PART ONE
THE REPRESENTATION OF PUBLIC OPINION
CHAPTER 1
WHY NUMBERS MATTER AND WHY WE SHOULD BE SUSPICIOUS OF THEM
In Norton Justerâs The Phantom Tollbooth, there are two competing kingdoms: Dictionopolis, the kingdom of words, and Digitopolis, the kingdom of numbers. The two kingdoms are divided by a philosophical squabble. Words, according to King Azaz of Dictionopolis, are indisputably more important than numbers. Numbers, argues the Mathemagician of Digitopolis, are measurably more significant than words. The dispute is, of course, irresolvable: a fact that only feeds the intransigence of the two positions.
If this argument has not exactly been replicated in the weightier tomes of academic writing, there are moments when it is possible to catch a whiff of a similar kind of territorial obduracy. Opinion pollsâand quantitative surveys in generalâare mechanisms for turning words into numbers. For some, this mechanism is one of the fundamental tools of social science, the apotheosis of an enlightenment project in which hypotheses about social structures can be measured, evaluated, proved, or invalidated. It is a mechanism that allows, as Melvin DeFleur puts it, âthe accepted epistemology for researchâ to be âdefined by the rules of natural scienceâ (DeFleur 1998, 92). For the statistically well-equipped, Herbert Kritzer points out, âdata are funâ (Kritzer 1996, 26). Indeed, Kritzer ingeniously borrows Roland Barthesâs exposition of free-floating literary criticism (Barthes 1975) to write about âthe pleasures of the statistical textââone that is a creature of context and interpretation, both confirming and revealing.
For others, the process of reducing words to numbers is like sucking the life out of language, a remorseless exercise in Gradgrindery whose limited world-view masquerades as objectivity and universal truth. The pollster is like Justerâs Mathemagician, transforming the complexities of social life into numbers on a page, into a form where they can be manipulated. As Andrew Ruddock puts it, those who reject a quantitative approach see surveys as âblunt objects bludgeoning heterogeneity out of audiences and rendering them amenable to the control of politicians and marketing executivesâ (Ruddock 1998, 116).
There are many fault lines that divide academics in the social sciences and the humanities, but the use of quantitative surveys as a way of probing what, how, and why people think is perhaps one of the most pervasive. Those who embrace the technology of surveys find themselves exasperated by social scientists who make claims without recourse to systematic forms of evidence. For social scientists like DeFleur, the quantitative survey is the only apparatus that allows us to progress. âThe milestones of the future,â he writes, âwill be products of carefully conducted scientific research, as opposed to qualitative writings or ideological criticismâ (DeFleur 1998, 94). Others reject these forms of quantification, arguing that you do not understand something as complex as language by turning into something neat and quantifiable, or by turning complex human beings into what Ien Ang calls âtaxonomic collectivesâ (Ang 1991). In the corridors of academe, the two camps can be singularly less polite, adopting monikers for one another that display anything from mockery to withering contempt.
The opinion survey, in other words, is a contested object, one that is both revered and reviled. It is either the quintessential scientific tool of a democratic age or the discarded machinery of a bygone intellectual era. This is partly because, as Susan Herbst points out, while the act of quantification is regarded by some as a scientific process, it is also a symbolic one (Herbst 1993a). Changing words into numbers is, in short, a semiotic act. It takes soft, rambling, shambling views of the world and turns them into hard figures and percentages. In so doing, what was once meandering and imprecise is pushed into a world governed by a series of inarguable mathematical truths. Humanity becomes science. And in a culture where science is often elevated above the petty squabbles of history and ideology, this gives the pollster a degree of power and authority. The argument about the use of quantitative surveys is, in this sense, an argument about the useâor abuseâof power.
If the use of quantitative surveys is a symbolic bid for authority, the public opinion poll is especially fraught with notions of power. Its subject, after all, is the theoretical basis for political legitimacy in a modern democracy. In a system in which politicians are supposed to be the publicâs representatives, the power to define and interpret public opinion is paramount. The debate about public opinion polls is therefore an issue of political legitimacy, a matter of conferring upon representatives the right to govern in the publicâs name. The long-term stakes could scarcely be higher.
For many on the nonquantitative side of the academic divide, the opinion poll is derided as the epitome of empiricist social science, a clumsy old technology that cranks and creaks its way toward untenable, simple-minded conclusions. Dismissals are usually made in the short-hand language of social theory: quantitative survey work is described as being mired in functionalist, empiricist, and positivist views of the world. In plainer language, there are two broad thrusts to the criticism of polling and quantitative surveys: the failure of pollsters to see the âbig pictureâ and the crudity of the methodâcriticisms I shall elaborate on shortly.
The purpose of this chapter is not to come down on one side or the other or to offer some form of compromise. My argumentâand the premise of this bookâis that the process of enumeration is an important (and in many instances, inevitable) part of any study of culture and society, but that enumerative formsâsuch as the opinion surveyâhave specific limits and operate in ideological contexts that give meaning to those limits. To interpret poll responses therefore requires an understanding of the constrained, ideological conditions in which they are produced.
THE VALUE OF NUMBERS
Although few academics would admit it, the nature of the divide between those who rely on numbers and those who donât often becomes territorial, where membership on one side or the other is often less a matter of rigorous analysis than a product of simpler preferences. As Kritzer writes: âThe debates that surround methodological choices are important, but they often are specious. Most analysts make their broad methodological choices based on what they like doingâ (Kritzer 1996, 25). Regardless of the origin of oneâs position, both sides have often focused more on the egregious aspects of the other rather than on their more careful or incisive moments. If polling has its limits, so too do some of its criticsâmany of whom feel that the opinion poll literature is so full of erroneous assumptions that it is of little interest.
My own background, for example, is influenced by cultural studies, which, like other fields of poststructuralist inquiry, has always maintained a strong suspicion of quantitative research. When empirical work has been done, less intrusive, more qualitative methodsâsuch as the focus group interview or more ethnographic forms of inquiryâare invariably preferred (e.g., Morley 1980, 1986, 1992; Radway 1984; Ang 1985; Hodge and Tripp 1986;Corner, Richardson, and Fenton 1990; Lewis 1991;Press 1991; Jhally and Lewis 1992; Heide 1995; Tulloch and Jenkins 1995; McKinley 1997). Qualitative research imposes frameworks and categories upon its subjects, but it waits longer before doing so, allowing people more time to speak in their own way. Rather than turning words into numbers, words are turned into other words. In this qualitative milieu, quantitative surveys of thought and opinion have been out of fashion for so long that they have, for many, been consigned to a methodological junkyard. The technology lies abandoned, out of sight and out of mind.
For those who are used to dealing in words rather than numbers, the very notion of reducing one to the other is to lose the infinite complexity of language. If qualitative methods are, like polls, a kind of intervention into the social world rather than an innocent means of reflection, they nevertheless retain something of that worldâs discursive form.
While a preference for qualitative form is often well conceived and fruitful, the lingering suspicion of numerical data has, I think, degenerated into habit. It is as if the argument with these methodologies was so comprehensively settled that one can be spared the time and effort of any further thought on the subject. But more than this, survey-generated statistics are neglected partly because of how they have traditionally been used, partly because of what they have come to signify. Favored by empiricists and positivistsâpeople who believe an objective world will simply reveal itself in columns and percentagesâthe honest toil of number-crunching is interpreted as a signifier of empiricism itself, dismissed with little more than the patronizing wave of a hand.
The symbolic power of numbers to connote science and scientific rigor is inverted on this critical terrain: numbers are seen to symbolize a narrow, controlling view of the world, an arrogant, anal-retentive, and characteristically male approach to social science. Thus the term ânumber cruncher,â with its connotations of empty-headed manual labor, becomes a pejorative term. Since many of the characterizations in the disputes between practitioners of quantitative and qualitative methods are symbolic, it is worth noting the class-bound snobbery in this dismissalâone that connotes a stand by an educated class against philistine incursion.
More importantly, we have to acknowledge that the moment we begin the process of sorting and categorizing qualitative data, we are treading, ever so lightly, into the world of numbers. This is necessarily so. As Andrew Ruddock writes, âThe very notion of culture depends on the relative coherence of meaning within a given society,â and that while understanding difference is integral to any sophisticated analysis of media and audiences, âso too is the recognition that the possibilities for difference are not boundlessâ (Ruddock 1998, 122) or chaotic.
Qualitative audience research, for example, is full of inferences about the wider applicability of particular cases. It suggests how certain people understand TV programs, films, or other cultural formsânot always or absolutely or to the nearest percentage point, but in the general run of things. And if a critical reading of a text is written as cultural (rather than purely literary) criticism, it implies a quantitative presence of that reading in contemporary culture. We may not be able to enumerate it, but in describing its presence we assume that it is, in some measure, significant or quantifiable. We assume that it counts.
It is difficult to discuss political power without, at some point, implicating majorities within civil society. We might prefer to gloss over the nitty gritty, arithmetic aspects of these implications, but they still lurk, operational but unspoken, beneath the tentative conclusions of qualitative research. Underlying discussions of resistance, dominance, or the significance of media are enumerative questions of space and place. As Stromer-Galley and Schiappa argue, a great deal of the work that has gone on within the more qualitative traditions of textual or rhetorical criticism has made implicit claims about the way media texts are understoodâwhat they call âaudience conjectures.â The fact that âonly a handful provided evidence for such claimsâ (Stromer-Galley and Schiappa 1998, 34) may grant them a degree of poetic license, but it does not free them from evidentiary burdens, which, in turn, forces them to consider quantities as well as qualities.
If quantitative surveys (particularly those carried out by the television and marketing industries) tend to reduce citizens or television viewers to crude typologies, or else forget that the production of social science is itself a discursive enterprise, it is partly a function of design, partly of interpretation. The general technology of data production does not automatically prescribe or determine the meaning of the data. And although the act of counting is not theoretically innocentâwe must categorize before we can countâit implies a process integral to social science. Most forms of social science are dependent upon categories and typicalities. Educational levels, race, income, sexuality, or gender may be constructions, but it is difficult to talk about society or history without them (Christians and Carey 1989).
Similarly, to the argument that âpublic opinionâ itself is partly a construction of the person who defines and describes it, one might respond with: so what? All forms of social categorizationâsuch as race or classâcan be seen as constructions. This does not mean that those constructions do not refer to real practices and real objectsâor even that those constructions are not meaningful in peopleâs lives. What is important is how those constructions are used and how they relate to people and practices.
It is also apparent that many of those in qualitative fields are unaware of the volume of critical work by those closer to quantitative traditions. Contributions to the critical history of polling have come from various disciplines and perspectives; from political science, mass communications, sociology, and cultural studiesâin short, from those who use quantitative methods to those who regard machines for turning words into numbers as inherently flawed. The overview that follows reflects this range of critical thinking. It is not, however, offered as the precursor to a rejection of the use of surveys but rather as a means for establishing the specific and limited ways in which those surveys might be useful.
THE LIMITS OF NUMBERS:
A CRITICAL HISTORY OF POLLING
Early opinion researchâin the 1930s and 1940sâtook place in an age of mass production and mass entertainment: everything, it seemedâfrom the sensibilities of art to the practicalities of scienceâhad become almost infinitely reproducible. New philosophical and sociological conceptions were required to make sense of an emerging era of mechanized uniformity at a time when fascism and Stalinism raised troubling questions about the politics of propaganda and mass culture. These doubts were manifested in films like Modern Times and in the work of the Frankfurt School (e.g., Adorno and Horkheimer 1979; Benjamin 1985), Walter Lippmann (Lippmann 1922, 1925), and others.
The opinion survey was an appropriate tool with which to begin to make sense of the emergence of a mass society informed and entertained by mass media. In the early research on mass culture, public opinion was therefore conceived as part of a social process, a way to chart the meaning of an emerging mass culture and the power of mass media. James Beniger describes how the first issue of the journal Public Opinion Quarterly (published in January 1937) promoted the idea of using polling technology to examine the relationship between mass communication and mass opinion within the more general conception of a mass society (Beniger 1987).
The problems with the âmass cultureâ and early media effects traditions have been elaborated elsewhere (see, for example, Morley 1980; Radway 1986). Suffice it to say that early research in this field was methodologically limited (Lewis 1991), partly because it had a narrow conception of how media might be influential. Research in this tradition focused, for example, on overtly propagandist forms and on easily observable changes in behavior rather than more subtle or long-term shifts in views of the world; as a consequence, its findings were limited and inconclusive. According to Beniger, the comparative failure of the âmass societyâ and âeffectsâ models to advance an understanding of the relationship between mass society and mass opinion meant that âsociologists and political scientists lost interestâ and hence âthe study of mass communication separated from that of public opinionâ (Beniger 1987, 50).
Subsequently, the development of the representative survey method allowed a slippage to occur, so that public opinion came to be understood as no more than the aggregation of responses to survey questions. Once this slippage became institutionalized, public opinion tended to be portrayed as an empirical âfactâ to be measured by polling rather than as a manifestation of social processes, like mass media consumption.
In this context, we might say that for all the flaws of the early mass communication approaches, they nonetheless involved a conception in which public opinion was seenâhowever crudelyâas part of a social process. Public opinion was conceived, at least potentially, as a construction that may or may not be a product of ideological apparatuses (such as the media).
Once freed from the specters of âmass communicationâ and âmass society,â public opinion research became something of an end in itself, a subdiscipline of political science. The dominant notion of public opinion polls that replaced the mass society conception was pseudoscientific and straightforwardly empiricist: public opinion simply existed as an independent entity to be gathered by neutral and objective experts. The socially constructed nature of public opinion surveys was implicitly denied by this empiricist model. Public opinion was simply out there, a set of facts to be collected and analyzed.
The move toward the study of public opinion in isolationâas a fact of political lifeâis in many ways an understandable one. If we see public opinion as a social constructionâa product of a complex interplay of personal circumstances and ideological influences (such as the education system and the media)âit becomes a more difficult objec...