POLITICAL INSTITUTIONALISATION OF PUBLIC OPINION THROUGH POLLING slavko splichal
Controversial Effects of Polling on Public Opinion
The idea of public opinion is inseparable from techniques, instruments, and institutions related to the expression and representation of opinions. Regardless of whether the notion of public opinion presupposed the public as a corporate social entity or merely as a (statistical) aggregation of individuals, or whether it considered public opinion as originating from rational discussion or merely as a widespread diffusion of elite opinion, even by coercion, it was always assumed that public opinion is (at least) publicly expressed opinion that represents the (majority of) people or the citizenry.
Since early modern times, public opinion was typically institutionalised in three distinct forms, but none of them genuinely represented an ideally defined public:
1.	Parliament. If anything, parliament may generally be called an "organ of opinion of the public," although even this is not always justified (Tönnies 1923, 77). Tönnies' belief reflects a widely held assumption, advocated by normative political theories and sociological theorizing of public opinion, notably by American Pragmatism (Park and Dewey). However, representative government has never been a system in which parliamentary representatives had to regard opinions of the electorate; it has never been a direct form of popular sovereignty. Rather, representative government, since its foundation, has been a rule by elites, distinguished from the majority of citizens by their social status, education, or way of life.
2.	Newspapers. Since its very beginning, the press played an important role, according to Tönnies. It delivered not only information to the public and, thus, was an important element in the process of public opinion
Slavko Splichal is Professor at the Department of Communication, Faculty of Social Sciences, University of Ljubljana.
formation, but it was also the main "means of expression" or "organ" of the public, constituting a virtual public. In addition, the press was a general medium for more restricted means of expression, like associations, meetings, or demonstrations, while the public was always, and nearly exclusively, a "newspaper reading public." However, newspapers not only express the opinion of the public, but also influence public opinion. In reality, newspapers are neither organs of opinion of the public nor are they identical with it; rather, they are primarily organs of political parties and commercial corporations.
3. Polling. Public opinion polling developed during the decline of the critical (reading) public. Public opinion became an object of research after it largely lost control over its former institutions, i.e., parliament and the press. One of the first prominent U.S. pollsters, George Gallup, believed that polling ought to compensate for the growing limitations of a parliamentary democracy. During the last fifty years, polling was largely institutionalised in Western democracies. But similarly to the press and parliament, polling was soon criticised for its effects on public opinion and democratic life, in general, and "accused" of manipulation and antidemocratic support.
Beniger (1992) identifies five types of "social and behavioral changes" potentially brought about by the development of opinion polling; they are changes in: (1) the definition of public opinion and (2) what may affect public opinion, (3) the effects public opinion may have in society, (4) the behaviour of individuals because of polling and (5) because of the publication of polling results.
We may definitely agree with Beniger on the substantial questions of what constitutes public opinion and how the spread of opinion polling in the 20th century has influenced and/or transformed the nature of public opinion. However, Beniger's attempt to answer these questions is controversial and inaccurate since he claims that — in contrast to the "classical conception of public opinion of ancient Athens" — public opinion has increasingly become (1 ) something in which not just the (lite but everyone might at least potentially participate; (2) an aggregate of individual opinions in which all are assumed to be of equal importance and uniformly informed; (3) something which might be unconscious in individuals and yet subject to external measurement and manipulation; (4) something abstracted and isolated from actual political controversies and discourse; and (5) something wholly independent of the uses to which it might be put (1992, 208; emphases added).
Despite two substantial flaws in Beniger's analysis — denoting public opinion as "something" without specifying how it is constituted, and misconceiving the first two characteristics of polled opinion as contrasted to the equality of citizens in ancient Athens rather than to the opinion of the liberal bourgeois public — Beniger is certainly right in believing that questions of how polling influences "conceptions of public opinion, and public opinion's role in modern economies, polities and societies" are "crucial to any reasonable conception of a democratic or free society" (Beniger, 1992: 218). Tönnies, for example, argued similarly that the opinion of the public became a new world power with eighteenth-century rationalism, when governments became more "popular". The development of newspapers as new means of expressing opinions, made possible by developments in economics, technology, and politics (particularly political rights and freedoms, e.g., freedom of expression and the press), was a necessary condition. In other words, a new "technology" of forming and expressing opinions may commonly be associated with changes in the nature of public opinion.1
Explicit or implicit assumptions about the nature of public opinion introduced by polling as a new means of expressing opinions brought about two types of partly interrelated changes:
•	the transformation of traditional or classic understandings and practices of public opinion processes and
•	the development of a new kind of social institution functional in the political system.
Polling helped transform autonomous public opinions into much more manageable "mass opinions" that could be created and shaped to suit particular interests. Moreover, polling became largely institutionalised2 in all democratic societies. Controversies over value and function of opinion polls for democratic political processes are as old as polling itself, and they largely reflect much older efforts at theorising public opinion and democratic government. The first eminent "representatives" of fundamental differences concerning the political function of polling were George Gallup and Lindsay Rogers; their disagreements in the 1940s reflect ideas presented by Dewey and Lippmann during the 1920s. For instance, Gallup claimed polling results as a "mandate from the people" to the government; that is, in a society in which direct democracy is impossible, polling ought to compensate for the limitations of electing political representatives. Public opinion revealed through polling was believed to provide a democratic counterweight to the growing independence of political representatives and, therefore, a separation of representation from popular rule. From that perspective, as Albig argued, public opinion polls may be an indication of democratic developments. In pre-democratic societies, customs, beliefs, and convictions are subjects of early indoctrination and remain very stable over a person's life-time, so that "(p)eriods of limited opinion do not have need for the recorders of opinion, for the straw-vote takers and pollers, for the study of opinions as important phenomena" (Albig 1956, 175). Similarly, V. O. Key related the interest of governments in the distribution of public opinion among their citizenry to "the ethical imperative that government heed the opinion of the public (which) has its origins in democratic ideology as well as in the practical necessity that governments obtain support of influential elements in society" (1967, 4). In contrast, Rogers defended the liberal-conservative criticism of the tyranny of public (= majority) opinion (particularly Edmund Burke's position). Since political leaders are expected to be responsible to the "true" and general public interest rather than to any particularistic interest, representatives should not follow the dictates of the public (Crespi 1989, 3-5).
Early Critiques of Polling: William Albig and Herbert Blumer
The early critiques of public opinion polls concentrated on problems of validity, i.e., on the question of whether polls actually measure what they aim and claim to measure — public opinion — and on the political implications of polling. Proceeding from his differentiation between "opinion" and "attitude," Albig maintained that no record of opinions is adequate unless it leads to accurate assumptions regarding the attitudes which underlie opinions (1939; 1956, 174). He saw two fundamental problems in attempting to measure public opinion: (1) the development of attitude tests sufficiently comprehensive to include, at least, the more typical attitude patterns of most individuals comprising a public ("of millions of individuals"); and (2) the development of sampling methods adequate to the task of reporting on large publics by means of the smallest feasible, representative sample.
While the problem of adequate sampling seemed "rather simple" to Albig, Blumer (1948, 546) emphasised that "the inherent deficiency of public opinion polling, certainly as currently done, is contained in its sampling procedure." On the other hand, both Albig and Blumer were concerned with the validity problem of "attitude tests," an enduring controversial issue. Albig was among the first prophets of polling who, nevertheless, critically warned against uncritical massive use of opinion polls that might lead to their invalidity. Public opinion polling may be devaluated on the ground of (1) the reduction of data gathering to interview response data, neglecting alternative methods, and the limitation of data analysis to "response counting", and (2) misunderstandings in interaction between the researcher and respondent, and manipulative question-wording.
Albig discussed seven crucial questions of polling which he related to news reporting (1939, 229-234) by looking at the problems of validity, social and political implications, and ethics of polling, primarily through newspaper publications of poll results.
(1)	The importance and quality of public opinion. When answering the question of what is and should be the role of the "opinions of members of large publics on public affairs," Albig suggested that pollsters too often fail to assess the quality and quantity of public opinion: they tend to overemphasise the significance of public opinion on hundreds of issues on which no really significant opinion exists. Thus, it is dubious, whether reporting the very large number of polls in the media is in the public interest.
(2)	Display of public ignorance. One of the most valuable results of the polls is, according to Albig, the reporting on broad areas of popular ignorance abundantly revealed by polling. Although the portrayal of popular ignorance of public affairs is more extensive than systematic, it is, nevertheless, an important public service performed by polls.
(3)	Influence on legislative and administrative decisions. One of the politically most critical questions is whether polling tends to "distort and degenerate the quality of the decisions of legislative representatives and administrative leaders?" Albig suggested that the absence of polls certainly would not keep politicians from attempting to assess public opinion; they would have to find other ways of polling the opinions of their followers, as they did in the past, although less frequently and unsystematically. However, "if the polls are more accurate than other sources of reporting on opinion, they should be required reading for men in public life" (Albig 1939, 232). As a matter of fact, Albig assumed a considerable influence of the polls on legislative and administrative decisions; although he admitted the difficulty of assessing direct effects, he did not consider "the poller the villain in this drama."
(4)	Checking interest claims. Also related to the influence of polling is the question of controlling otherwise essentially uncontrolled claims of interest groups. Albig accepted Gallup's belief that polls support democracy because they "can limit the claims of pressure groups to the facts, and thus prevent many insupportable demands for special privilege." Polling could provide political representatives and administrators with information about the distribution of opinions to counter the claims of pressure groups, although perhaps not to the extent argued by Gallup.
(5)	Influence, conformity, and band-wagon effect. Albig agreed that reporting poll results may occasionally influence the opinions of readers, especially on topics on which they do not have (firm) opinions.3 The publication of poll results in the media may possibly have an agenda setting function and increase adherence to majority
positions, i.e., produce a band-wagon effect. Since polling greatly oversimplifies complicated issues, it is highly unlikely that the publication of poll results increases reflective thought. However, polls do not bear any crucial responsibility for simplifying the thinking of the public, according to Albig, because such oversimplifications are also characteristic of the mind-life of the majority of those polled, i.e., they are not caused by polls.
(6) Ethics. Two ethical issues were raised by Albig in relation to polls: (1) Do commercial pollsters correctly inform their publics about the quality and quantity of the opinions of their respondents? (2) Is there a danger of corruption in the polling process and, therefore, a need to regulate polling formally or informally? He believed that major commercial and academic pollsters were generally competent and ethical, but these questions became increasingly relevant with the growth of polling organisations and widely differing forms of professional competence, ownership, and control (Albig 1956, 234). Regarding adequate social control, Albig did not see realistic possibilities for voluntary trade association agreements or legal regulation.
In contrast to Albig, Blumer was concerned with the validity problem in the strict sense — "whether public opinion polling actually deals with public opinion" (1948, 542). He was among the first to severely critique the transformation of public opinion from "a property of groups" to an "attribute of individuals," explicitly carried out by those who justified polling as an instrument of measuring public opinion. Blumer conceptualised the formation of public opinion as a function of society in operation through the interaction of groups rather than disparate individuals, who share equally in the process of forming and expressing opinions. He declared, "By expression of public opinion I mean bringing the public opinion to bear on those who have to act in response to public opinion [...] through such means as letters, telegrams, petitions, resolutions, lobbies, delegations, and personal meetings (of) interested groups and individuals" (1948, 544-45). In other words, polling cannot be used as a valid "measure" of public opinion because:
Public opinion which was a mere display, or which was terminal in its very expression, or which never came to the attention of those who have to act on public opinion would be impotent and meaningless as far as affecting the action or operation of society is concerned (Blumer 1948, 545).
The basic deficiency of public opinion polling is contained, according to Blumer, in its random sampling procedure which advances the notion of society as a mere aggregate of disparate individuals. Opinion polls tend to extract the attitudes of citizens and make them available to subsidisers (governments, parties, or lobbies) who decide on their own what actions (not) to take, but they ignore all other non-subsidised or subsidised sources and forms of information about opinion(s) of the public, e.g., letters, petitions, resolutions, lobbies, delegations, personal meetings, press conferences, and perhaps most importantly, the mass media through which interested individuals and groups can influence those who have to act in response to public opinion. At best, then, opinion polling may serve as only one indicator of public opinion whose validity has to be proven in comparison with other indicators or means of expression. In that case, polling "supplements rather than supplants other modes of opinion expression, each with its own limitations" (Miller 1995, 111; emphases added). It obviously cannot serve as a (universal) procedure to measure public opinion, or objectify itself to become a neutral means of expressing an individual's' opinion. At worst, it may be used as a means of manufacturing "public opinion."
The central component of Blumer's critique of "public opinion polling" as a method of recording and measuring public opinion relates to the problem of (in)adequate sampling:
(1)	Aggregation of individuals. Sampling procedures enforce a treatment of society as an aggregate of disparate individuals (Blumer 1948, 546). Compared to "common sense empirical observations of public opinion," there is obviously no guarantee in polling that those individuals who truly participate in the formation of public opinion are included in the sample.
(2)	Demographic variables. Age, sex, occupation, economic status, educational attainment or class status are rarely "the marks of significant functional position in the formation of public opinion on a given issue." Since these are the only variables included in public opinion polling, "We know essentially nothing of the individual in the sample with reference to the significance of him or his opinion in the public opinion" (Blumer 1948, 546).
(3)	Aggregate data. Blumer emphasised that "the collective findings have no assurance of depicting public opinion on a given issue because these findings ignore the framework and the functional operation of the public opinion" (Blumer 1948, 547). An individual who is responsive to public opinion must assess public opinion as it comes to his/ her attention in terms of the functional organisation of society to which he/she is responsive. Different individuals and groups will certainly find different specific issues most relevant for them. Thus, on the one hand, public opinion polls must ignore concrete, specific questions relevant only for specific environments, because they are directed to the aggregate level of society. On the other hand, many questions will remain unanswered by respondents because they lack relevance for specific environments or groups.
(4)	Validity is not transferable. Polling is regarded as intrinsically valid (only) because of its rather spectacular success in predicting elections. However, "the casting of ballots is distinctly an action of separate individuals wherein a ballot cast by one individual has exactly the same weight as ballot cast by another individual" (Blumer 1948, 547). In other words, the validity of sampling ballots is not proof of the universal validity of random sampling! Polls exhibit a high, predictive validity when used to predict election results, but (a) the predictive validity of a measuring instrument is only one dimension of validity which does not yet guarantee its validity, in general (see Splichal 1990), and (b) the accuracy of a prediction of one type of findings is not transferable to (all) other types of findings.
(5)	The loss of generic subject. In toto, public opinion polling is not able to "isolate 'public opinion' as an abstract or generic concept which could thereby become the focal point for the formation of a system of propositions" (Blumer 1948, 542). Blumer proved this general observation by arguing that there are (a) no efforts to try to identify public opinion as an object of study; (b) no specific studies are used to test a general proposition about public opinion; and (c) no generalisations exist about public opinion despite a large amount of polling studies.
Nevertheless, in opposition to both Albig and Blumer, "scientific pollsters" continued to discuss and improve primarily sampling reliability as it were the most important factor of an unsatisfactory (predictive) validity of polling. They largely ignored Blumer's theoretical critique that emphasised a highly problematic operational conceptualisation of public opinion in polling, and Albig's questioning of the accuracy of measuring instruments used in polling (e.g., how wording and contextualisation of questions influence opinions).
Modern Critique: Public Opinion Polls and Democracy
Contrary to critiques of public opinion polling in the 1930s and 1940s and their primary concern with the question of validity, more recent critiques focused on social and, particularly, political implications of polling. Undoubtedly, the importance of polling is closely related to profound societal, and particularly economic changes during the period when polling was "invented." It is another question, however, whether changes in 20th-century political systems, which brought about the prevalence of opinion polling, denote a democratic development or not. Although polling probably represents a more democratic means of expressing opinions than some alternatives, such as demonstrations, for example it may, at the same time, deprive public opinion of its potential to create an obligation for the authorities to heed. Thanks to polling, "public opinion" became more predictable and could be managed more easily: instead of promoting the influence of public opinion on governmental policy, polling may help governments adapt public opinion to their interests. The interests of the modern democratic state, managed by professional political elites who seek to minimize electoral and social disruptions from below, may be defined and pursued by various strategies of opinion management. A tendency of marginalising the power of public opinion is typical in a "managerial democracy" which generated a major transformation from an adversary relationship between public opinion and government, typical before the 20th century, to a managerial one. As Ginsberg (1986, 62-85; 1989, 274-293) argued, the rise of opinion polling cannot be connected to a truly democratic development because polling:
1.	changes both what is expressed and what is perceived as public opinion "by transforming public opinion from a voluntary to an externally subsidized matter;"
2.	transforms public opinion from a behavioural to an attitudinal phenomenon;
3.	shifts public opinion from "a property of groups" to an "attribute of individuals;"
4.	transforms public opinion from a "spontaneous assertion" to a "constrained response," partially removing the control of individuals over the subject matter of their public expression.
Ginsberg differs from Gallup and Rogers (as well as Albig). His critique of polling tacitly assumed — in contrast to Rogers — that public opinion should influence governmental policy; but — in contrast to Gallup — he does not see any possibility for polling to perform such a function. The more polling becomes the dominant mode of "measuring public opinion," the more these four transformations result in a "domestication or pacification" of public opinion, according to Ginsberg, and change its relationship to government. Opinion polling is a typical example of (direct) information subsidies: it subsidises the costs of public presentation of (mass) opinions with which power actors may attempt to influence the actions of others and/or ensure their willingness to occupy themselves with specific themes, primarily through defining "public issues," i.e., the issues for public discussion. In other words, "Clients who provide the essential funds for survey research, therefore, exercise at least de facto control over what research gets done and how it is conducted" (Miller 1995, 109). The argument resembles Tönnies' critique of the press, which he considered an organ of political parties, and the dependence on political parties, which prevented the press to be a means of public expression or an organ of the public.
Pacification of public opinion is basically related to the fact that polls assume an equal value of all opinions regardless of their intensity. If the costs of public expressions of opinions are borne by opinion-holders themselves, those with more intense or extreme opinions are more likely to bear the costs of publication. Consequently, more intense opinions were more likely to appear in public. If the costs are subsidised or, as in the case of polling, entirely financed externally rather than by opinion-holders, less intense opinions are as likely to be publicly presented as the most extreme opinions. A large proportion of individuals who "don't know" how to respond to a survey question would never participate in public discussion or actions initiated by citizens themselves. Similar to Blumer's criticism in1948, Ginsberg argues that
The polls, in effect, submerge individuals with strongly held views in a more apathetic mass public. [...] A government wishing to maintain some semblance of responsiveness to public opinion would typically find it less difficult to comply with the preferences reported by the polls that to obey the opinion that might be inferred from letters, strikes, or protests. Indeed, relative to other modes of public expression, polled opinion could be characterized as a collective statement of permission (1989, 276).
The concept of a "permissive consensus" originates in Key's theory of public opinion, where he defined it as a specific form of consensus that is not directive but only loosely connected with governmental decisions or actions (Key 1967, 32-35). A simple majority agreement with an opinion or option in an opinion poll may largely arise from persons not strongly attached to the opinion they agreed with. On the other hand, "a 10 per cent dissent may include small pockets of the most determined opposition whose members command controlling points in the governmental mechanism." When such a simple majority agreement exists, a government may be relatively free to act or not to act. Although Key's conception of permissive consensus is clearly related to opinion polling, it is not a deficiency in the democratic political process. However, he recognized the need "to go beyond the survey data and make assumptions and estimates about the role of [...] the political elite," when not enough persons from this stratum are included in a random sample to enable a systematic analysis of the "leadership eshalon."According to Key, the under-representation of "political influentials" — "the activist subculture" — who significantly affect public opinion is the most important "missing piece of the puzzle" of public opinion (1961/1967, 536). Similarly, Key's ideas of "multiple consensus" and the differentiation between attentive and nonattentive publics also imply a critical attitude towards the conceptualisation of public opinion as an agreement among individuals which does not differentiate among the values of individual opinions.
The relations between public opinion and government have been fundamentally changed in the 20th century, both in terms of how the public is informed and influenced, and how public opinion is expressed. Key related these profound changes to the concentration of power in large corporations and the increasing complexity of government. Before the advent of polling, public opinion was primarily expressed in, and inferred from different forms of political behaviour. Through the 19th century, public opinion was often equated with mass behaviour, e.g., riots, strikes, or demonstrations. In the sociological tradition of the early 20th century, the expression of public opinion was considered "in the form of direct influence on those who are to act in response to public opinion" (Blumer 1948, 545). Tönnies identified a number of "means
of expression" used by public opinion, like the formation of associations, assemblies, and demonstrations, but he considered newspapers by far the most important means, regardless of their controversial function, because they were also used by political parties to manipulate and deceive the public. With the rise of daily newspapers, previous institutions that organised meaning, identity, and authoritative information for the public, which shaped political preferences of the people and simplified the process of democratic power-seeking — political parties, mainstream religion, the nuclear family, the workplace, neighbourhood and social-class groupings — have all but waned in importance and influence. The press and other mass media developed as power centres, and media-based strategies for shaping public opinion and winning support became increasingly important.
Since the American elections of 1936, "scientific" opinion polls — which followed earlier "straw polls" — occupy a politically most significant position. During that period, newspapers, periodicals, and other commercial companies, in addition to political parties and candidates, became key financial supporters of polling and increasingly subsidised the expression of opinions regarding issues they selected according to their commercial (advertising) or political (propaganda) interests. While before the 20th century public opinion could often, or even primarily, be inferred from non-verbal behaviour, the press and polling largely verbalised and standardised public opinion or, as Ginsberg (1989, 278-279) put it, "polling transformed public opinion from behavioral to an attitudinal phenomenon. The polls elicit, organize, and publicize opinion without requiring any action on the part of the opinion-holder" (emphases added). As any other form of information subsidy, polls may help direct attention, control specific demands, and influence behaviour. Poll results do not directly constrain the expression of public opinion through behaviour, but they do reduce the probability of behavioural expression, because they provide an opportunity to "subsidisers" to predict changes in popular opinions and react accordingly. A fundamental dimension of critiques of polling as an anti-democratic process is related to the question of whether public opinion is a mere expression of opinion in the form of answers in polls or whether it includes action. Arthur Schlesinger, Jr., argues (Glynn and McLeod 1984, 65) that polling "elicits essentially an irresponsible expression of opinion. [...] The measure of responsible (in the sense of 'real') opinion is not answers, but acts."
While it is obvious that, in modern democratic societies, a growing tendency of the reduction of public opinion from a more complex (behavioural) phenomenon to a merely attitudinal phenomenon exists, the "theoretical solution" suggested by Schlesinger is rather disputable. The idea to relate a genuine public opinion to the chance that opinion formed by public discussion may be effectively translated into action, even against the authorities, if necessary, goes back to C. W Mills' (1956) differentiation between "public" and "mass." For Mills, the authoritative control over the communication channels and over any action to make the opinion effective (i.e., blocking any practical action) was not a consequence of polling, but rather brought about by the industrialisation of culture and structural changes in the society. In the first place, then, the "irresponsibility" of opinion expression is not a unique characteristic of polling.
In addition, it is questionable if actions are a more valid instrument of attitude expression than opinions which respondents express anonymously, or that a sharp
difference or even opposition may exist between the two. Albig would certainly not agree with such an assumption. In his Public Opinion, he argued that even anonymity may not modify the subject's tendency to give a conventional answer which is also typically the case in more "public" expression of attitudes, or else that individuals' public actions are no less congruent with socially desirable or at least acceptable behaviour than are the anonymously expressed opinions.
Most investigators have assumed that the individual's hidden attitude rather than the conventional response given in such situations, is the real attitude. If by "real" is meant that which more likely results in action, it by no means follows that the individual's hidden attitude is more real than the conventional response. Although usually more willing to disclose the conventional attitude and thus avoid antagonistic responses, the individual may also be much more willing to act in accordance with the conventional attitude. Thus, action as well as opinion is a fallible indication of all the attitudes involved in a situation (Albig 1939, 171).
Since the famous Literary Digest straw poll in 1936, politicians became increasingly interested in public opinion polling. Such an intense interest in a (new) means of expression is not a phenomenon characteristic of polling: before, political parties and governments were inclined to influence and control public opinion through the contents of newspapers; at the end of the 20th century they are becoming more sensitive to feedback through new computer mediated communication channels. Similarly to opinion polls, telephone, telefax, and increasingly electronic mail, enable politicians to establish "contacts" with the electorate and measure its "pulse." These modern means of expression and control of public opinion — mass media, opinion polls, and computer mediated communications — spurred populism. Broadcasting, in particular, elicits the experiences, views and priorities of ordinary people and encourage them to discuss social and political problems via various new communication vehicles — call-in-programmes, electronic town meetings, televised citizen juries, and, especially, talk shows. Polling enables political elites to anticipate and avert the electorate's displeasure, even if they subordinate themselves imprudently and disfunctionally to popular sentiments by passing dubious laws — not to solve societal problems but to avoid popular irritation and retain the electorate's votes. Electronic mail further stimulates this sort of speculative coorientation: the new "democratic rhetoric," emerging both in legislatures and public discourse (or public relations), draws primarily on public sentiment. Yet, there is no simple answer to the most simple question: whether these spectacular developments in information and communication technologies in the 20th century revive direct democracy and open universal possibilities for direct citizen participation in political decision-making — or are routinely used to influence and manipulate public opinion. On the one hand, these new means of communication may help the public influence decisions by those in power to respond to its demands, which is perfectly congruent with the democratic idea and reflects optimism among advocates of polling. On the other hand, polling may also help develop a generally "permissive consensus" which may provide a government more freedom to act (or not to act) and, thus, opportunities to manipulate (educate) public opinion.
Undoubtedly, polling has controversial effects. Indeed, the deficiencies of polling are not just an academic matter, or only a question of an (in)appropriate method. Polling may provide a "more representative picture of the public's views than would usually be obtained from group leaders and notables," but the price for such an "antidote
to inaccuracy as well as to mendacity" of group representatives and leaders may be rather high. By delegitimising the claims of group leaders and activists to speak for membership opinions, polling reduces the political effectiveness of public opinion. It seems, then, that — in principle — both Gallup and Rogers and their followers were right. However, the specific historical conditions are decisive for determining the (anti)democratic nature of polling: as I have demonstrated elsewhere, particularly in the former socialist countries opinion polling had a clearly anti-authoritarian character. On the other hand, as Ginsberg argued, the introduction of polling was most damaging to the representation of working class interests in capitalist societies, because it "erodes one of the major competitive advantages that has traditionally been available to lower-class groups and parties — a knowledge of mass public opinion superior to that of middle- and upper-class opponents" (Ginsberg 1989, 284).4 Although polls may be used, in principle, by any commercial or political group, they were particularly useful for non-working-class associations and parties which were more loosely organised and, thus, had less knowledge of public sentiments and public opinions. In modern democracies, however, no political party — regardless of its political orientation — can avoid the use of polling due to its massive political institutionalisation.
Because of the transformations introduced by polls into the formation and expression of (public) opinions, polls became a major power in structuring public issues and providing agendas of public discourse. Similarly to media or other carriers of subsidised information, polls reflect content and priority of issues determined primarily by groups that subsidise polling. In fact, since the late 1800s, newspapers were both, major promoters and supporters of the polls. However, it is probably an overstatement to conclude, as Ginsberg (1989, 287) did, that polling "erodes individuals' control over the agenda of their own expression of opinion" It is one thing for polling to produce a misleading picture of a public agenda, it is another thing to "erode control." Polling results may still create an obligation for authorities and may be considered an expression of public opinion, which is one of the central ideas in a theory of public opinion. Most likely, polls create more effects in the other direction, i.e., on public opinion: the selection of questions (and answers) for questionnaires and, particularly, the reporting of polling results by the media help determine the relative importance of "public issues" the public(s) ought to discuss. However, media, political elites, and even the public(s) are autonomous in developing their own agendas; there is no evidence of a causal connection between various agendas in any possible direction. Nevertheless, one may expect that those subsidising polling will influence the agenda set by polls rather than vice versa — which does not mean that pollsters are merely objective inquirers into public opinion without their own ideologies and particular interests. In other words, polling plays largely an instrumental (or secondary) role in the process of agenda setting; it intensifies rather than establishes the power of political and commercial elites. Ginsberg provided relevant evidence when he reported how in the late1960s and early1970s U. S. polls "took little interest in the issues which aroused so much public concern," thus, clearly indicating the existence of different agendas.
An erosion may have occurred, however, when particularly during the 1940s, some students of U.S. opinion polls considered polling a partial substitute for democratic political procedures, including the conduct of elections. It implied a constitutional position for polling that equals "traditional" political procedures and would, indeed, essentially erode the control of individuals over their agenda(s).Wilson argued that in
such a case, "the public would have no initiative of its own, and there might be some question whether there is any public at all. It has been sarcastically suggested that all we need in modern government is a competent body of civil servants and a United States Polling Authority" (1962, 171).
Public Opinion Polling — Monitoring or Manufacturing Public Opinion?
Setting aside abstract, ideological protests from criticisms of public opinion research, the following major points of critiquing polling remain under review:
1.	Opinion polling presupposes that each individual has, or must have an opinion about everything (Bourdieu 1979, 124; Keane 1984, 148; Peer 1992, 231).
2.	Opinion polling presupposes that opinions could be statistically sampled, tabulated as results, and mathematically reconstructed (Keane 1984, 148).
3.	It is presupposed that all opinions have an equal value or realistic importance, respectively (Blumer 1948, 543; Bourdieu 1979, 124).
4.	Polling presupposes that opinions "extracted" from respondents are their "true" opinions, independent from interactions with those conducting polls.
5.	Polling presupposes that there exists in society a consensus about which questions are important and, therefore must be put to the respondents (Bourdieu 1979, 124).
On the distribution of opinions. The thesis that each individual must have an opinion about everything, widely postulated as a necessary although not explicit presupposition of opinion polling, is simply erroneous. Leemor Peer mentioned two reasons why opinion polling must assume that (all) people have opinions. Accordingly, the theoretical reason ought to originate from "the democratic principle of self rule [...] that all people have opinions, that they have the same value, and that they should be expressed and acted upon" (Peer 1992, 231). If this were the basic assumption of democracy, to be materialised in opinion polling as an instrument of democratically expressed opinions, how then is one to explain a citizen's right to abstain from voting? For classical Greek democracy, the basic principle was liberty which included, according to Aristotle, two defining elements: "ruling and being ruled in turn" and "to live as you like" (Held 1987, 19). Based on equality, these two elements are noncontro-versial only as long as each individual, indeed, has the opportunity of "ruling and being ruled in turn." Otherwise equality may conflict with liberty, which was the case throughout history, including democratic regimes.
In ancient Athens, only men over the age of 20 were eligible for citizenship: women, immigrants, younger men and, particularly, slaves were politically marginalised. The notion of equality as the basis of liberty was restricted to those with equal status (male and Athenian born), but even they did not really enjoy the equal opportunity of ruling (Held 1987, 23). For instance, the Assembly was too large to organise and propose public decisions; that responsibility was with a Council of 500. Since the Assembly had a quorum of 6,000 citizens, the majority could abstain from voting and elections. In sum, even in the legendary Athenian democracy, only citizens who represented an exclusive minority of the Athenian populace were expected to have opinions.Citizens had the right to speak and be heard by attending a political assembly, but only few of them would, in fact, speak. Similarly, in contemporary democracies, citizenship rights
and freedoms represent key features, particularly freedom of expression, freedom of association, and the equal right to vote (one person, one vote), but there is no assumption that everybody expresses opinions, associates, and votes. Even if democratic theories assume that citizens are well informed and interested in public issues, which enables them to participate in decision making, they do not assume that all people have or even express opinions on all issues. Consequently, there is no theoretical reason why opinion polling should assume more active citizens than (all the other) traditional democratic institutions. On the contrary, as Albig claimed, the portrayal of popular ignorance of public affairs is one of the most valuable results of polling. Apart from it, it is probably safe to maintain that, despite the increasing disproportion of the number of people who express opinions and the number of those who only receive them, more societies have become more rather than less democratic over time; the institutionalisation of public opinion polling in the 20th century does not change this fact.
Nevertheless, the "practical" reason why opinion polling ought to assume that people have opinions seems related to sampling: its methodology assumes that every unit in the population should have a specific value on the variable to be measured. Accordingly, Peer (1992, 231) believed that opinion polls "are not designed to determine whether opinions exist, but to measure the variation in existing opinions." She wrongly believes that the (non)existence of opinions is a practical question of sampling. As it is true for any variable, what is found in a questionnaire is only an indicator of a variable (concept) that always requires theoretical definition. It is another question, of course, whether such a theoretical definition exists or not. In other words, what are the "true values" of a variable is not (primarily) a practical, but rather a theoretical question; it does not change the problem if the variable is (a specific) opinion or attitude intended to be measured. The respondent's refusal to answer, or a "don't know" answer are not obvious cases of "non-opinions" and, thus, simply missing values; they may be considered true values, at least as much as "random responses." Methodological reasoning may even suggest quite the contrary, namely that opinions of a population are spread in the form of a normal distribution, which would imply that on the majority of issues the majority of people has no opinion and only small minorities have definite positive or negative opinions, respectively. Thus, according to a critical "insider" of the public opinion industry, polling assumes that "the public is a population of individuals (often citizens or residents of a geographic region) who may have an opinion about the subject matter of the survey" (Miller 1995, 107). Peer contradicted herself when she argued that the analysis of respondents who tend towards a non-response "puts them in categories," a clear indication of "normalisation," that assures the perpetuation of the "democratic discourse" (that all people have opinions and that all opinions count equally). If polls really "are not designed to determine whether opinions exist, but to measure the variation in existing opinions," where does an interest in non-opinions come from? On the other hand, what is the consequence of the same sort of analyses ("putting in categories") of respondents who do express (different) opinions? Is the "act of categorising people into groups" according to different opinions they have expressed an act of power as well?
Yet, I am far from believing that no serious theoretical and methodological problems are associated with polling. However, the problems are not caused by a tacit assumption that "people have opinions." Interpretations of polling results ignore most
frequently answers such as "I don't know," "I'm not aware," or "I cannot answer," instead of accurately analysing them; especially, since they are not the result of randomness, but represent the actual position of the respondent in society and are related to his or her "political competence" (Bourdieu 1979, 125). It is not, therefore, a matter of inability in principle, but a question of why the possibility of analysing no-answers is not affirmed as an integral part of testing validity in research practices (e.g., Michelat and Simon 1985). As later established by Bourdieu (1985), this is, above all, a question of the autonomy of the social sciences in a given society. Because of financial demands, empirical research relatively often succumbs to the influence of those who order and finance it — not only in the selection of "important" questions. but also in consenting to descriptive presentations of interview response data and rejecting methods which would allow an explanation of the findings, since those ordering research are usually not interested in them. This position certainly counters Peer's controversial thought that respondents with non-opinions should not be analysed and "categorised" since that would help perpetuate the disciplinary discourse: exactly the same would hold true for any analysis or categorisation of respondents regardless of whether they do or do not express opinions. The development of polls into a form of social institution did not necessitate the assumption that all people have opinions. In other words, the problem of disciplinary power is associated with polling in general (which became a social institution) rather than with a specific methodological feature within polling procedures.
It is undoubtedly difficult to find in practice a clear and fully expressed class, strata or group understanding, in the broadest sense, unless there are fortuitous historical circumstances, but they must be helped to emerge from the "public" through research interviews. As important differences exist in everyday life between what people say and what they think (or just imagine) about themselves and others, so it is necessary to distinguish in empirical research the imagination of the individuals and groups from their actual attitudes, opinions and interests. This is equally true of published opinions, usually the opinions of the power elites, e.g., in the mass media, as well as private, anonymous expressions of respondents in opinion research. Both inevitably have a socio-political nature. Both may be adapted, changed, or manipulated. Politics, in order to be effective, must consider the situation and development of mass consciousness and mass opinions — whether politics is authoritarian or emancipatory in nature. Empirical opinion research, because of political interest in it, is always a political activity, even if researchers wish to be liberated in their relation to the "ruling power."
The most apparent expressions of political interests in opinion research in contemporary democracies are manifested in: many surveys conducted for and sponsored by government agencies and officials; systematic monitoring of public opinion polls by governments, primarily to identify the form and intensity of public support for or opposition to governmental actions; and the legal regulation of public opinion research, in particular research and forecasting of voting behaviour (Lazareff 1984). The legal restrictions of voting research, and above all the publication of polling results in the mass media, not only pertain to the question of actual press freedom, but also contain a warning about the possible intervention of opinion polling in the mass "production of opinions" as a critical impulse by dominant industrial (i.e. state and party) forces. The subordination of opinion research and its publication to the ruling legal order undoubtedly demonstrates that public opinion polling does not in itself
have an inherently administrative character; it may even be opposed to it, like the first "social statistics" of the 19th century which were, to a great extent, directed against the primordial forms of capitalist accumulation.
On the consequences of the "mathematical reconstruction" of opinions. Keane demonstrated on the basis of well placed arguments the commercialisation of forming opinions, but he did not see the difference between the actual process of a disintegrating public and consenting to the results of this process, which is implied by uncritical opinion polling. Keane's reproach of statistical mystifications in opinion polls, which represent "antidemocratic vindications of the measurement and manufacture of public opinion"(1984, 148), is an example of the simplified (and thus mistaken) understanding of the relevant possibilities of statistical analysis of interview response data. Polls are usually very accurate in estimating (or "predicting," as criticised by Keane) voting behaviours of citizens or consumer decisions, because they simulate the very concrete private act that an individual will "practically" undertake in a voting booth or a store. Interaction during the interview does not necessarily represent a significant deviation from individual voting or buying practices, because in both cases the individual privately and anonymously responds to a well defined stimulus (questions). The accuracy of any estimate is clearly based on the use of sampling procedures and statistical data analyses. However, using the same data gathering techniques and analyses in (public) opinion studies represents an epistemological problem, because — as Salmon and Kline (1984, 12) reproduced Blumer's argumentation — elections are conceptually very different from public opinion phenomena. Public opinion is a complex communication process (rather than a concretely defined situation) in which individuals (1) are confronted with innumerable and constantly changing choices for action, including the possibility of not acting at all, and (2) when taking actions, they act publicly in conjunction with other members of their reference or interest groups. It is true that opinion polling frequently remains only on the level of "summing the empirically existing beliefs of individuals," but this is not to say that "statistical sampling," "tabulations of results," and "mathematical reconstruction" of opinions (whatever it should mean) generate an "automatic opinion of all and the considered opinion of none," as Keane (1984, 148) argued. He paradoxically referred to Tönnies as one of those who challenged "these antidemocratic vindications of the measurement and manufacture of public opinion," because Tönnies explicitly pleaded for and practised the application of qualitative and quantitative methods in empirical sociology and advocated a connection between pure sociology and sociographic research, e.g., for a combination of statistical geography, ethnography and anthropology.
In summary: empirical research and statistical data may significantly contribute to adequately understanding a problem provided that the interpretation of data is based on comprehensive theoretical knowledge. Needless to say, this requirement prevents the researcher from equating public opinion with a simple aggregate or an average of all opinions. It is true that opinion polling usually does not go beyond the level of "summing the empirically existing beliefs of individuals;" added description and the manufacture of public opinion through public opinion polling are not consequences of the use of statistical and quantitative methods in polling, i.e., a logical deficiency of research, but rather a consequence of the subordination of polling to the interests of those commissioning the research (i.e., to the research aims which are externally defined by political or commercial clients).
On the "equal value" of opinions. Of all the arguments regarding the (in)validity of operationalising public opinion through opinion polling, the most justified ones seem to be Blumer's and, thirty years later, Bourdieu's arguments that public opinion research wrongly presupposes an equal validity, or realistic importance, for all individual opinions in society (or, rather, in the sample of respondents). According to Blumer, the formation of public opinion does not occur through an interaction of disparate individuals who share equally in the process. Instead, the formation of public opinion reflects the functional composition and organization of society. The formation of public opinion occurs in large measure through the interaction of groups (Blumer 1948, 544).
Similarly, Bourdieu (1979, 126-7) claimed that not all opinions have equal importance. There are two general principles involving the formation of opinion which the model of public opinion polling does not take into account: (1) the uneven distribution of "political competence," which is based on differences in (political) knowledge and education — which is why the possibilities of an individual having any opinion at all are also not constant! — and (2) "class ethos," a system of implicit values which the individual internalises from earliest childhood onwards and which is the basis for generating answers to the most different questions. According to Bourdieu (1979, 128), a considerable difference exists between "the opinion which people produce in an artificial situation such as survey and the opinion they produce in a situation closer to the daily-life situation in which opinions are confronted and confirmed."
The assumption of the equal value of opinions, however, does not prove the "administrative character" of opinion polling, but only important deficiencies and, thus perhaps, the invalidity of concrete research. I do not believe that opinion research, in principle, uses an impossible, objectivised measurement of "political competence" among respondents as a basis for explaining the distribution of opinions (including "don't knows") in society. Yet, empirical opinion research can only reveal "political competence" as a fact without being able to reveal the process of its origination and production. To explain the presence/absence, or perhaps better, the various degrees of political competence among individuals, special research is needed, whose results may also be used sensibly for explaining the findings of "public opinion research," although certainly not in all its details. However, it is important to point out that the actual process of political socialisation, which results in an individual's "political (in)compe-tence," also remains hidden in the actual processes of generating opinions. Therefore, public opinion research is not more deficient than the actual process of forming opinions or the historical formation of the liberal bourgeois public.
Certainly, public opinion polling is usually not interested in what determines opinions held by individuals. Such a reduction of the problem disregards the key questions of who forms and who only accepts opinions expressed in private discussions (interview). If so, opinion is defined from the consumption (in accordance with understanding the public as an "opinion market") rather than the production side. Public opinion polling is restricted to studying results and is not concerned with their underlying social processes. This "lack of interest" in the causes — or at least the presuppositions — of forming opinions is also expressed in the fact that public opinion polling usually lacks any kind of critical reflection on the dominant problems operationalised in opinion questionnaires. Opinion researchers scarcely scrutinise those for whom topics defined in the questionnaires are dominant political topics
and others for whom "possible answers" in questionnaires are realistic answers in the sense of actually constituting possible ways of solving problems in society. Bourdieu firmly established the notion that dominant topics, as they appear in opinion polling, are problems which essentially interest people with power and who consider themselves well informed on the means of organising political action.
It is possible, however, to address the criticism of public opinion research concerning the neglect of a different validity of opinions — which means that different individual opinions, with respect to their objective circumstances, have different probabilities to become a group or even "common" opinion — quite differently. Thus, when public opinion actually does not exist, the opinion of the ruling power elite(s) is legitimated as "public," "common," or even as the "only intelligent" opinion. Bourdieu's criticism (1979, 128-9) that opinion research is incapable of generating any kind of sensible prediction of what will happen in a crisis situation — although it is capable of accurately forecasting whether (and how) opinions will change or not in stable situations — may also be understood as a belief that public opinion research aims to describe and forecast the actually dominant opinion. Such a demand is, of course, precisely the opposite of a criticism of public opinion research as administrative research, which is subordinated to authoritative institutions. Dominant opinions are created by power elites and the media, which also ensure publicity; thus, there is no need for opinion research to ascertain the dominant opinion. On the other hand, research into the opinions of the masses reveals the differences between dominant opinions and opinions which "do not have equal value." Revealing the existence of latent opinions which are not "politically competent" and, therefore, have no possibility of becoming dominant in a given society, is precisely the aim of all (including empirical) critical research.
It often happens that forecasts based on public opinion polls fail. The most famous failure was probably that in 1948 in the United States when all the "scientific pollsters," including Gallup, predicted that the presidential candidate Thomas Dewey would defeat Harry Truman for at least five per cents, but in fact Truman won the elections for almost five per cents. On the one hand, the difference between respondents' expression of intent as measured by polls and individuals' actual behaviour (e.g., citizens' voting behaviour) may be considered a difference between politically relevant and irrelevant opinions. From this perspective, Gallup was in a way right when he claimed that "the selection, not the polls, had turned out wrong" (Hogan 1997, 163). In other words, pollsters succeeded to identify those latent opinions (intended voting behaviour) which did not materialise in a manifest form but, clearly, did not succeed to predict what will happen in the "true" elections.
However, it would be deceiving to claim that polls represent the least accurate type of human behaviour prediction. Serious limitations which exist in opinion poll predictions are not unique to polling. Predicting human behaviour generally represents a central and unresolved problem in all social sciences, and relates both to theory and empirical research. While predictive validity is an important dimension of validity, it is still only one dimension; consequently, a low predictive validity doe s not yet indicate an absolute invalidity of a theory or research.
On the independence of respondents' opinions. In his critique of public opinion polling, Ginsberg challenged its validity because "the data reported by opinion polls are actually the product of an interplay between opinion and the survey instrument"
(1989, 273). One certainly must agree with Ginsberg's thought, but it should be added that polling represents only a specific instance of the general relationship between subject (researcher), method (the tool), and object in the development of scientific knowledge; the validity of research cannot be equated with objectivity in the sense of independence from the subject of research, i.e., the researcher. William Albig (1939) was among the first who, in a more specific and concrete way, warned of the limitations of validity in the use of questionnaires, particularly when respondents' reactions are limited to dichotomous "yes" or "no" responses. He saw two major sources of systematic errors that lead to invalid opinion surveys: (1) a prejudiced or even exclusive reduction of data gathering to survey responses, and (2) specific language difficulties which may cause misunderstandings in interactions between researchers and respondents.
Attitudes as expressed in opinions are, according to Albig (1939, 157), mistakenly considered as entities or units which may "profitably be handled quantitatively;" in fact, attitudes are always interrelated in a complex way and in varying proportions. An appropriate solution to the problem of "simply counting the responses" is the true measurement of attitudes, that is, an attempt to indicate the relation of answers to some standard. The Thurstone test largely but not completely overcomes "technical" objections to the construction of the tests, but certain problems remain. Apart from the methodological problems, validity of survey response data is seriously limited, because the judges' attitudes may significantly affect the scale. The standard in attitude tests is constructed by the experimenter or by a limited group of judges. The latter, however, do not provide a valid scale for respondents, because they themselves come from another group, often with varied attitudes. In addition, time and labour involved in the construction of a Thurstone-type scale are often discouraging or even prohibitive, which may help explain why public opinion polls, and even surveys, usually "simply count responses."
These problems, Albig believed, are not resolved by multiple-choice and cross-out tests, rating, ranking, and attitude scales. Approaches alternative to survey responses exist and should be used to infer attitudes from opinions expressed in written essays, letters, case-history descriptions, autobiographies, diaries, and oral or written interviews. However, while text or discourse analysis and in-depth interview permit the use of extensive, often unstructured, long and wordy messages and documents, and exert the limitations of simple interview response data, they present serious methodological problems in attempts to use statistical data analysis (Albig 1939, 158).
The language difficulty leads to more serious errors in simple questionnaires then in any other form of measurement. Different respondents interpret words differently, or misinterprete them, which may cause a complete opinion change because of misunderstanding language. But in addition to problems of (mis)understanding on the part of respondents to which Albig referred, questions may be worded intentionally in a manipulative way to predetermine the respondents' answers in an effort to achieve supportive or at least permissive consensus to legitimise policy decisions. At any rate, history of polling abounds with examples of significant question-wording effects, which provide ample evidence that respondents' answers depend not only on variations in question-wording, but also on the number, character and order of questions in the questionnaire, and response options to individual questions.5 Experimental studies also revealed a tendency of people to say "yes" when asked about unfamiliar topics ("response
acquiescence"), or that in the absence of the "don't know" response, many respondents would select an answer even on non-existent policies (Hogan 1997, 169-171).
On the consensus about dominant questions. We, thus, return to the fundamental question concerning social subjects, who order public opinion research and make use of its findings, who define research aims and so-called "dominant questions," which represent a sort of "agenda setting" for public opinion. Here again, the question is about the autonomy of social-scientific research and its critical self-reflection, that is, its endeavours to ask and answer the question of who opinion polls are for? However, if in a given historical period and a particular society, public opinion polling was (is) primarily performed for the carriers of dominant opinions, it is still impossible to derive a general conclusion about the administrative nature of public opinion research. Such a conclusion can easily be falsified by a closer examination of the social function of public opinion research in former socialist societies. This does not change the fact, however, that public opinion polling is a substitute for the opinion of the public, which is repressed by economic and/or political constraints, and that public opinion polling in such circumstances is necessarily a form of repression. However, if we consider both — repressive and emancipatory — dimensions, all forms of opinion research cannot be simply reduced to "administrative research," which is subordinated to the interests of commerce and the ruling power. In principle, polling does not inherently "manufacture" public opinion more than any other method of empirical research. According to Miller, "It is fair to say, however, that survey evidence is often used to manufacture public opinion when the constraints of the measurement process are ignored or hidden, and when the sample public is reified and inferences go beyond the intended survey framework" (1995, 110). This is exactly the same sort of criticism Allport addressed in the 1930s as "fictions and blind alleys" which were widespread "even in textbooks of political and social science," to formulate "a workable, scientific approach" to public opinion — i.e., a behaviouristic definition and research into public opinion (1937). He criticised those non-empirical understandings of public opinion which perceived behind it a collective entity, a "group mind," or even a personalised collective being rather than an instance of behaviours of individuals. While Allport believed that empirical research could demystify such conceptions, presentations, and interpretations of poll results which reify the Public, they apparently go in the opposite direction, thus producing the same "fallacy."
Indubitably, public opinion polling is concerned with opinions which are not the result of free discussions and reasoning or are publicly expressed in themselves. This enables survey organisations and their clients to make free use (or non-use) of poll results. In other words, polling is an accurate operationalisation of Key's definition of public opinion, according to which public opinion "may simply be taken to mean those opinions held by private persons which governments (or any other client) find it prudent to heed" (1967, 14). This certainly raises serious doubts about the first part of Moore's controversial belief that
Gallup's vision of polling as an instrument for improving the practice of democracy seems largely vindicated. Despite response effects due to question wording and placement and interviewer characteristics [...], polling can, indeed, provide a continuous monitoring of the elusive pulse of democracy (Moore 1995, 357; emphases added).
Public opinion polling is concerned with opinions which are not the result of free discussion and reasoning — they are private in the sense that they are expressed without intention to reach the public, and they remain unpublished. In addition, the respondent is also guaranteed anonymity. Such a guarantee of the anonymity of personal opinions is not a direct consequence of the pressure of "administrative research," but rather merely a research stimulus (not always effective) to encourage the "free" expression of opinions. The lack of publicness may, however, stimulate the transformation of passive or neutral "monitoring" into active social control that would further reduce an individual's participation in the formation of public opinion. By definition, this would corrupt rather than improve "the practice of democracy."
A possible, critical orientation of such research — although it always has a manipulative character, too — depends on the real interests of the (empirical) social sciences in revealing the emancipatory potential of the masses or the public. Empirical opinion research has, similar to other research methods and means of expressing opinions, numerous limitations in relation to "problems" it is able to "solve." For example, it cannot measure objectively the structural, relational characteristics of groups, but only the state of individual consciousness. It does not allow for individual expressions of opinion and/or individual actions, but limits "expression" to answering questions. Certainly, opinion polling cannot replace the information deficit which is the consequence of non-use, absence, or non-accessibility of other sources of scientifically relevant information or other means of expressing opinions. However, after considering the realistic interpretative possibilities and methodological presuppositions of public opinion polling, we can conclude that it can neither save nor bury a disappearing public opinion. Opinion polling as a measuring instrument and/or means of expressing opinions is neither good nor bad by itself. Neither it is neutral, as Michael Hogan (1997, 175) convincingly demonstrates with George Gallup's 50-year attempts to influence the ideological climate in the United States with the publication of the results of his polls. Because public opinion polling develops from institutional needs, its impact is always structurally determined and mediated through institutional arrangements. The key question for researchers focuses on the client for whom research is being done, which is also reflected by how the topic (object) of opinion research is restricted and tailored by the particularistic interests of the client as well as by science and scientists themselves.
While public opinion polling is largely politically institutionalised in modern democracies, this is not yet a guarantee of either its scientific validity or its democratic character. Polling is, at the same time, a commercial enterprise closely linked to recent trends in journalism to adopt polling results in the form of a special journalistic genre. Mass media do not publish polling results with the intention to influence policy makers, but to attract the attention of "the public." Scientific validity and political implications of poll results are less relevant for reporting, or not relevant at all; what counts is the possibility of a simplistic dramatisation of a news story. It seems that, with polling, history of newspapers repeats itself. The more the critiques of the commercialisation of the press and other media requested substantial reforms to (re-)establish the newspaper (or, for example, public service television) as "the organ" of public opinion, the more the media developed into the opposite direction. Similarly, public opinion polling primarily develops as an "industry of news events" rather than an "instrument of democracy," despite all the warnings and critiques.
Notes:
1.	This also pertains to the present and the future: computer mediated communication may again affect the formation of public opinion and its nature.
2.	Berger and Luckmann (1969) published a very illuminating discussion of institutionalisation, or formation of institutions, its relation to habitualisation, typification, historicness and social control. I refer here to their complex definition of the concept "institution."
3.	Albig quotes the results of a 1945 poll in Mexico, in which 81.2 per cent of social scientists and 72.5 per cent of journalists interviewed stated that public opinion polls had some influence on opinion.
4.	Similarly, Noelle-Neumann pointed out in her PhD dissertation defended in 1940 in Berlin that, in contrast to the United States where public opinion polls had to be used in order to obtain information about people's opinions, "our biggest popular leaders had and have such a vital relationship with the masses they lead, that they can comprehend them without such instruments of knowledge" (Noelle 1940, 1).
5.	One of the most often reported cases of almost mysterious effects of variations in question-wording on responses is that reported by Rugg who analysed split-ballot data gathered by Elmo Roper. 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. While 46 per cent of respondents opposed "forbidding" such speeches, only 25 per cent of them would "allow" them — an astonishing difference of 21 per cent! However, in a similar experiment on "forbidding" or "not allowing" pornographic films, the effect of question-wording disappeared.
In studying polls on abortion, Schuman and Presser found a difference of 15 per cent in two surveys asking 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?" However, in one survey was the question was placed 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." In this case, the support for abortion rights in the other question was 15 per cent lower than in the case when no such "introductory" question was asked (Hogan 1997, 169, 171).
References:
Albig, William. 1939. Public Opinion. New York: McGraw Hill. Albig, William. 1956. Modern Public Opinion. New York: McGraw Hill.
Allport, Floyd H. 1937. Toward a Science of Public Opinion. Public Opinion Quarterly 1, 1, 7-23. Beniger, James R. 1987. Toward an Old New Paradigm. The Half-Century Flirtation with Mass
Society. Public Opinion Quarterly 51, 4 - Part 2, S46-S66. Beniger, James R. 1992. The Impact of Polling on Public Opinion: Reconciling Foucault, Habermas,
and Bourdieu. International Journal of Public Opinion Research 4, 3, 204-219. Berger, Peter L. and Thomas Luckmann. 1969. Die gesellschaftliche Konstruktion der Wirklichkeit.
Frankfurt am Main: Fischer. Blumer, Herbert. 1948. Public Opinion and Public Opinion Polling. American Sociological Review 13, 542-554.
Bourdieu, Pierre. 1972/1979. Public Opinion Does Not Exist. In A. Mattelart and S. Siegelaub (eds.),
Communication and Class Struggle. Vol. 1. New York: International General. Bourdieu, Pierre. 1985. Remarques a propos de la valeure scientifique et des effets politiques des
enquetes d'opinion. Pouvoirs 33, 131-40. Crespi, Irving. 1989. Public Opinion, Polls, and Democracy. Boulder: Westview. Ginsberg, Benjamin. 1986. The Captive Public: How Mass Opinion Promotes State Power. New York: Basic Books.
Ginsberg, Benjamin. 1989. How Polling Transforms Public Opinion. In M. Marglois and G. A. Mauser
(eds.), Manipulating Public Opinion, 271-293. Pacific Grove, CA: Brooks/Cole. Glynn, Carroll J. and Jack M. McLeod. 1984. Implications of the Spiral of Silence Theory for
Is-.
no
Communication and Public Opinion Research. In K. R. Sanders, L. L. Kaid, and D. Nimmo (eds.), Political Communication 'Yearbook, 43-65. Carbondale: Southern Illinois University Press.
Held, David. 1989. Models of Democracy. Cambridge: Polity Press.
Hogan, Michael J. 1997. George Gallup and the Rhetoric of Scientific Democracy. Communication Monographs 64, 2, 161-179.
Keane, John. 1984. Public Life and Late Capitalism. Toward a Socialist Theory of Democracy. Cambridge: Cambridge University Press.
Key, V. O., Jr. 1961/1967. Public Opinion and American Democracy. New York: Alfred A. Knopf.
Lazareff, Alexandre. 1984. Le droit des sondages politiques. Paris: LDJ.
Michelat, Guy and Michel Simon. 1985. Les "sans réponse" aux questions politiques. Pouvoirs 33, 41-56.
Miller, Peter V. 1995. The Industry of Public Opinion. In T. L. Glasser and C. T. Salmon (eds.), Public Opinion and the Communication of Consent, 105-131. New York: Guilford.
Moore, David W. 1995. The Superpollsters. How They Measure and Manipulate Public Opinion in America. New York: Four Walls Eight Windows.
Noelle, Elisabeth. 1940. Amerikanische Massenbefragungen über Politik und Presse. Inaugural Dissertation. Berlin: Friedrich-Wilhelms-Universität.
Peer, Limor. 1992. The Practice of Opinion Polling as a Disciplinary Mechanism: A Foucauldian Perspective. International Journal of Public Opinion Research 4, 3, 230-242.
Salmon, Charles T. and Gerald D. Kline. 1984. The Spiral of Silence Ten Years Later: An Examination and Evaluation. In K. R. Sanders, L. L. Kaid and D. Nimmo (eds.), Political Communication Yearbook, 3-29. Carbondale: Southern Illinois University Press.
Splichal, Slavko. 1990. Statistična analiza besedil. (Statistical Text Analysis.) Ljubljana: FDV.
Tönnies, Ferdinand. 1922. Kritik der öffentlichen Meinung. Berlin: Julius Springer.
Tönnies, Ferdinand. 1923. Macht und Wert der Öffentlichen Meinung. Die Diouskuren: Jahrbuch für Geisteswissenschaften 2, 2, 72-99.
Wilson, Francis Graham. 1962. A Theory of Public Opinion. Chicago: Henry Regnery.
00 no