup thus simultaneously worried about public apathy and sounded his usual refrain that the public was "far ahead" of its leaders. In a speech at Georgetown University, for example, Gallup (1952b) worried that the American people might "delude themselves" into complacency, yet again the public was "years ahead of Congress" in its desire "to go all-out to sell our point of view to the world," even if it meant "the expenditure of a billion dollars and more to do the job" (p. 502).

A decade later, Gallup (1968b) focused on electoral reform and declared that never before in history had "the man in the street been so conscious of all the shortcomings and faults in the whole electoral process from beginning to end" (p. 41). Noting that presidential campaigns had "degenerated" into a "phony business," a "matter of showmanship" and "a test of the candidate's stamina" rather than of "his intelligence" (Gallup, 1968a, p. 133), he again claimed that the polls showed the public ahead of its leaders. For "many, many years," Gallup (1968c) told an interviewer, the people had "favored doing away with the Electoral College" and establishing "a nationwide primary." They also favored "shorter campaigns" and "more serious and shorter conventions." They would "put an absolute top on campaign spending by any candidate or any party," and they would "give each candidate five or six half-hour broadcast periods to present his case-at which time everything else would be ruled off the air." In short, the public supported "a whole series of reforms that would change virtually every step and stage of the business of getting elected." Again, the public not only was "right" about electoral reform but "far, far ahead of the parties and the politicians" (p. 34).

In the twilight of his career, Gallup continued his crusade for electoral reform and also became deeply concerned about public education. Inaugurating an annual Gallup Poll on public attitudes toward education in 1968 (see Elam, 1984), Gallup became a major voice for educational reform. Meanwhile, he continued to chart public opinion on a wide variety of foreign and domestic issues and to celebrate the public's wisdom and foresight on even the most complex and divisive issues of the late 1960s and 1970s. On Vietnam, for example, Gallup (1972) again declared the public to be "far ahead of its leaders" for supporting Vietnamization three years before Nixon even proposed it (p. 17). Even in the post-Watergate climate of political alienation and disenchantment, Gallup's faith in the wisdom of "the people" remained undiminished. At the age of 77,  allup (1978) sounded the refrain that he first introduced in the 1930s in summarizing what he had learned from "conducting thousands of surveys on almos  every conceivable issue for nearly half a century": "One is that the judgment of the American people is extraordinarily sound. Another is that the public is almost always ahead of its leaders" (p. 59).

In preaching the democratic faith, Gallup obviously went beyond merely "reporting" his data. Recalling "very few instances . . . when the public was clearly in the wrong on an issue" (as cited in McDonald, 1962, p. 28), Gallup judged the wisdom both of policies and of the public's opinions. Gallup also routinely reflected on the historical significance of findings, discerned causes and trends, and even predicted the future. Most significantly, however, Gallup invoked the moral authority of vox populi in a number of his own pet causes. Notwithstanding his professions of scientific objectivity and political independence, it seems more than coincidence that Gallup-the conservative businessman, the close personal friend of Thomas Dewey, the staunch Cold Warrior-almost invariably discovered a public with the "right" political views.  In short, George Gallup was no mere "weatherman" of public opinion. To the contrary, he sought to alter the ideological climate of his era with the results of his own polls.

The Faint "Pulse of Democracy"

Academic survey researchers now generally agree that if polls are to be anything more than "entertainment . . . or propaganda for political interests," they must be integrated with "additional sources of information into a fuller, textured, portrait of public opinion" (MacKuen, 1984, p. 245). In other words, individual polls, if they are to be meaningful at all, must be interpreted both in terms of larger historical or social trends, and within the context of public debate and discussion. At a minimum, polls must be compared to related polls if one hopes to speak of trends, shed light upon the impact of question-wording, or explore the salience and logic of the public's opinions. Indeed, recent studies suggest that meaningful conclusions about public opinion can emerge only through comprehensive, comparative analyses of all available opinion data on particular topics or issues (see, e.g., Hogan & Smith, 1991).

George Gallup's legacy discourages such comprehensive and integrative analysis. Establishing polling as a commercial enterprise and aligning it with journalism, Gallup pioneered the institutional arrangements that today not only discourage comparative and in-depth analyses of polling data but have corrupted the mission of polling. Some polls, of course, are commissioned by politicians or special interest groups, which creates an obvious incentive for bias. Less obvious, yet even more troubling, however, is how journalistic imperatives have corrupted even the syndicated media polls.

Early in his career, Gallup offered media syndication as "a pretty realistic guarantee" of polling's honesty and impartiality. As Gallup (1938) summarized that guarantee, the Gallup Poll derived "all its income" from some sixty newspapers "of all shades of opinion"-some "left of center," some "right of center," and some in "the middle of the road." With that arrangement, Gallup asked rhetorically, "how long do you suppose we would last if we were anything but honest?" (p. 138).

"Honesty" is not the issue when considering how journalism has affected the quality and character of public opinion polling. Rather, as Everett Carll Ladd (1980) has argued, there is a "clash of institutional imperatives" between  ournalism and polling, with contemporary notions of "news" encouraging technically deficient polling. So serious is this "clash," and so different are the requirements of a good news story and sound polling, that the "linkage of polling and the press raises serious questions as to whether opinion research does, or even can, enhance the democratic process" (p. 575). Indeed, there is no longer much question about it: as essentially wholly owned subsidiaries of a handful of media conglomerates, polling firms do not, in fact, aspire to "scientific" assessment of public opinion. Instead, they strive to make news.

The highest priority among journalists using polls is "timeliness," yet this creates a built-in incentive for methodological shortcuts: smaller samples, fewer efforts to track down respondents, truncated and unverified interviews, and less data processing. Once the data are collected, moreover, time and space limitations, along with the dictates of a good news story, encourage interpretations that oversimplify and overdramatize the results. As Ladd (1980) pointed out, good news reporting "has focus and arrives at relatively clear and unambiguous conclusions," whereas "good opinion research typically reveals . . . tentativeness, ambivalence, uncertainty, and lack of information or awareness" (p. 577). As Moore (1992) further reminds us, news organizations pay dearly for their "exclusive" polls and thus have a "vested interest" in making the results "sound important" (p. 250). As a result, both journalists and the pollsters themselves look for something more than disinterest or ambivalence in polling data: something new or surprising, something dramatic, a "lesson" to be learned, or some portent of conflict.

Journalistic imperatives have, in a sense, relieved pollsters of the problem of framing good questions. Failing to discern news in the public's relatively stable opinions on persistent political issues, journalists now demand more quirky polls from the pollsters-polls that pursue fresh, often non-political, and occasionally even silly angles on "old news." During the Persian Gulf War, for example, the pollsters typically asked the public not to assess the Bush administration's war policy, but to speculate about the unknowable: "How long do you think the war will last?" And "how many Americans do you think will be killed before the war is over?" The Gallup poll asked Americans not only to speculate about whether America would "win" the war, but even whether they thought "prayers" might be "effective" in resolving "a situation like this one in the Persian Gulf" (Gallup Poll Monthly, 1991, pp. 7, 25). As Mueller (1994) has observed, the Persian Gulf War was the "most extensively polled" issue in American history-the "mother of all polling events" (p. xiv). Yet despite asking some 5,000 questions (Ladd & Benson, 1992, p. 29), the "picture of public opinion" that emerged from polling on the Gulf War remained "fuzzy," as Moore (1992) has written, and "the crucial question of whether the public supported or opposed a war with Iraq could not be summarized with a percentage point estimate" (p. 353).

Polls have become "news events" in and of themselves. As a result, they substitute for substantive information about political issues and stifle debate. Indeed, as Herbst (1993) has observed, polls often make political debate seem "superfluous," since "they give the illusion that the public has already spoken in a definitive manner" (p. 166). When debates do occur, they often focus on the polls themselves, with partisans debating not the merits of a policy but the "message" of "the polls" (see, e.g., Hogan, 1994, pp. 119-137). Such an emphasis, in turn, discourages bold leadership, since it pressures elected representatives to defer to the polls. It may even discourage the public itself from other modes of political expression (such as letter-writing or demonstrations), since it suggests not only that the public already has spoken but that the public "speaks" only through the polls.

During election campaigns, polls likewise substitute for substantive information about the records or proposals of candidates. During the 1988 presidential primaries, for example, nearly a third of all election stories on the network news noted poll results, and during the final months of the campaign, a new poll was reported, on average, every other day (Jamieson, 1992, p. 175). More importantly, the polls have redefined campaign "news" itself and have fueled an astonishing increase in so-called "horserace journalism"-an emphasis on the strategic "game" of politicssince the 1960s (see Patterson, 1994, pp. 53-93). Not only does such reporting deprive voters of the information they need to assess candidates, but it may discourage political participation altogether by casting voters as mere "spectators" (see Jamieson, 1992, pp. 186-188).

Despite all the historical "lessons" and "scientific" breakthroughs, we thus remain a long way from George Gallup's vision of a "scientific democracy." Indeed, the public opinion poll not only has failed to become "the most useful instrument of democracy ever devised," as Gallup predicted (Meier & Saunders, 1949, p. 218), but it has had a number of deleterious effects on public discussion and political leadership in America. Properly designed and interpreted, polls might well serve George Gallup's dream of a more inclusive and efficient democracy. In a sad irony of history, however, Gallup's own rhetorical legacy subverted that dream, and the "pulse of democracy" only beats fainter in the age of saturation polling.

Footnonets:

www.pseudology.org