From Skopein to Scraping: Probability, Agency, and the Politics of Public Opinion Research

Introduction

Polling has become a ubiquitous activity in modern societies. In its early years, pioneers in the field, such as Elmo Roper (1900-1971), praised it as the ‘greatest contribution to democracy since the introduction of the secret ballot’.¹ George Gallup (1901-1984) saw it as a means to protect the common man or woman from the ‘tyranny of the majority’.² It was seen as a means to grant political agency to those voices that might not otherwise be heard. The point being made here has to do with the logic of survey sampling. The principle is that it must not depend on the individual and on whether they enter the sample, but that, for each person belonging to the population for which poll results are to be generalised, the probability of the individual entering a sample must be equal; if unequal, it must still be calculable and accounted for. Polling, it seems, was and is viewed by many as an endeavour that gives everyone in the populace an equal say in matters of public concern.

Drawing on work in the history and philosophy of statistics, political theory and sociology, this article further explores this claim regarding the relationship between probabilistic reasoning in survey statistics and political agency. At the core of the article lies an exploration of polling through the lens of the work of Jacques Rancière. Despite the understanding of public opinion research as a means to give equal voice and political agency to everybody, this article argues that polling removes the political sphere, which is needed for there to be political agency, for the appearance of politics. In substituting the always-partial representations of polls for the real, measured public opinion becomes identified with the actual body of the people and their opinions.

To arrive at this point, I will begin with a short elaboration of the notion of political representation and its role in modern democracies. I will explore how practices of gauging public opinion have been presented as a means to improve and enable democratic governance and decision-making, as it allows the constant assessment of the will of the people. I will argue, however, that rather than representing the will of the people, polling practices construct the people that they purport to represent. In a second step, I will show how the origin of the view of polling as a deliberating democratic force has its roots in the egalitarian presupposition of sampling, giving each member of a population an equal voice and possibility for political participation. In this way, opinions are estimated based on other people’s responses and create a representation of the people that is always visible and complete, overriding the sphere in which political subjectivities can appear. Thus, the probabilistic calculation of opinions does away with, to some extent, the promise of opinion research as a means to enable political participation and to grant equal agency. In the last step, I will explore new developments in the field of public opinion research, arguing that new methods of what has been termed ‘Demos Scraping’ reinforce this very observation.

Notes on Political Representation

Modern liberal democracies are characterised by a form of responsiveness, in the sense that the political leadership acts ‘in the interest of the represented, in a manner responsive to them’.³ For there to be a possibility to be responsive to a populace, a certain form of knowledge about the populace is necessary. Democracies thus develop practices of knowledge creation, statistical techniques of doubling and representing, that give the populace a politically intelligible form through which public attitudes, for example, become visible. As indicated in the introduction, studying public opinion has, from its earliest stages onwards, been seen as a way through which the responsiveness that political leaders should perform can be enabled. Polling pioneer Archibald Crossley, for instance, stated in 1937 that ‘[s]cientific polling makes it possible within two or three days at moderate expense for the entire nation to work hand in hand with its legislative representatives, on laws which affect our daily lives’. This, he states, ‘is the long-sought key to “Government by the people”’.⁴ These claims included the argument that ‘democracy’s auxiliary ballot box’, as Elmo Roper once termed it, is even more democratic than voting, since sampling allows the inclusion of the voices of those who do not vote.⁵

As it will become crucial for the overall argument of this essay, we need to distinguish between two logics of representing public opinion, which can be called Demoscopy and Demos Scraping. While the former refers to classical opinion polls, the latter refers to the more recent use of (big) data and web scraping tools to understand public opinion based on digital trace data. The neologism ‘Demoscopy’ was first introduced by Stuart Dodd in 1946 as ‘the observing of people by sampling’, combining demos, the Greek word for citizenry, and skopein, which means to view or observe.⁶ This notion has not found prominence in the Anglophone world but is widely used in European countries including Germany and France. Contrary to the observation of people by sampling, new means of data gathering aim to access the populace through the digital traces people leave for other purposes. Lena Ulbricht suggested calling this practice Demos Scraping, a combination of the term ‘web scraping’ and demos, describing the ‘practices of gaining information about citizens through automated analysis of digital trace data which are re-purposed for political means’.⁷ This form of representation also implies that, rather than ‘getting a deep understanding of their object’, those epistemic practices ‘scrape the surface’ of the constituency.⁸

Seminal work on the notion of representation understands it as an interaction in which the representant responds in a certain way to the represented. In her classic study on the concept of representation, political theorist Hanna Pitkin defines it as re-presentation, in the sense of ‘making present again’. Rather than merely bringing something into presence, representation is about the ‘making present in some sense of something which is nevertheless not present literally or in fact’.⁹ One subset of representation is that of standing for, within which descriptive representation, defined as ‘the making present of something absent by resemblance or reflection, as in a mirror or in art’, is the form we find in surveys and polls.¹⁰ A subset of a population stands for the whole; it holds up a mirror through which the population can observe itself. Within this model, Pitkin counts random sampling as the best approach, ‘no doubt because of the power of that technique in scientific research and because it is familiarly linked with representation in the idea of the representative sample’.¹¹

Following this model, the idea of representation constitutes a gap between how things really are and how they appear to be. There is a mirroring relation between the represented and representation, but the mirror must not be mistaken for the image that it produces. For instance, the fact that the 2021 UK Census asked its respondents for the first time about their sexual orientation and gender identity can be seen as a correction of a previous miscount that kept LGBT populations invisible.¹² As we shall see later when discussing Rancière, one can say that this constitutes an intervention into the ‘distribution of the sensible’.¹³ Representation thus involves a dual metaphysics of reality and appearance, within which one can appeal misrepresentations. Those occasions in which miscounts can be appealed are, however, rare, and statistical representation becomes more a form of presentation. As we shall see, the agency of groups to make their voice heard and to disrupt the regime of the perceptible is very limited.

When focusing on the process, the performative character of representations becomes visible. As Lisa Disch states, the core of Pitkin’s work actually lies in its processual character, stating that ‘[i]n politics, acts of representation do not simply reflect constituencies and their interests but help to bring them into being’.¹⁴ Disch arrives at this point in her remarks on what she terms the ‘constituency paradox’, by which she means that, if the interests of constituents develop not before but in the representative process, responsiveness to those interests cannot count as an indicator of good democratic representation. Acts of representation are ‘quasi-performative’, in that they bring into being the phenomena which they purport to describe. The statistical representation becomes the represented itself. Following Rancière, public opinion research is not only a science of opinion, but a ‘science immediately accomplished as opinion, a science that has no meaning except in terms of this process of specularization where an opinion sees itself in the mirror held up by science to reveal to it its identity with itself’.¹⁵ The practice of public opinion polling determines the consensus, not in assessing it but in producing it.

Samples, Surveys and Strata: On the Logic of Demoscopy

To acquire a better understanding of the logic of survey sampling, there are three important areas to address. The first concerns history, the second equality, and the third strata. I will begin with history. Demoscopy, or the Science of Public Opinion, emerged in the first half of the twentieth century; it is based on the principle of studying a sample from which to extrapolate to a population. Towards the end of the nineteenth and the early twentieth century sampling was suggested by the Norwegian statistician Anders Nicolai Kiær (1838-1919), and over the years a specific technique emerged from the various approaches that were tried. In the beginning, newspapers began to print questionnaires that readers could fill out and send back. The Literary Digest played an important role here, as, since their polls began in 1916, the magazine had always correctly predicted the winning American presidential candidate, and was thus highly praised and trusted. Although the procedures were at first not particularly sophisticated, predicting the outcome of presidential elections based on opinion polls was successful five times in total. However, the highly regarded publication failed spectacularly in 1936 when it incorrectly forecast Landon to win over Roosevelt. The cause of the Digest’s wrong forecast is considered to be coverage bias: respondents were recruited mostly from the phone book and a list of automobile owners, resulting in a sample skewed towards wealthy people.

At the same time, George Gallup drew a much smaller (quota) sample and accurately predicted Roosevelt's victory. In the aftermath, Gallup’s forecasts gained trust and authority. Twelve years after his brilliant prediction, however, a major polling disaster happened to Gallup. In the course of the 1948 American presidential election, Gallup inaccurately predicted a victory for Dewey over Truman. Gallup’s election prediction enjoyed so much confidence at the time that the Chicago Tribune erroneously headlined with ‘Dewey Defeats Truman’ the day after the election, and the German Newspaper Münchner Merkur ran the headline ‘Thomas E. Dewey America's New President’. The reason for the misprediction was that the interviewers were allowed to choose who to interview, given certain quotas. In each of the fixed categories (including gender, age, and economic status) Republicans were apparently easier to reach. One of the results of these investigations was a critique of quota sampling, and Gallup subsequently began to use random sampling as the basis for his polls. This technique had been discussed among survey researchers since the early twentieth century, but at this point a consensus on its superiority was established and random sampling became the characteristic approach of the polling industry.

With regards to equality, survey and poll randomisation is based on a form of egality, in the sense that, as noted, every person belonging to the universe for which the results are to be generalised must have an equal or calculable probability of entering the sample. I would like to exemplify this notion by drawing on Sidney Verba’s 1995 presidential address before the American Political Science Association. Verba contrasts surveys with other forms of political participation, such as voting, protests, letter writing, or campaigning and states that while those ‘ordinary modes of citizen activity […] allow quiescence […] [,] [s]urveys do not let people be quiescent; they chase them down and ask them questions’.¹⁶ Surveys thus bring social science technology, the political theory of representation and real political issues together. They allow for equal responsiveness by governing elites towards the citizenry in the sense that this responsiveness requires ‘the capacity to provide equal consideration’ and ‘equal information about the needs and preferences of all citizens’; it also means that ‘citizens have to supply that information’, as ‘if some citizens are invisible, one cannot respond to them’.¹⁷

Verba acknowledges the inevitability that in any given constituency there will be active people and quiescent people; however, he points out that it is important to differentiate between the reasons, whether it is a result of preference or constraint. In this regard, the epistemology of survey sampling has a special character, as participation in a survey does not hinge on whether people have the resources or the motivation to do so: everybody participates in that their responses are calculated based on other responses. Despite polls never being perfectly representative, Verba argues that they, nevertheless, offer a better cross-section of the populace than almost all other modes of citizen activity. They achieve a ‘relatively unbiased view of the public by combining science and representativeness, indeed, by achieving representativeness through science’.¹⁸ Without surveys, political leaders would still ‘sway with the wind of opinion’; ‘[t]he wind would just blow from different quarters, more likely from the better parts of town’.¹⁹ To Verba, ‘[s]urveys produce just what democracy is supposed to produce – equal representation of all citizens’.²⁰ In his view, the ‘sample survey is rigorously egalitarian; it is designed so that each citizen has an equal chance to participate and an equal voice when participating’.²¹ On this view, polls become a crucial instrument for democracy as they allow for equal representation and responsiveness; everybody, it seems, is equal before the polls.

This egalitarian view has also been challenged. Herbert Blumer, for instance, stated in a 1948 article that ‘polling gives an inaccurate and unrealistic picture of public opinion because of the failure to catch opinions as they are organised and as they operate in a functioning society’.²² In the case of polling for election forecasting, ‘a ballot cast by one individual has exactly the same weight as a ballot by another individual’, which means that voters are a population in which each individual ‘has equal weight’.²³ However, this changes when moving from election forecasting to issue polling. The formation and expression of public opinion is not simply ‘an action of a population of disparate individuals having equal weight’ but is shaped by the structure of society and the various groups and individuals within it.²⁴ These groups and individuals hold different levels of influence and occupy different strategic positions of power, which influence the formation and expression of public opinion. Treating everyone as equal thus does not take into consideration the actual power dynamics and social factors in the real world. This argument has also been echoed by Pierre Bourdieu, according to whom society consists of fields, where people have different social, cultural, or economic capital; this is also why people have different backgrounds against which they form opinions. Bourdieu states that pollsters and survey researchers cannot expect everyone to have an equally well-informed opinion, or an opinion at all, just as polling presupposes that everyone has the same or a known chance to enter a sample; it also presupposes ‘that the production of an opinion is within everyone’s range of possibility’.²⁵ Without considering the power of those who hold an opinion, it is not a strength of polls to treat all opinions equally. Distinct from election forecasts, the reality of the world makes it impossible to assume that a given opinion means the same for every participant. Public opinion, as per Bourdieu’s conclusion, ‘does not exist in the form which some people, whose existence depends on this illusion, would have us believe’.²⁶ Notwithstanding these and similar arguments, polling became the dominant instrument for creating representations of people and their opinions, rendering public opinion the result of what public opinion polls measure.

The application of random sampling does not, however, mean that if you contact people at random you will actually obtain a random sample. In the words of pollster Nate Silver, the ‘dirty little secret about polling’ is ‘that if you randomly call people on the phone, you will not get a truly random sample’.²⁷ This, as Mark Pack states, is because some people are more likely to respond; women, for instance, are more likely to answer the phone than men, and the elderly are more likely than young people.²⁸ Those circumstances lead survey researchers and pollsters to the practice of stratification. The idea behind stratification is to subdivide a heterogenous population into separate and homogeneous units, called strata. While pre-stratification means that strata are created before the sampling begins, post-stratification describes the creation of strata during the sampling or after it has taken place.

Stratification, in order to create homogeneity, raises new questions, as, in each stratum, individuals with similar demographics are being put together to form a virtual community. Matthew Hannah argues that stratification ‘reduces individuals to collections of their publicly recorded attributes and lumps them together in new groupings without their consent’.²⁹ Those virtual communities have to do with ‘formal similarity’ and not with ‘any substantive connection or affinity’.³⁰ When survey sampling means that there is an equal or known probability to enter the sample, it also means that the opinions of those who do not enter are calculated based on others. Those who constitute the sample and respond thus have a ‘disproportionate say’ in determining how the opinion of those who do not respond will be estimated based on their neighbours in the virtual community.³¹ As in Michel Foucault’s description of the prison, here too ‘[e]ach individual, in his [sic] place, is securely confined to a cell from which he is seen from the front by the supervisor; but the side walls prevent him from coming into contact with his companions.’³² With Rancière, we can say that

[t]he science of simulations of opinion is the perfect realization of the empty virtue Plato called sôphrosunê: the fact of each person’s being in their place, going about their own business there, and having the opinion identical to the fact of being in that place and doing only what there is to do there.³³

In this sense, probability and agency are closely linked. In assigning each individual an equal or known probability of entering a sample, pollsters can in fact impose an opinion on these individuals.

The Disappearance of Politics

In his oeuvre Rancière refers to the political difference, in which he distinguishes between politics and police. The difference is in the mode of appearance. Whether a protest expresses politics depends on whether or not it is able to appear in the current regime of the perceptible: ‘nothing is political in itself. But anything may become political if it gives rise to a meeting of these two logics’.³⁴ To exemplify this distinction, Rancière uses the example of a protest that is removed from perception, because the police say: ‘[m]ove along! There is nothing to see here!’³⁵ Politics is a space for the appearance of the subject, agency, and for reconfiguring the space of the perceptible. Politics only happens in cases where those ‘distributions of the sensible’ are interrupted, thus when ‘moving along’ and not taking part in the protest is shaken.³⁶ Politics is ‘first and foremost an intervention upon the visible and the sayable’,³⁷ and thus an intervention into the order set up by the police, understood as the implicit law that governs the visible. Furthermore, democracy is described as the ‘kind of community that is defined by the existence of a specific sphere of appearance of the people’.³⁸ The introduction of certain groups or interests into the political sphere is the means by which those groups and interests are constituted. In contrast, postdemocracy is the ‘government practice and conceptual legitimization of a democracy after the demos, a democracy that has eliminated the appearance, miscount, and dispute of the people’.³⁹

In postdemocracy, the dualism between how things are and how they appear is replaced by a monist metaphysics in which there is ‘no interval between law and fact’.⁴⁰ This shift from representation to presentation follows a change in the technologies of governance. Technologies of datafication transform ‘all aspects of life and turn them into data’ through which ‘we can transform their purpose and turn the information into new forms of value’.⁴¹ The difference is that while words and linguistic expressions can always correspond to various meanings, data appears to exactly correspond to what it expresses; it is the ‘conjunction of science and the media’.⁴² In this sense, the people become identical with the sum of its parts, of which the ‘count is always even and with nothing left over’.⁴³ Echoing the notion of stratification, Rancière states that public opinion polling conducts ‘scientific modelling and forecasting operating on an empirical population carved up exactly into its parts’, and, further, that the ‘effectiveness of the sovereign people is exercised as strictly identical to the calculations of a science of the population’s opinions, which is the same as saying an immediate unity of science and opinion’.⁴⁴ The appearance of the people, the political moment in which agency is exercised, becomes overridden by calculations:

As a regime of opinion, the principle of postdemocracy is to make the troubled and troubling appearance of the people and its always false count disappear behind procedures exhaustively presenting the people and its parts and bringing the count of those parts in line with the image of the whole. The utopia of postdemocracy is that of an uninterrupted count that presents the total of ‘public opinion’ as identical to the body of the people.⁴⁵

In a similar vein, Ernesto Laclau understands the process of representation to include a situation where ‘somebody is present in a place from which he or she is materially absent’, resonating with Pitkin’s use of the term.⁴⁶ Representation always has it that somebody ‘substitutes for’ and ‘embodies’ the represented. For Laclau, the conditions for perfect representation, when ‘the act of representation is totally transparent’ and the ‘opaqueness inherent in any substitution and embodiment’ is reduced to a minimum, ‘never actually obtain’.⁴⁷ This is due to the inherent logic in the process of representation: if somebody or something needs to be represented, its basic identity is taken from one place to another, through which the identity of the represented remains incomplete. An interest that is being represented by, for example, a deputy, always undergoes transformation, as ‘the representative inscribes an interest in a complex reality different from that in which the interest was originally formulated and, in doing so, he or she constructs and transforms that interest’.⁴⁸ There can thus be no pure representation, there is always an ‘opaqueness, an essential impurity in the process of representation, which is at the same time its condition of both possibility and impossibility’.⁴⁹ If the representation could be fully reduced to the represented, it would no longer involve any representation. The representation of public opinion thus means that, in the process of representation, something is created that only exists by being represented. In this respect, opinion polls will always produce partial and opaque representations, although endowed with the illusion of capturing the whole.

The ubiquity of public opinion polls creates a sense in which a measurement is identified with the actual body of the people and with their opinions. This identification is the removal of the very sphere in which the people can appear. Polling replaces the struggle of who and what enters the visible, inherent in politics, with continuous measures and numbers, corresponding to the logic of the police. Rancière describes this identification as such:

What in actual fact is this identification of democratic opinion with the system of polls and simulations? It is the absolute removal of the sphere of appearance of the people. In it the community is continually presented to itself. In it the people are never again uneven, uncountable, or unpresentable. They are always both totally present and totally absent at once. They are entirely caught in a structure of the visible where everything is on show and where there is thus no longer any place for appearance.⁵⁰

In this sense, the regime of public opinion, ‘as gauged by the poll and of the unending exhibition of the real’ is, for Rancière, the ‘normal form the police in western societies takes’.⁵¹ Polling data forecloses the possibility for minorities to challenge their exclusions, since the data is no longer a doubling of the facts of the world but an alleged presentation of the facts of the world. This becomes even more troubling with the privatisation and opacity of the very methods to gauge public opinion. In this sense, the indeterminacy of how numbers and figures are brought about constitutes a form of overdetermination, as those numbers are stipulated as a representation of the whole. But resistance does exist, as can be seen from the Statactivism movement, for example, which is a French movement that mobilises statistical data to support social and political movements. Such endeavours can be understood as a way to interrupt the police and re-distribute the sensible. With the slogan ‘another number is possible’, statactivists denounce a certain state of reality and use statistics to create more equivalent representations of reality; their aim is to undo what a hegemonic logic of quantification has established.⁵²

On the Power of Platforms: Public Opinion for a Democracy Without Politics

In this last section, I will briefly focus on new methods of data gathering that exist alongside traditional surveys. In their 2014 report on ‘Social Media in Public Opinion Research’, the American Association for Public Opinion Research (AAPOR) opened with the observation that public opinion research is entering an era in which traditional surveys may play a less important role, referring to increasing difficulties of executing and maintaining traditional surveys and polls. The rise of new technologies changes the landscape in which public opinion researchers operate:

As these technologies expand, so does access to users’ thoughts, feelings, and actions expressed instantaneously, organically, and often publicly across the platforms they use. […] The ubiquity of social media and the opinions users express on social media provide researchers with new data-collection tools and alternative sources of qualitative and quantitative information to augment or, in some cases, provide alternatives to more traditional data-collection methods.⁵³

Resonating with this statement, Susan Herbst argues that with the advent of the Internet, public opinion has become much more aligned with ‘talk and texture, not with quantitative data’.⁵⁴ It is the linguistic behaviour of Internet users themselves that serves as a source for the study of public opinion, not their responses to questionnaires. Thus, while polls understand public opinion as an aggregation of individual opinions, under the new conditions of blogs and social media, ‘textured talk, dialogue, exchange, and conversation – not numbers – are the content of public opinion’.⁵⁵

One of the central differences between these modes of public opinion research is that while in the earlier form data was gathered with the aim of understanding public opinion, in the new form data is a surplus of other activities, repurposed to access public opinion. This data is organic, as if came from the users’ natural, everyday behaviour rather than from a constructed survey situation. Demos Scraping can here be seen as distinct from Demoscopy, as introduced earlier in this essay, but it comes with a similar enthusiasm and promises. As Lena Ulbricht has shown, defenders of those approaches claim that these new data ‘yield unprecedented insights into populations for policy makers’.⁵⁶ In doing so, policy makers are using a similar rhetoric to the defenders of early polling: Twitter data, for instance, is claimed to enable researchers to take ‘the pulse of the nation’, which is very similar to the proclamations of pollsters George Gallup and Saul Rae, whose book was entitled The Pulse of Democracy: Public Opinion and How It Works.⁵⁷

Demos Scraping is supposed to create a snapshot of the population on a continuous basis, and thus enable the continuous monitoring of the changing political mood within a population. It can thus be understood as a statistical representation of all sorts of areas of social life in real time. It is presented as having many advantages compared to traditional polls. As summarised by Lena Ulbricht, its defenders argue that ‘despite its multiple biases, it is a way of knowing the demos that surpasses the insights gained from traditional disciplinary data about citizens’.⁵⁸ For instance, opinion mining uses natural language processing to determine the public's mood or attitude towards a particular topic. It is usually applied to text data and allows one to determine whether the data is positive, negative or neutral.

In a similar way to how polling has been understood as a form of political representation, so too can Demos Scraping. It takes place, however, under the new conditions of multi-national corporations, which means that the emergence of public opinion is more and more tied to the possibilities and configurations provided by those platforms. Facebook’s News Feed, for instance, is an example of how public opinion is increasingly shaped through algorithmic decisions based on commodified and privatised datasets. In his seminal work on the power of platforms, Michael Seeman defines them as ‘structured spaces of possibility’.⁵⁹ Platforms anticipate and direct certain behaviour based on a private and standardised logic. As hegemonic platforms invariably shape the way communication unfolds, public opinion increasingly becomes a function of social media platforms or, to be more precise, a function of the possibilities afforded by those platforms. In other words, social media platforms determine the space of possibility in which political participation appears according to the platform’s logic. In using individual advertising, filtering and managing news and posts, platforms shape and structure the possibilities for political agency.

The main implication of this is related to the idea of political participation inherent in the practice of Demos Scraping, as this practice changes the context in which democratic and political participation unfolds. Lena Ulbricht points out that if data analysts see online behaviour ‘as an expression of political preferences and intentions, they radically re-interpret political participation’, overlooking the fact that ‘the observed behaviour was not intended to be decidedly political’, and thus that ‘it is by definition not a form of political participation’.⁶⁰ In this sense, this practice brings about what Antoinette Rouvroy has termed ‘data-behaviourism’, a ‘new way of producing knowledge about future preferences, attitudes, behaviours, or events without considering the subject’s psychological motivation, speeches or narratives but rather relying on data’.⁶¹ It constructs, as Lena Ulbricht argues, politically-passive citizens as political participants without their awareness. When the quantification of the linguistic behaviour of people in areas devoid of politics becomes the locus of their political participation, one can no longer speak of them as exercising political agency. In line with what has been discussed throughout the article, this practice thus reinforces the removal of the sphere in which politics and political agency can appear. It heralds a democracy that is devoid of politics.

Conclusion

After exploring the notion of political representation and tracing some of the crucial historical moments in the development of polling, I have shown how the view of polling as a deliberating democratic force has its roots in the egalitarian presupposition of (random) sampling. When sampling a representative subset of a population, everybody, even those who were not questioned, seem to be given an equal say in political matters and thus given political agency. George Gallup’s statement that ‘in many situations – particularly those in which a substantial portion of the population fails to take the trouble to vote – the poll results might be even more accurate as a measure of public sentiment than the official returns’ is a case in point here; since not everyone goes to the ballot, it is argued that polling is an even more accurate reflection of public opinion than an election.⁶² Throughout this essay, I have scrutinised this view that presents Demoscopy, the practice of observing the population through sampling, as an ultimately democratic tool that grants equal political agency to all individuals. In contrast, I have argued that caution must be exercised when considering it as a means of enabling equal political agency and participation; the calculation and presentation of the probability of holding a certain opinion must not be mistaken for the struggles often involved in making one’s voice heard.

In drawing on, alongside others, the work of Jacques Rancière, I have emphasised that for there to be political agency, more is needed than the mere assessment and calculation of preferences; what polling lacks is a moment in which political opinions and positions can appear and interrupt the distribution of the sensible. This is aggravated by the practice of Demos Scraping, the gauging of public opinion through digital trace data, in which data, often generated for non-political purposes, becomes identified with political participation. It suppresses the agency of the populace by reducing political participation to behaviour in areas devoid of political subjectivities. In both modes of assessing public opinion, Demoscopy and Demos Scraping, the people is, in Rancière’s words ‘caught in a structure of the visible where everything is on show and where there is thus no longer any place for appearance’.⁶³ It is in this sense that techniques to measure public opinion turn the probabilities of holding certain opinions into necessities, imposing on individuals opinions calculated by statistical techniques.

In considering the importance of knowing and understanding the views and opinions of the public in a democracy, it is crucial to approach practices that seek to do so with caution, and to not mistake these opinions for political participation in decision-making processes. Polling results should thus not be given undue weight in the construction of the legitimacy of political measures or policies. Instead, they should be considered as one factor among many in decision-making processes. This theme becomes particularly important when it comes to a lack of transparency on the side of those who produce estimates of public opinion. The oftentimes opaquely generated numbers about public opinion are not seldomly treated as a true representation of the actual body of the people, which feeds back into policy making or governmental programmes. In this sense, statistical statements about public opinion can be used to manufacture legitimacy and to replace democratic forms of political participation. It is in this sense that the scientification of polls can give possible political actions the character of a mathematical necessity.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Notes on contributors

Lukas Griessl

Lukas Griessl is a PhD candidate in Sociology at the University of Essex. Before coming to Essex, he studied at the CU Eichstätt-Ingolstadt, the University of Exeter and the London School of Economics and Political Science. His current research interest is in the history and sociology of empirical social research and polling, which is also the broader theme of his doctoral dissertation. In his research, he draws on work in political theory, science and technology studies and recent work in ignorance studies. Email: [email protected]

Notes

1 Igo, The Averaged American, 121.

2 Gallup and Rae, The Pulse of Democracy, 268.

3 Pitkin, The Concept of Representation, 209.

4 Crossley, “Straw Polls in 1936,” 35.

5 Roper and Woodward, “Democracy’s Auxiliary Ballot Box.”

6 Dodd, “Toward World Surveying”, 473.

7 Ulbricht, “Scraping the Demos,” 427.

8 Ibid., 427–8.

9 Pitkin, The Concept of Representation, 8–9.

10 Ibid., 11.

11 Ibid., 73.

12 Sherwood, “Census to Ask about Sexual Orientation for the First Time.”

13 Rancière, The Politics of Aesthetics.

14 Disch, “Democratic Representation and the Constituency Paradox,” 600.

15 Rancière, Disagreement, 105.

16 Verba, “The Citizen as Respondent,” 4.

17 Ibid., 1–2.

18 Ibid., 4.

19 Ibid., 6.

20 Ibid.

21 Ibid.

22 Blumer, “Public Opinion and Public Opinion Polling,” 547.

23 Ibid.

24 Ibid.

25 Bourdieu, “Public Opinion Does Not Exist,” 124.

26 Ibid., 129.

27 Pack, Polling Unpacked, 48.

28 Ibid.

29 Hannah, “Sampling and the Politics of Representation in US Census 2000,” 526.

30 Ibid.

31 Ibid.

32 Foucault, Discipline and Punish, 200.

33 Rancière, Disagreement, 106.

34 Ibid., 32.

35 Rancière, Panagia, and Bowlby, “Ten Theses on Politics,” 22.

36 Rancière, The Politics of Aesthetics.

37 Rancière, Panagia, and Bowlby, “Ten Theses on Politics,” 21.

38 Rancière, Disagreement, 99.

39 Ibid., 102.

40 Ibid., 112.

41 Cukier and Mayer-Schoenberger, “The Rise of Big Data,” 28.

42 Rancière, Disagreement, 105.

43 Ibid., 105.

44 Ibid., 103.

45 Ibid.

46 Laclau, Emancipation(s), 97.

47 Ibid.

48 Ibid., 98.

49 Ibid.

50 Rancière, Disagreement, 103.

51 Ibid., 31.

52 Bruno, Didier, and Prévieux, Statactivisme.

53 Murphy et al., “Social Media in Public Opinion Research,” 789.

54 Herbst, “(Un)Numbered Voices?,” 94.

55 Ibid., 95.

56 Ulbricht, “Scraping the Demos,” 429.

57 Mislove, “Pulse of the Nation.”

58 Ulbricht, “Scraping the Demos,” 429.

59 Seemann, Die Macht der Plattformen, 31.

60 Ulbricht, “Scraping the Demos,” 432.

61 Rouvroy, “The End(s) of Critique,” 143.

62 Gallup, “The Role of the Public Opinion Poll,” 21.

63 Rancière, Disagreement, 103.