Accueil du site > Essays > Maps and opinion polls

Maps and opinion polls

par Joël Gombin & Jean Rivière & traduit par Oliver Waine, on 6 juin 2012

Toutes les versions de cet article : [English] [français]

Presidential elections – a crucial moment in the democratic life of France – generate all sorts of analyses that seek to interpret and decipher voting behaviour. Joël Gombin and Rivière highlight the limits of opinion polls and the risks of monocausal geographical analyses, instead advocating an innovative approach that is able to situate voters’ choices within their territorial contexts.

As with previous ballots, the 2012 presidential election proved to be the perfect opportunity for public opinion research bodies to produce abundant amounts of material on electoral trends of a scientific, semi-scientific and general nature. Maps and opinion polls – or, to put it another way, space-oriented and individual-oriented approaches – have for a long time been the essential tools of those whose job it is to try to decipher the polls on a national scale. Accordingly, the field of electoral studies in France suffers from the fact certain types of data are used as a matter of routine, partly owing to divisions between disciplines. On the one hand (among geographers), the most frequent approach consists of producing maps on the basis of electoral results aggregated at different scales (département, canton, municipality, etc.), and then trying to get these maps to convey a message, sometimes at the risk of overinterpreting and oversimplifying different types of residential space (“urban cores”, “suburbs”, “periurban areas”, “rural areas”, etc.). [1] On the other hand (among political scientists and televisual electoral sociologists), the dominant approach is based on opinion polls of voting intentions, despite the limitations of this technique, which are well known following Pierre Bourdieu’s inaugural criticism in “L’opinion publique n’existe pas” (Bourdieu 1973). [2]

After a long period during which opinion polls dominated the field of electoral studies, a number of approaches that, without reifying space, are attentive to the geographical and social contexts in which individuals live have been developed in recent years, thus making it possible to obtain greater realism in the comprehension of electoral behaviours by resituating them in the environment where they occur.

Opinion polls : from their initial contribution to their current limits

Ever since George Gallup correctly predicted the re-election of President Franklin D. Roosevelt in 1936, opinion polls have gradually established themselves as the method par excellence for obtaining information about voting patterns, for both scientific and non-scientific purposes. In keeping with behaviourism, [3] which dominated social psychology – and indeed the social sciences in general – after the Second World War, the individual became the preferred unit of analysis. According to this theory, it is the individual’s specific characteristics that explain electoral behaviour. Although expensive, the technology implemented was initially highly productive from a scientific standpoint. In a context where the social sciences, particularly in France, were marked by the Marxist paradigm, opinion polls made it possible to obtain sociological regularities : the so-called “heavy variables”. Social position and religious practices were, at the time, seen as the key determinants of electoral behaviour and, as Paul Lazarsfeld puts it, “a person thinks, politically, as he is, socially”. In France, the classic work by Guy Michelat and Michel Simon, Classe, religion et comportement politique (in English, Religion, class and politics, 1977), represents the apogee of this period by showing the existence of two symbolic systems that dominate the French political system : on the one hand, Catholics tend to vote significantly more to the right than average ; on the other, manual workers, often de-Christianised, generally vote to the left, and communist in particular.

As with any scientific technology, however, the opinion poll has “decreasing cognitive yields” (Lehingue 2007), for several reasons. First, the generally accepted link between this method and the theoretical framework of rational choice – the idea that homo politicus, like homo economicus, acts according to cold cost-benefit calculations – tends to remove individuals from the social contexts in which they actually move. The well-known “paradox of voting” is a good example of the postulates of the theory of rational choice being called into question : if the electorate were “rational”, they would not vote at all, since voting involves costs that far exceed the benefits, as the probability that one person’s vote will change the outcome of the election is infinitesimal (Boudon 1997). Secondly, pollsters encounter increasing difficulty in finding respondents for their surveys, which means that this tool gradually loses its scientific interest, as its representativeness is less and less reliable. In parallel, social changes such as the destructuring of workers’ collectives or the expansion of the “employees” category (i.e. office workers) were bringing a significant limitation of opinion polls to the fore, namely that the behaviour of a socio-economic category can only be more or less accurately predicted if the category in question corresponds to a real and relatively homogeneous social group – in other words, if the statistical category matches the social reality on the ground. However, this was clearly less and less the case for the “aggregated socio-professional categories” in France (a six-category system of socio-economic classes [4]). This may explain the smokescreen effect whereby the “heavy variables” (and in particular the six statistical socio-economic classes used) would appear to no longer take account of voting habits – a common finding of electoral studies in the last 20 years. The individual described by opinion polls increasingly resembles an abstract intellectual construction : an individual free of any social attachments and any socialisation environment, as if weightless.

And yet, as the scientific interest of opinion polls has decreased, their usage in the media has become more widespread : despite the significant “corrections” to which they are subjected, backed up with figures presented by established experts, they are supposed to predict who is going to win for the benefit of the “general public” (Lehingue 2007 ; Garrigou 2006 ; Hubé and Rivière 2008). These considerable limitations explain why, from the 1970s and 1980s onwards, French political scientists have tried to explore other avenues, stressing that opinion-poll procedures “have the disadvantage of ignoring geography and, to a certain extent, also ignoring history” (Michelat and Simon 1977) and that “the individual ‘captured’ by the sample survey is nothing more than a statistical abstraction” (Dogan and Derivry 1986).

Avoiding the trap of “spatialism”

This diminishing merits of opinion polls and the increasing accessibility of mapping software have led to a resurgence of studies that use spatialised data as their principal empirical basis (e.g. election results, public statistics such as census data from the French statistics office, INSEE). Consequently, since the end of the 1960s, many studies have been produced – initially in English-speaking countries – that take the spatial aspects of vote production into consideration. In France, such research has become very popular in the last 10 years or so, no doubt encouraged as much by the strong geographical trends associated with the far-right FN (National Front) vote as by Jean-Marie Le Pen’s propulsion to the second round of the 2002 presidential election – and, in particular, the opinion polls’ spectacular failure to predict this outcome on the now (in)famous date of 21 April 2002.

Current research draws on a rich tradition of political geography, initiated well before the appearance of opinion polls by André Siegfried and his famous Tableau politique de la France de l’Ouest (“Political Portrait of Western France”, 1913). In this groundbreaking work, Siegfried notes a strong relationship between voting tendencies and the nature of the soil, often summarised as follows : “granite votes to the right and limestone votes to the left”. Beyond this simplistic formula, it in fact highlights a system of social determinations that was highly innovative at the time : the geological substrate (granite or limestone) has an impact on the form of settlement (concentrated or dispersed), which, combined with other sociological aspects (type of land tenure, influence of the Catholic church, influence of the nobility), produced local social configurations that helped to explain regional electoral orientations.

This recent “rediscovery” of spatial aspects is not, however, without its pitfalls – which certain authors have already observed in Siegfried’s work (Veitl 1995) – and more specifically the temptation to believe that spatial factors alone can be used to explain voting trends. The spectre of “spatialism” lurks in the background, as can be seen in studies that seek to isolate a “spatial factor” in a given electoral dynamic. This is the case, for example, with regard to distance from the city, sometimes accorded great importance and deemed a determining factor, whereas, of course, it is far from being the only element to influence voting patterns (Bussi, Colange and Rivière 2011). In other words, spaces tend to be confused with the social processes that occur within them. This is especially problematic when certain media outlets, always looking for new angles, seize upon this type of interpretation and put explanations relating to distance from the city on the same footing as explanations relating to inhabitants’ social characteristics. [5] Here, we touch upon the fundamental ambiguity of the cartographic approach : is it a tool whose aim is to highlight (or even measure) “spatial effects”, that appear to be governed by a specific logic ; or is it a means of gaining a better understanding of social processes by resituating them in the contexts where they occur ? The answer, far from being unequivocal, clearly depends on researchers’ epistemological and theoretical options.

Accordingly, approaches where spatial aspects are presented as the only explanatory factors for voting patterns seem to us at best tautological, and at worst deeply anti-sociological. Indeed, on the contrary, methods that take account of the spatial dimension of social dynamics should ultimately be used to achieve increased realism by considering the production of votes in their contexts, which in turn makes it possible to propose other explanatory systems based on variable geography. Certain recent work has showed, for instance, that the electoral behaviour of the farming population varies considerably in spatial terms according to localised social configurations (Gombin and Mayance 2010).

Considering votes within their context

Obviously, much work still remains to be done in order to fully understand contemporary electoral behaviour. However, we can highlight the development of two directions of research that, together, help us to gain a clearer understanding of electoral dynamics, and which both seek to incorporate these dynamics into the system of social inequalities and embed them in the geographical contexts where they take shape.

The first is linked to the recent reinforcement of contextual approaches in France (Braconnier 2010). By associating different kinds of empirical materials (e.g. election results and national census data from INSEE, to characterise the spaces studied, together with voter signature lists, exit-poll questionnaires, interviews and localised observations), these approaches mean that the processes under way in various types of residential contexts can be better understood. As a result, it has, for example, been possible to establish that, in working-class neighbourhoods composed of large social housing estates, electoral registration rates and voter turnout rates vary considerably, depending on residents’ degree of integration within social groups, starting with the family, which always forms the principal framework for political socialisation (Braconnier and Dormagen 2007). In a very different context, the monitoring of the municipal election campaigns of 2008 in the central arrondissements (city districts) of Paris revealed the various mobilisation strategies of parties, candidates and militants on the ground (Agrikoliansky, Heurtaux and Le Grignou (dir.) 2011 [6]). Similarly, the historical analysis of stages of settlement and the exploration of the way in which the transformations of the sociological composition of “lower middle-class” housing estates are perceived make it possible to shed light on the electoral choices of their residents (Cartier et al. 2008, 2012). In periurban towns, all too often considered homogeneous entities, statistical comparisons between forms of social segregation, residential careers and the geography of votes helps to establish a connection with the social embedding of voter preferences (Rivière 2011, 2012). Finally, in working-class rural contexts, it is by analysing the reorganisation of worker sociabilities and the destructuring of traditional groups that we are better able to understand why certain social categories are tending to vote for the far right (Pierru and Vignon 2006). The comparison and cumulation of the results of all these studies will no doubt help, in years to come, to make advances in the analysis of electoral dynamics and draw conclusions whose scope extends beyond that of individual research areas.

The second direction lies in the use of innovative quantitative techniques, at least in French electoral sociology, such as multilevel modelling (Jadot 2002 ; Gombin and Mayance 2010). This type of modelling makes it possible, to a certain extent, to combine maps and opinion polls, by considering individual data together with the characteristics of these individuals’ places of residence (political profile of the place of residence, unemployment rate for the local employment area, etc.), so that it is possible, for instance, to compare the voting patterns of manual workers living in working-class neighbourhoods with those of manual workers living in more middle-class neighbourhoods. Indeed, 50 years ago, Klatzmann (1957) had already shown that manual workers tended to vote further to the left in districts where they were more numerous. However, the relevance of such an approach largely depends on the quality of materials available at the outset – not just the opinion-poll data, but also the scale at which the communities where individuals live are considered. Until now, however, this information has only been available at municipal level, at best, which presents a significant problem if we wish to have a detailed understanding of the electoral dynamics of urban settings, where the vast majority of electors live. But this deficiency is gradually being eliminated, thanks to the Cartelec research programme, which brings together geographers and political scientists, and which aims to create a vast database compiling electoral results and public statistics (INSEE census data, indicators used by the family benefits office concerning economic insecurity, and household income data from the tax office) at the scale of polling districts for all major French cities. The use of these data sets makes it possible to establish strong relationships between social segregation and city-dwellers’ electoral choices on a very precise scale. [7] Here too, the cumulative contributions of these quantitative methods at a highly detailed scale should enable important progress to be made in the field of electoral analysis.

The contributions made by these recent works, outlined briefly here, must be evaluated in greater detail, especially as this is a field that is evolving rapidly. But it already seems that the two complementary directions of research mentioned above make it possible to achieve a degree of sociological realism that is in stark contrast with the figure of the desocialised individual without attachments found in opinion polls, or with the cold, calculating, rational voter of econometric voting models, as well as with the traps of spatialism presented by certain fields of electoral geography. Voting choices constitute a social practice which – like any practice – responds to determinations that only make sense within the context of a system of relations between individuals and social groups that must be considered in spatial terms if it is to fulfil its explanatory potential.


  • Agrikoliansky, E., Heurtaux, J. and Le Grignou, B. (dir.). 2011. Paris en campagne. Les élections municipales de mars 2008 dans deux arrondissements parisiens, Bellecombe-en-Bauges : Éditions du Croquant.
  • Bargel, L. 2012. “Les municipales à Paris en 2008. Une approche contextualisée du processus électoral”, Métropolitiques, 6 April.
  • Bondon, R. 1997. “Le ‘paradoxe du vote’ et la théorie de la rationalité”, Revue française de sociologie, vol. 38, no. 2, pp. 217–227.
  • Bourdieu, P. 1973. “L’opinion publique n’existe pas”, Les temps modernes, vol. 318, pp. 1292–1309.
  • Braconnier, C. and Dormagen, J.-Y. 2007. La démocratie de l’abstention : aux origines de la démobilisation en milieu populaire, Paris : Gallimard.
  • Braconnier, C. 2010. Une autre sociologie du vote : les électeurs dans leurs contextes. Bilan critique et perspectives, Paris : Lextenso/Laboratoire d’études juridiques et politiques (LEJEP).
  • Bussi, M., Colange, C. and Rivière, J. 2011. “Distance(s) à la ville et comportements électoraux”, in Mattei and Pumain (dir.), Données urbaines, vol. 6, Paris : Anthropos.
  • Cartier, M., Coutant, I., Masclet, O. and Siblot, Y. 2008. La France des « petits-moyens » : enquête sur la banlieue pavillonnaire, Paris : La Découverte.
  • Cartier, M., Coutant, I., Masclet, O. and Siblot, Y. 2012. “Les petits pavillonnaires : perdus pour la gauche ?”, Métropolitiques, 21 March.
  • Derivry, D. and Dogan, M. 1986. “Religion, classe et politique en France. Six types de relations causales”, Revue française de science politique, vol. 36, no. 2, pp. 157–181.
  • Gombin, J. and Mayance, P. 2010. “Tous conservateurs ? Analyse écologique du vote de la population agricole lors de l’élection présidentielle de 2007”, in Mayer, Muller, Hervieu, Rémy and Purseigle (dir.), Les mondes agricoles en politique, Paris : Presses de Sciences Po, pp. 193–216.
  • Jadot, A. 2002. “(Ne pas) être un électeur européen. Une analyse multiniveaux des déterminants individuels et contextuels de l’abstention en 1999”, Revue internationale de politique comparée, vol. 9, no. 1, pp. 31–45.
  • Klatzmann, J. 1957. “Comportement électoral et classe sociale. Étude du vote communiste et du vote socialiste à Paris et dans la Seine”, in Duverger, Goguel and Touchard (dir.), Les élections du 2 janvier 1956, Paris : Armand Colin, pp. 254–285.
  • Lehingue, P. 2007. Subunda : coups de sonde dans l’océan des sondages, Bellecombe-en-Bauges : Éditions du Croquant.
  • Michelat, G. and Simon, M. 1977. Classe, religion et comportement politique, Paris : Presses de la Fondation nationale des sciences politiques.
  • Pierru, E. and Vignon, S. 2008. “L’inconnue de l’équation FN : Ruralité et vote d’extrême-droite. Quelques éléments à propos de la Somme”, in Antoine and Mischi (dir.), Sociabilité et politique en milieu rural, Rennes : Presses Universitaires de Rennes, pp. 407–419.
  • Rivière, J. 2012. “Trajectoires résidentielles et choix électoraux chez les couches moyennes périurbaines”, Espaces et Sociétés, no. 148–149.
  • Rivière, J. 2011. “La division sociale des espaces périurbains français et ses effets électoraux”, in Mattei and Pumain (dir.), Données urbaines, vol. 6, Paris : Anthropos.
  • Siegfried, A. 1913. Tableau politique de la France de l’Ouest sous la troisième République, Paris : Armand Colin.
  • Veitl, P. 1995. “Territoires du politique. Lectures du ‘Tableau politique’ d’André Siegfried”, Politix, vol. 8, no. 29, pp. 103–122.


[1This homogenisation of types of residential space is all the more problematic as the categories produced by academics are then reused in the speeches and the strategies of the various political parties. A case in point where such transfers between scientific categories and political communication occurs is described in the article “Le livre de gauche qui inspire la droite” (“The left-wing book that inspired the right”), the lead of which specified that, “With Fractures françaises, Christophe Guilluy puts the new working classes of the outer suburbs at the heart of the campaign” (Libération, 30 March 2012).

[2In this text, sociologist Pierre Bourdieu identifies three postulates upon which all opinion polls are implicitly based : (1) everyone can have an opinion ; (2) all opinions are of equal value ; (3) there is a shared consensus regarding the questions that deserve to be asked. He then shows that these three postulates are, in reality, not very well founded and that opinion polls thus derive from the construction of an opinion that does not exist.

[3Behaviourism is an explanatory theory whereby stimulus and reaction form the central elements of the scientific investigation. The notion of “attitude” acts as the interface between the stimuli and behaviours observed.

[4The six “socio-professional categories and professions” (abbreviated to “PCS” in French) are : (1) farmers ; (2) craftsmen, tradesmen, retailers and company directors ; (3) executives and intellectual professions ; (4) intermediate professions ; (5) employees [office workers] ; (6) manual workers.

[5In the 29 February 2012 edition of Le Monde, an article by journalist Thomas Wieder specified that : “Age, sex, and socio-economic category are the variables that spontaneously come to mind to explain voter behaviour. But they are not the only factors. For a number of years now, some specialists in electoral geography have been looking into what they call the ‘urbanity gradient’.” See also the criticism of this explanation proposed by the geographer Guy Burgel in an opinion piece entitled “Le périurbain n’est ni de droite, ni de gauche” (“Periurban areas are neither right-wing nor left-wing”) (Le Monde, 9 March 2012).

[6See the review of this work by Lucie Bargel (2012).

[7For an illustration of the utility of this approach at polling-station level, see the article (in French) in Métropolitiques by Jean Rivière entitled : “Vote et inégalités socio-territoriales : Paris et sa petite couronne” (“Voting and socio-territorial inequalities : Paris and its inner suburbs”).

Commenter cet article

modération a priori

Attention, votre message n’apparaîtra qu’après avoir été relu et approuvé.

Qui êtes-vous ?
Ajoutez votre commentaire ici
  • Ce champ accepte les raccourcis SPIP {{gras}} {italique} -*liste [texte->url] <quote> <code> et le code HTML <q> <del> <ins>. Pour créer des paragraphes, laissez simplement des lignes vides.

Ajouter un document

Suivre les commentaires : RSS 2.0 | Atom

To cite this article:

Joël Gombin & Jean Rivière & traduit par Oliver Waine, « Maps and opinion polls », Métropolitiques, 6 juin 2012. URL:
haut de page

Les textes publiés sur Métropolitiques sont protégés par le droit d'auteur. Toute reproduction interdite sans autorisation.

Site propulsé par Spip2Mentions légales | Contact | Plan du site