CAIRN-INT.INFO : International Edition

1The 2017 French presidential election prompted a number of analyses, in particular those of a cartographical nature—and of variable rigor—emphasizing the divergence between two different versions of France, one of which brought Emmanuel Macron to power, the other having turned to Marine Le Pen. There are numerous different formulations to describe this divide: a “higher” France versus a peripheral France; the France of those who have gained from globalization versus the France of those who have suffered from it; the France of optimism versus the France of pessimism; a France of universal values versus an ethnocentric France; a France open to the world versus a closed France; an urban France versus suburban France. However it is presented, this schism involves various sets of values that are taken to be of importance to the electorate, as well as geographical factors. Indeed, the roots of the divide may be found in internal controversies inherent to the geographical factors influencing why votes were cast as they were in urban and periurban areas. [1] The split also relates to transformations in political science and the competing efforts to desocialize the analysis of electoral behavior. [2]

2In addressing these two academic contexts that come together in the field of electoral studies, [3] the present article uses the example of the French presidential election of 2017 to argue for an electoral social geography that emphasizes the spatial dimension of the electoral effects of social inequalities. Rather than essentializing geographical categories (in this case that of urban dwellers and their supposedly homogenous voting habits), this article focuses on the electoral behavior of the inhabitants of France’s 35 cities with populations exceeding 100,000 (including more than 5,000 polling districts, representing 11% of the electorate of 47.5 million). [4] The choice of these highly robust empirical resources (given that they relate to the votes actually cast) signals a return to the tradition of applying an ecological analysis to voting in France. [5] Furthermore, this involves applying the finest level of aggregation, which has been used increasingly over the past few years, both in geography [6] and in political science. [7] However, working at this level of analysis does give rise to a certain number of methodological difficulties, which explains the paucity of studies at this scale [8] (see box 1). In order to assess this material—and because the statistical techniques employed make use of certain theoretical assumptions—we have favored a geometric data analysis (which includes factorial techniques for reducing information as well as classification techniques), [9] as this allows us to address the relationships between the voting space and the social space (in the Bourdieusian sense of the space of positions), bearing in mind that, “instead of substantialist simplifications (i.e. the existence of unvarying, fixed classes, as defined in and by themselves), the social space should be understood as a multidimensional space of positions defined by multiple factors (e.g. qualifications, age, sex, income levels, heritage, independence at work, difficulty of work, size of the workplace, and place of residence)”. [10] From a geographical perspective, we might add that as many social spaces may be constructed as there are urban contexts. The social space of Paris is not structured in the same way as that of Marseille, for instance.

Box 1. Construction of the Mobilized Data Set

The urban polling district base maps were produced in the course of updating the CARTELEC database, which was led by Céline Colange (research engineer at CNRS unit no. UMR IDEES 6266). She has our thanks for her greatly valued work. Because of technical restrictions, the study of French cities is limited to their respective municipalities or centers, which weakens the analysis sociologically and has probably contributed to less highly correlated statistics being calculated, given that this means losing, in particular, some of the large peripheral working-class areas, whose profile would potentially reveal more heterogeneity in statistical series. The election results associated with these base maps were provided by the Elections Office of the French Ministry of the Interior. Census data from the Institut national de la statistique et des études économiques (INSEE—the National Institute of Statistics and Economic Studies) down to the level of blocks grouped together for statistical purposes (îlots regroupés pour l’information statistique, or IRIS) were used to describe intraurban social structures. The variables selected to capture these social structures in their various dimensions were: the age of residents, their level of qualifications, the employment status of those aged 15 to 64 (taking into account those without work as well as the sociooccupational groups [SOGs] of those in work), their status if in work, the time they have spent living in their neighborhood, and their occupancy status. The two types of linked data do not, however, cover exactly the same populations: electoral results concern only those who are registered to vote, whereas the INSEE data describe the entire resident population, including foreign nationals and minors, who are unable to vote, as well as French adults who are not on the electoral register. Finally, the problem of discrepancies between the spatial grids used for the different base maps (with splits in polling districts in 2017, the 2012 polling district boundaries, and INSEE’s IRIS areas in 2014) was resolved by disaggregating all the data in the spatial grid for the 2017 polling districts by means of a unit surface contribution procedure using Geographic Information System (GIS) software. [11]

3The first part of this article highlights the structure of the electoral space of cities during the first round of the presidential election, embedding it within the underlying social space, thus placing the discussion among the current debates on the relevance of “heavy variables”, as they are known. [12] The second part develops this articulation of the electoral sociology of neighborhoods and the geography of social inequalities by establishing a classification of localized electoral configurations, for which interurban mapping is proposed. The final part seeks to qualify the general observations made by contextualizing them and, in doing so, to emphasize the spatial variability of the relationships between voting patterns and the social structures of different cities. It must be borne in mind that the scope of validity of the conclusions set out in this article relates to the cities of France, which are marked by an overrepresentation of those voting for left-wing candidates (+3 points for Jean-Luc Mélenchon and +2 points for Benoît Hamon) and an underrepresentation of those voting for the far right (with only 9.6% of votes cast for Marine Le Pen in the cities as opposed to 16.1% nationally).

Structure of the Electoral Space in French Cities

How Was the Electoral Space Structured in 2017?

4In order to show how the electoral space was structured in France’s cities, a principal component analysis (standardized PCA) [13] was carried out at the level of the 5,006 polling districts in the 35 urban centers studied. The different options available to voters for the first round of the 2017 presidential election (abstention, a blank ballot, a void ballot, and votes cast for the various candidates expressed as percentages of the electorate) were used as principal variables (the variables that help generate factorial axes), while the results of the first round of the 2012 presidential ballot were added as supplementary variables in order to show the changing dynamics since the previous election and in an attempt to clarify the 2017 vote in political terms. These variables have been situated in the initial factorial design in an indicative manner, but without influencing its construction (see figure 1). [14]

Figure 1

The Structure of the Electoral Space in 2017

Figure 1

The Structure of the Electoral Space in 2017

5The data show that, contrary to what several commentators said following the 2017 election, the left-right divide is not just far from gone, but continues to profoundly structure the electoral space of French cities. Axis 1 of the PCA opposes the polling districts that came out in favor of François Fillon (and secondarily in favor of Emmanuel Macron) with those where voters preferred Jean-Luc Mélenchon (voting secondarily for Philippe Poutou or abstaining)—the respective political positions of the candidates representing Les républicains and France insoumise helping to consolidate the left-right split. Axis 2 is based on an opposition between polling districts that tended towards the center or center-left (the best represented variables along this axis being votes for Emmanuel Macron and Benoît Hamon), and those where the candidates at the periphery of the political space, in particular those on the far right (the top-contributing variables along this axis being votes for Marine Le Pen and then Nicolas Dupont-Aignan) attracted high levels of votes. It is interesting to observe the relative positions of the candidates in relation to these four main poles (left, right, center, and periphery [15]) as shown in the factorial design.

6Hence, Emmanuel Macron is positioned between François Fillon (who is very close to the value indicating the Nicolas Sarkozy vote in 2012, with a 0.95 correlation between the two) and Benoît Hamon in an area of the electoral space that was occupied by François Bayrou, president of the Democratic Movement (MoDem), in 2012. In the cities, Macron’s victorious campaign is sited in the center-right, inasmuch as the votes that he received have a strong statistical link with Bayrou’s successive campaigns. There is a 0.73 correlation with the votes cast for François Bayrou in 2012, and secondary correlations of 0.59 with Eva Joly and 0.36 with Nicolas Sarkozy. There is even a 0.78 correlation with the turnout for Bayrou in 2007—François Bayrou managed to attract over 18% of the vote, coming third in the first round. With a historically poor result, the Parti socialiste (PS) candidate ended up between Macron and Mélenchon on the factorial design, in an area of the electoral space that was occupied by both Eva Joly (r = 0.73) and François Hollande (r = 0.69) in 2012. The Hollande vote was transferred at least as much to Benoît Hamon as it was to Jean-Luc Mélenchon, whose results have almost as strong a correlation with those of the outgoing president (r = 0.67 with François Hollande in 2012) as with his own previous performance (r = 0.85 with Mélenchon in 2012). Not far from the value indicating Mélenchon’s vote, a little higher on the design, we find the values representing the various far-left candidates (Philippe Poutou and Nathalie Arthaud), whose results positively correlate with those of François Asselineau, with abstentions and with blank and void ballots. Finally, the upper part of the design contains the far-right candidates, Marine Le Pen and Nicolas Dupont-Aignan. Following the first round, Dupont-Aignan urged voters to choose Le Pen in the second round (r = 0.50 between the two in the first round). [16]

7Despite the candidates’ rather artificial positions in the first round of the 2017 election, this social space becomes entirely clear when we add the sociological characteristics of residents of different polling districts.

Identifying the Intra-Urban Social Structures that Can Explain the Electoral Space

8Here, we have drawn up a similar PCA as for the previous sub-section, but this time the supplementary variables are indicators describing the social characteristics of voters in the different polling districts on the bases of socio-demographics (age group and qualifications), socio-economics (socio-occupational group and employment status), and residency (period of establishment in the neighborhood and occupancy status). [17] By examining this factorial design (figure 2) and the correlations between voters’ decisions and the social structure of polling districts (figure 3), we can form an up-to-date picture of the deeply intricate links between the electoral sociology of neighborhoods and the geography of social inequalities.

Figure 2

Mapping the Electoral Space in the Urban Social Space

Figure 2

Mapping the Electoral Space in the Urban Social Space

Figure 3

Correlation Coefficients Between Voting Decisions and Social Structures of Polling Districts

Figure 3

Correlation Coefficients Between Voting Decisions and Social Structures of Polling Districts

The light grey boxes indicate correlations of over 0.20; the dark grey boxes mark the strongest correlation in each line. The values in bold differ from zero with a significance level of alpha = 0.01.

9As with the two axes describing the electoral space, the social space of urban polling districts may be described according to the two dimensions that are graphically represented using the grey arrows on figure 2. The first of these represents the opposition between polling districts whose SOGs consist of working-class groups (employees and laborers), those where there are low levels of qualifications (e.g. those with no qualifications and those with only primary school leaving certificates, vocational diplomas, or other vocational qualifications), and those polling districts situated in more comfortable areas with better-off SOGs (management level and the “higher intellectual professions” [18]) and residents with a high level of academic qualifications (baccalaureate + 2 years of study, bac + 3 years, or more). This dimension, which crosses the design from the top-right to the bottom-left, highlights the importance of the volume of capital (both economic and cultural) in generating political alignments and voting decisions. [19] The second dimension, which crosses the design from the bottom-left to the top-right, makes clear the opposition between polling districts inhabited predominantly by a younger population (18-24 and 25-39) and those who have arrived in the area more recently (within two years or between two and four years previously) and those districts with a more elderly population (65-79 and 80+) and residents who have been living locally for a longer period (for 5 to 9 years or for 10 + years).

10Going into more detail, the geography of the François Fillon vote reflects a social environment that symbolizes the traditional right-wing vote that Fillon embodies. The results for the Les républicains candidate are higher in those polling districts where there is a high concentration of older residents (r = 0.24 among 65 to 79-year-olds and 0.32 for those over 80), of those with university degrees (0.24 among those with a baccalaureate + 2 years and 0.62 for those with a bac + 3 years or more), and above all of people in higher sociooccupational positions (r = 0.60 for the MHIP group and 0.57 among farmers, artisans, traders, and business owners) and owner-occupiers. In those places where Emmanuel Macron achieved his best scores, the area profile is fairly close to that just described, thus confirming the link between the winning candidate and those categories of voter with greater capital. In Macron’s case, however, the emphasis is on younger age groups (r = 0.31 among 25 to 39-year-olds) and hence those who are less established as residents in their neighborhoods (r = 0.38 among private sector tenants), but also on those with more qualifications (r = 0.78 among those with qualifications exceeding bac + 3 years) and those in senior salaried positions (r = 0.81 among the MHIP group—this being the strongest correlation recorded in the entire matrix).

11We find that the area of social space favoring Benoît Hamon also has a fairly young resident profile (r = 0.49 for 25 to 39-year-olds and 0.29 among 18 to 24-year-olds), albeit its position is a little lower in the social hierarchy (r = 0.36 with MHIPs and 0.24 among mid-level professionals), with a lack of either occupational or residential stability (r = 0.31 with those with an unpredictable income and r = 0.32 with residents who have lived in their neighborhood for less than two years), which also indicates the presence of students in urban environments. More generally, these areas of the social space represent a space of transition towards social spheres populated by the young and by the working class, which supported the more left-wing candidates such as Philippe Poutou and, most of all, Jean-Luc Mélenchon. Indeed, Mélenchon’s results correlate both with high concentrations of young residents and with areas whose residents have few academic qualifications (r = 0.34 for those with no qualifications, and 0.22 for those with vocational qualifications), belong to working-class groups (0.45 with the unemployed and 0.28 with laborers) and have modest conditions of material existence (r = 0.45 with those with unpredictable pay and 0.36 with social tenants). These empirical elements indicate a shift compared with the 2012 results, at least in the cities. In that year, it was a moot point whether Mélenchon’s campaign had a “dynamic that had little to do with social variables” and drew from “a non-working-class vote”. [20] Above all, however, it is the “electoral exodus”, [21] whether temporary or enduring, [22] that has characterized the attitudes of those in the most precarious groups when it comes to voting, [23] bringing to mind the very strong correlations between abstention rates and polling districts containing high numbers of residents with no qualifications (r = 0.76), unemployed people, laborers, and those listed as “not working (other)” (r = 0.77, 0.66, and 0.64 respectively), as well as those unable to avoid living in the least desirable residential areas (r = 0.57 with social tenants).

12Showing weaker correlations, the same social indicators appear in the top-right quarter of the factorial design, close to the results of the far-left candidates, the blank ballots, and the void ballots. In the upper part of the design, the scores of the far-right candidates (Marine Le Pen and, secondarily, Nicolas Dupont-Aignan) appear to show a statistical link with the quarters containing residents in the middle age groups (r around 0.35 with 40 to 54-yearolds, 55 to 64-year-olds, and even 65 to 79-year-olds—an indication that the generational effect among voters born around the time of the Second World War that limited the Front national (FN) vote may now be coming to an end). Sociologically, the Le Pen vote in the urban centers appears to be strongly connected with a low rate of qualifications among voters (r = 0.65 for those with vocational qualifications, 0.55 for holders of a primary certificate, and 0.43 for those with a national diploma), with working-class voters (0.63 with employees, 0.54 for laborers, and 0.38 for those not in work), with those at the end of their careers (0.51 with people who have taken early retirement), and with those living in areas with a preponderance of people on stable salaries (0.47 where there are holders of permanent contracts and civil servants). This set of indicia must be interpreted with caution, but it does lead us to the conclusion that it was not the most vulnerable segments of the population (who are not registered to vote or who abstain, as we have just shown), but those sections of the working class who feel a “triangular conscience”, [24] in other words an awkward sensation of not belonging with those in a more dominant social position, while being adamantly distinct from those in a lower position, who turned towards the FN in greater numbers. [25]

13It remains to be seen how these different voting behaviors came together at the polling district level, and how the vote can be mapped in the intra-urban space of French cities.

Classifying the Intra-Urban Socio-Electoral Configurations

14The aim of this second section is to provide an overview of the electoral landscape of France’s cities through a classification of their polling districts. A hierarchical agglomerative clustering (HAC) exercise has allowed us to identify seven profiles for the urban districts that have been mapped (see figures 4 and 5) on the basis of their voting characteristics (see figure 6). [26] Each profile description is included in these two figures and in the table in appendix 1, which shows the average sociological profile of residents with each profile. While regional effects may be seen (as with the prominence of type B3 districts in the Mediterranean cities with a notable far-right vote), this classification is highly transversal in its nature throughout the cities studied, with an average of 5.3 profiles out of 7 present in each city. In 15 of the 35 cities featured, six or even seven profiles are present). As the hierarchical classification tree shows, the electoral landscape of these cities features two major groups of polling districts: those in the city centers, which are oriented towards the right and center of the electoral space; and those on the edge of the center or on the periphery, which, depending on their profiles, had more voters abstaining or voting for candidates on the left or the far right.

Figures 4 and 5

Polling District Classification Mapping

Figures 4 and 5Figures 4 and 5

Polling District Classification Mapping

Figure 6

Electoral Profiles According to Classification

Figure 6

Electoral Profiles According to Classification

The light grey boxes indicate values above the average for cities as a whole; the dark grey boxes indicate the profile for which each candidate achieved their highest score.

Domination of the Right and the Center in Central Urban Polling Districts

15Most A1 districts in Paris are situated in the west of the capital, which generally corresponds with the pleasant historical areas of other cities high up in the urban hierarchy [27] (Lyon, [28] Strasbourg, Nantes, and Bordeaux [29]). Having already been identified after the 2007 presidential election in the seventh, eighth, and sixteenth arrondissements, [30] they offer a clear illustration of how the middle class are the last social class who feel mobilized, with a shared sense of common interests to maintain along with the means required to defend those interests in the political sphere. [31] This grouping of districts is clearly marked by the highly pronounced overrepresentation of votes for François Fillon (44% of the vote, i.e. three times the average in the cities), which is linked to the high turnout for Emmanuel Macron (26%), with the total votes for left-wing candidates therefore coming to less than 10%. The proportion of elderly residents (in the 65-79 and 80+ age groups) is slightly higher in these beaux quartiers compared with the average, [32] and more than half of the residents have completed three years of higher education (baccalaureate + 3 years, with two thirds of the electorate in these areas having completed at least bac + 2 years). There are also marked overrepresentations of the MHIP group, artisans, traders, business owners, and owner-occupiers, who make up 46% of residents, even though three quarters of districts with this profile are concentrated in Paris, where property costs are highest.

16The A2 districts often form a geographical extension of those that have just been described in the major cities such as Lyon, Marseille, and Toulouse, of better-off areas of cities lower down in the urban hierarchy, such as Tours, Angers, and Le Mans, or of conservative cities such as Aix-en-Provence, Orléans, Dijon, and Nancy. These districts also reveal favorable voting patterns for the right and the center, albeit in a less clear-cut way (with 25% voting for Fillon and 23% for Macron). The demographic characteristics of these social areas are close to those previously described, with a slight overrepresentation of older residents, although their position is a little lower in an academic respect (47% of those living there have a bac + 2 years of study, compared with a 39% average figure) and vocationally (with the MHIP group and mid-level professionals accounting for 39% of residents, which compares with an average of 33%). Given the small presence of Parisian districts in this category, slightly more residents of the better-off group A2 districts (48%) own their own homes.

17Conversely, group A3 contains half of all polling districts in Paris: all of the first, second, third, fourth, fifth, ninth, thirteenth, and fourteenth arrondissements that were gentrified in the 1980s, together with large sections of the east of the capital (the tenth, eleventh, twelfth, and eighteenth arrondissements) that underwent a gentrification process in the 1990s. [33] Similar district profiles are prominent in the electoral landscape of cities in western France (in 33% of districts in Nantes, [34] in 32% of districts in Rennes, and in 30% of those in Caen). These are relatively young areas (with nearly a third in the 25-39 age group) and contain the highest proportion of private sector tenants (45%), with a high level of qualifications (43% hold a bac + 3 years or more) and double the concentration of MHIPs compared with elsewhere (31% compared with a 16% average). The profiles of A3 districts are marked by high levels of votes for Emmanuel Macron (31% compared with an average 21%), combined with a slight overrepresentation of votes for François Fillon (+2 points) and for Benoît Hamon (also +2 points—and this is the district profile with the highest Hamon vote). These polling districts are hence closest to the center of the electoral space, and represent transitional spaces between the more desirable, right-leaning districts and the traditionally left-voting urban areas.

Edge-of-Center and Peripheral Polling Districts: A Split Between Abstentions, the Left, and the Far Right

18The category B1 districts are those where the total number of votes cast for left-wing candidates is highest (33%), with particularly high scores for Jean-Luc Mélenchon (23%). These districts are located in some of the historical urban left-wing bastions, which form an arc around the east side of Paris—from the the eastern edge of eighteenth arrondissement through the nineteenth and twentieth arrondissements down to the southern areas of the thirteenth and fourteenth—as well as in the fourth electoral district of Bouches-du-Rhône, where the high turnout for Mélenchon has highlighted the forgotten patterns of Marseille’s support for the Parti communiste français (PCF), as identified by the few maps drawn up in the 1970s and 80s that went down to the level of polling districts. [35] More generally, these districts are situated in the central and edge-of-center areas in the top fifteen cities in the French urban hierarchy, which have stronger left-wing leanings. This relates to more than a third of polling districts in Grenoble, [36] Toulouse, Rennes, Bordeaux, Nantes, Montpellier, Rennes, Lyon, and Lille. In these parts of the large service-oriented metropolitan areas, [37] which host the country’s leading universities, we find the highest concentration of younger residents (with the 18-24 and 25-39 age groups together making up 51% of the population), the highest proportion of pupils, students, and interns among those aged between 15 and 64 (18%), and also the highest proportion of those with unpredictable pay (those on fixed-term contracts and temporary workers). As may be expected given the presence of these younger residents at the early stages of their careers, the B1 districts are those where residential turnover is highest, with nearly half of residents not having lived in the same home five years previously. Indeed, this suggests great caution is required in drawing correlations from aggregate data, especially because of under-registration among students.

19The profile for B2 districts appears similar to that for B1, albeit the overrepresentation of votes for Jean-Luc Mélenchon is associated as much with abstentions and blank and void ballots as with the far-left candidates (Philippe Poutou and Nathalie Arthaud) and the far-right candidates (Marine Le Pen and Nicolas Dupont-Aignan). This voting pattern is highly prominent within historically working cities lower down in the urban hierarchy (with populations of between 100,000 and 200,000), and prevailed in more than a third of districts in Le Havre, Saint-Étienne, Clermont-Ferrand, Brest, Limoges, Amiens, Metz, Besançon, and Avignon. The inhabitants of these districts are in no way different from the urban population in general as far as demographics is concerned, although they have a lower level of cultural capital, with 54% not having a baccalaureate, in contrast with an average level of 44%). In line with these lower qualification levels, voters here are slightly more likely to hold working-class socio-occupational positions than is the case elsewhere, with the employee, laborer, and unemployed categories making up 42% of the population aged between 15 and 64, compared with a 35% average. Moreover, the numbers of social tenants are slightly higher than in other urban settings.

20There are more owner-occupiers living in the B3 districts (51%, compared with 40% on average), although this relates to their particular presence in France’s Mediterranean cities, where there is a high proportion of comfortably retired people alongside a significant level of social inequality The B3 profile, which notably features a high concentration of votes for Marine Le Pen (22% of votes cast, compared with a 10% average across the cities studied) in line with the regional trend, accounts for 70% of voters in the polling districts of Toulon, 63% in Nice, [38] 53% in Perpignan, 57% in Marseille, and 38% in Nîmes. Residents aged 65 and over make up 26% of B3 voters, providing a generational explanation for the slightly lower preponderance of people with higher education qualifications compared with elsewhere. Finally, the B3 profile is close to the average in relation to occupational hierarchies, albeit with an underrepresentation of the MHIP group (7 points lower).

21The B4 group is markedly different from the B1, B2, and B3 categories, because of the very high abstention level (39% of voters, compared with an average of 23%), combined with slight overrepresentations of votes for Jean-Luc Mélenchon, the far left, and the far right. It should be emphasized that the abstention rate in these areas is widely underestimated, in that their residents are as often unregistered or incorrectly registered as not, [39] not to mention the element of foreign nationals among the working-class population, who do not have the right to vote. [40] The location of these districts matches the large residential zones in most French cities. These include the northern arrondissements of Marseille, the southern edges of Lyon adjoining Vénissieux, le Mirail and la Reynerie in Toulouse, la Paillade in Montpellier, [41] and Caucriauville in Le Havre. The sociological profile of C4 polling districts symbolizes that described in various works on the quartiers in the context of French urban sociology: 37% of residents in these districts have no qualifications, and fewer than 15% have university degrees; 67% of those aged between 15 and 64 come under the categories of employees, laborers, unemployed, and not working (other); 56% are social tenants. This range of indicators helps us understand why there is such a persistent distancing of people from the democratic institutions in these areas, where their material conditions tend to deteriorate irrespective of who is in power. [42]

22These categorizations allow us to describe a number of overarching intra-urban electoral categories (the analysis may be refined by adding more categories to show the internal nuances within the profiles already established), to apply further sociological clarity to them, and to map them on a large scale. Nonetheless, it may also be possible to ascertain the spatial dimension between voting orientations and local social structures by other means.

Making Explanatory Models Using “Variable Geography”

23Whereas the foregoing analyses carry a certain level of validity covering French cities as a whole, we can make the hypothesis that the local and regional patterns that have been established, such as the size of the FN vote in the Mediterranean cities (group B3) and the very high level of support for François Fillon in the well-off neighborhoods of the west of Paris (type A1), will alter the statistical relationships on a national scale by aggregating the systems of relationships that apply to each city—although these do not necessarily all go in the same direction. This final section therefore underlines the spatial variability of the assertions made hitherto. This does not invalidate them, but in fact makes it possible to reinforce them by taking local specificities into account.

What is the Spatial Variability of the Statistical Relationships Already Established?

24Before examining the variations in the correlations between voting behavior and social structures in a set of urban contexts, we shall first consider the relationship between the presence of a working-class population and the votes for Marine Le Pen, even though the theory applies to all candidates and all SOGs. This relationship is a subject of controversy among election experts, [43] to the extent that investigating the relationship between the working-class presence in an area and the FN vote has become a sort of empirical laboratory for the renewed uptake of the ecological approach to voting patterns, which has shown, for instance, that “the most dominated social categories, starting with workers, are those whose behavior [in terms of voting FN] seems to vary most in spatial terms”. [44] Correlations between the locations of working-class SOGs and the pattern of votes for Le Pen were calculated separately for each of the 35 cities studied (r = 0.54 according to figure 1), [45] and these correlation coefficients were then mapped (see figure 7). This analytical strategy is related to the associated approach to the initial stages of multilevel analysis, where recent developments have proved fertile ground for French electoral sociology. [46]

Figure 7

Understanding the Geographical Variation in the Correlations Between Working-Class Populations and the FN Vote

Figure 7

Understanding the Geographical Variation in the Correlations Between Working-Class Populations and the FN Vote

25It should be noted, first of all, that the correlation coefficients in figure 7 are not strictly comparable with each other, as Pearson’s r correlation test uses the number of “degrees of freedom” to integrate the population of the statistical series. Hence, a coefficient must reach 0.33 to be significant for a threshold of 1% (p<0.01) for Caen (56 districts), for example, but it only has to be 0.25 for Saint-Étienne (100 districts), or just 0.18 for Nantes (200 districts), and a mere 0.12 for Marseille (480 districts). However, while the implication from the tables of critical values is that the required threshold is lower for a higher number of individuals (in other words, in cities with more polling districts), the coefficient values are broadly the same whatever the size of the city. It may therefore be deduced (while exercising caution with regard to the current debates about using correlation tests) that the statistical relationships between votes cast and social structures are slightly stronger further up the urban ranking. [47] Levels of inequality and social segregation (which generally increase with the size of towns and cities) provide one line of explanation to shed light on this hypothesis. [48]

26In mapping the correlation coefficients of the FN candidate with the location of the workingclass SOG, we find a contrast between cities in the south-east of France and those in the north-west, which brings to mind the apparent divide between the FN in the north, which is more social and populist and attracts more workers, and the FN in the south, which is more identity-based and economically liberal in nature and is more geared to attracting retired people. [49] Above all, however, the map shows an urban France with the highest levels of populations who come from the other side of the Mediterranean (in the sense of being immigrants or foreigners), possibly indicating internal divisions within working-class spheres as regards attitudes to immigration. Within the city of Marseille, for instance, there is only a weak correlation between the level of the working-class population and the FN vote (at 0.34), although a multilevel analysis has shown that, when the analysis reaches the level of municipal arrondissements, these relationships are markedly different in both degree and direction, depending on the share of immigrants within the population of the area concerned. [50] The scatter plot on the right-hand side of figure 7 attempts to test this hypothesis for polling districts in French cities as a whole, crossing the strength of the correlation between working-class areas and the Le Pen vote (the information mapped on the left of figure 7) with the “immigrant” population share in each city. [51] If we remove three individual statistics from the series where the share of “immigrants” is greater than 20% (Paris, which, as the capital city, effectively acts as an airlock for migration movements, and the two cities in Alsace, which have a particular history as border towns), a strong negative correlation may be established (r = -0.73—significant at a threshold of 0.01%), in defiance of common sense. In western cities such as Caen, Tours, and Rennes, whose populations have only a small “immigrant” component, there is a very strong relationship between the working-class presence and the FN vote. Conversely, in the cities of the south-east, where the “immigrant” presence is both stronger and longer established—and where many workers are themselves “immigrants”—there is only a weak or even insignificant relationship between the workingclass population and the FN vote. It should further be noted that it is in these cities of the Mediterranean basin and what used to be the “red south” (Nice, Toulon, Marseille, Aix, Avignon, Nîmes, and Perpignan), whose Gini coefficients reveal some of the highest levels of inequality in the country, where the relationship between the working-class presence and the Mélenchon vote is the strongest, with the France insoumise candidate competing with the FN in the polling booths.

27It seems essential, therefore, to take some time to observe how the intra-urban social structures manifested themselves in the election results in what are, by definition, individual contexts.

Focus on One Urban Area: Looking Beyond the Core City and Disaggregating the SOGs

28The aim of this final sub-section is to extend the reasoning adopted up to this point by applying it to the capital—Paris being the grande métropole par excellence, given the macrocephaly that is the French urban pattern. [52] Two steps shall be taken to enhance the sociological accuracy. Firstly, the geographical scope of the analysis shall be widened. Secondly, the classification used for the social structure will be refined so as to reintroduce the distinction between public and private [53] into the analysis. One source of bias in the foregoing analysis stems from methodological constraints: the 35 cities studied were limited to their central municipalities, which cuts them off from their conurbations and urban areas. Although the periurban area of the Île-de-France largely transcends the confines of the petite couronne or inner ring, [54] it is this zone that comprises around a hundred communes to be analyzed. Given how much the specific sociology of the French capital (characterized by its high concentration of what INSEE has termed “superior metropolitan jobs”) has been observed to compromise the analytical usefulness of the socio-occupational category in its limited SOG-based version, we propose to test the electoral effects of disaggregating this classification in the upper part of the social stratification, whose growing concentration in the urban setting is now regularly to be found at the heart of public debate. Aside from the set of technical problems inherent in working at the level of polling districts (see box 1), this approach poses a number of other specific challenges (see box 2).

Box 2. Achieving a Detailed Classification in an Intra-Urban Context

If we are to access these data down to the level of the IRIS group of blocks, this requires either an enhanced level of statistical security clearance or a specific product request to INSEE via the Quetelet Center. While the latter option was chosen, INSEE limits the number of variable modalities to twelve and, for reasons of statistical confidentiality, blanks out those fields that do not contain a sufficient number of individuals. In order to take these two rules into account, the request that was formulated proposed using twelve sets with the aim of minimizing the number of blanked out fields, rather than using the 24-set French PCS categorization (professions et catégories socioprofessionnelles, or occupations and socio-occupational categories), which is impractical for maintaining statistical robustness at this scale with the updated census. Despite all this, of the 2,752 IRIS areas that make up Paris and the inner ring, 666, or 25% of the total, were blanked out. These IRIS areas were located throughout the space studied, although the blanking out of several areas in the eighth and sixteenth arrondissements of Paris is sure to result in a fairly significant underestimation of correlations. We thank the staff at the Quetelet Center (in particular Alexandre Kych and Benoît Tudoux) for their time and their work in extracting the data. However, data from 174 IRIS areas proved to be unusable, because some of the divisions into polling districts were not available for the corresponding municipalities. The following analyses cover a total of 1,911 IRIS areas (69% of the total number) where election results covering each polling district have been collected, noting that a disaggregating of INSEE data derived from a specific product request using polling districts as geographical divisions would be too risky in terms of reliability. The IRIS subdivision is approximately of the same size as that of polling districts, albeit slightly less detailed, with the space studied comprising 2,752 IRIS areas as opposed to 3,347 districts.
We should clarify that, because of its complexity, this initiative under the CARTELEC project, [55] funded by the French National Research Agency (ANR), could not be run again for the 2017 presidential election that is the subject of this thematic issue. The same reasoning could still apply here, however. More fundamentally, this article seeks to contribute to the theoretical and methodological debates concerning voting analysis beyond the scope of 2017—hence the relevance of this temporal decoupling for the reader.
tableau im8
SOG SCs xith 2 figures together Abstention Blank/void Schivardi Laguiller Besancenot Bové Buffet Royal Voynet Bayrou Sarkozy De Villiers Nihous Le Pen SOG1+2 Small business owners (11+21+22) -0.18 -0.29 -0.08 -0.24 -0.28 -0.12 -0.17 -0.26 0.11 0.10 0.30 0.07 -0.05 -0.08 -0.32 -0.34 -0.14 -0.41 -0.44 -0.20 -0.28 -0.29 0.03 0.23 0.43 -0.03 -0.16 -0.32 Larger business owners (23+31) -0.30 -0.28 -0.13 -0.37 -0.40 -0.18 -0.26 -0.23 -0.01 0.22 0.38 -0.05 -0.15 -0.32 SOG3 Pub. sect. mgt., intell. and art. prof. (32) -0.55 -0.45 -0.14 -0.48 -0.56 0.00 -0.40 0.09 0.42 0.60 0.19 -0.19 -0.24 -0.65 -0.70 -0.54 -0.23 -0.62 -0.76 -0.26 -0.55 -0.25 0.37 0.74 0.51 -0.06 -0.19 -0.69 Company executives (36) -0.70 -0.51 -0.25 -0.63 -0.77 -0.38 -0.57 -0.43 0.28 0.72 0.64 0.03 -0.13 -0.61 SOG4 Public sec. mid-level prof. (41) -0.23 -0.05 0.19 0.08 0.06 0.19 -0.01 0.08 0.46 0.31 -0.14 0.15 0.14 0.00 -0.33 -0.16 0.16 -0.01 -0.07 0.12 -0.15 0.00 0.55 0.43 -0.03 0.16 0.15 -0.06 Private sec. mid-level prof. (46+47+48) -0.31 -0.19 0.09 -0.07 -0.15 0.04 -0.21 -0.06 0.47 0.40 0.06 0.13 0.12 -0.08
The values in bold are significant up to 1%; the shaded boxes indicate significant positive correlations.

29There are three scenarios that arise and provide further nuance to the observations made up to this point (see figure 8). The first of these scenarios is rather marginal, but nonetheless interesting, and relates to situations where the correlation for a given SOG is greater than for the most similar social categories (SCs). This is especially the case for the votes cast for François Bayrou and Dominique Voynet. These two candidates, close to the political center, gained from the cumulative effects of a residential concentration of two SCs with apparently similar behaviors (r = 0.60 with the presence of public sector managerial grades for Bayrou, and 0.72 with private sector executives, for instance), bearing in mind that the correlation between the distribution of these two sections of the management-level workforce is itself 0.65 in the space studied. The second scenario relates to the existence of significant gaps between the correlations of two sections of SOG, albeit that there is no sign of the coefficients being reversed. The SOG’s coefficient is in an intermediate position between those of its corresponding SCs, and is in a sense averaged out by the coexistence of different behaviors within it. The most symbolic example is without doubt the 0.51 correlation between the Sarkozy vote and the presence of the MHIP SOG, which masks the divide between the public sector management group (0.19) and the company executives group (0.64). The hypothesis may be proposed that a similar phenomenon took place in Paris in 2017, producing particularly strong correlations between the presence of private sector executives and votes for François Fillon and Emmanuel Macron. Although they are less clear cut, the examples of the correlations between the presence of the SOG3 SCs and the Dominique Voynet vote and of the SOG4 SCs with the Marie-George Buffet vote also correspond with this situation. The third scenario confirms the existence of major gaps between the correlations of two SC divisions, yet with the coefficient sign reversed. Once again, it is an internal division among those at management level that emerges most clearly, as the -0.25 correlation between the Ségolène Royal vote and the presence of SOG3 hides the actual opposing behavior of the two CSs specified here: -0.43 with private sector executives and 0.09 with public sector managers. It would be interesting to apply this to the votes cast for Benoît Hamon. The same phenomenon applies, albeit to a less significant extent, to the relationship between the two components of SOG3 and the José Bové vote, between the two parts of SOG4 and the Nicolas Sarkozy vote, and between small and large business owners and the Philippe De Villiers vote.

30Moving onto a final analysis, while we have been able to establish that the location of the upper middle and upper classes in the petite couronne or inner ring around Paris corresponds with the favoring of centrist and right-wing candidates, each SOG appears to feature internal divisions to a greater or lesser extent. This is the case within SOG3, where, in the various IRIS areas, significant proportions of public sector managers lean more to the center-left, whereas those urban areas where there is a higher presence of company executives show a stronger correlation with votes for Bayrou and Sarkozy. As for the distinctions to be drawn among mid-level professionals, the spatial distribution of those employed in the education, healthcare, and social work sectors may show particular correlations with the left-wing candidates. Indeed, an analysis of the support for Mélenchon among this classification would be highly informative. Among the self-employed labor force, finally, the distribution of small business owners (farmers, artisans, and traders) is a little less correlated with the results of right-wing candidates, and in some cases is slightly more correlated with the far right compared with the distribution of larger business owners (i.e. company heads and those working in the liberal professions).

31* * *

32In sum, the proposed analytical strategy consists of contextualizing the analysis, rather than isolating the contextual effects as per British and American studies—and as criticized by Julien Audemard in a recent highly thought-provoking theoretical article. [56] Although this may be an acceptable criticism, it nonetheless appears debatable, for two reasons. Firstly, as is typical with interdisciplinary dialogue, there is a certain lack of understanding born of the vocabulary employed—in this case, by election analysts. Whereas, for the political analyst, the meaning of the term “context” appears to relate directly to social collectives (peer groups in particular) along the lines of the approach initiated by Paul Lazarsfeld, geographers will use this term in the sense of local context (which is itself inseparable as a social context). What is more, this is in a disciplinary framework where references to “ecological” or “environmental” approaches may give rise to further misunderstanding, this time with physical geography. Semantic uncertainty aside, the second reason why this criticism may appear unjustified is that the use of aggregated data must be considered in a “double context challenging the heuristic potential of the concept of social class in explaining electoral behaviors and the development of competing illustrations based around the concept of individual rationality”, as Audemard himself notes. [57]

33In this double context, the criticism leveled against the use of aggregate data appears to have partly missed its target. Contrary to what sometimes prevails among electoral analysts, where the specter of “Robinson’s paradox” [58] has frequently been deployed to discredit ecological analysis (regarded as the “ecological fallacy” to encourage the use of individual data, albeit that this is affected by the atomistic fallacy [59]), this article demonstrates that statistical correlations do not inevitably reach high values when we work at the most detailed geographical level. [60] Indeed, the very high levels reached by some correlations (established among several thousand statistical individuals) reminds us that the notion of voting by class in urban areas riven by stark inequalities is alive and well: r = 0.81 between the presence of the MHIP group and votes cast for Emmanuel Macron; 0.76 between the presence of residents with no qualifications and abstentions; 0.63 between employees and the Marine Le Pen vote; and 0.57 between the self-employed and the vote for François Fillon. It is clearly necessary to take time to observe how electoral choices can become socially entrenched in the localities where the votes are cast and where they can be made sense of. Certain regional or urban economic fundamentals will produce specific social configurations, which, as with local political developments, can markedly diverge from national tendencies when it comes to voting alignments—and they can even generate local sub-systems providing different explanations that cancel each other out, thus making it impossible to update the statistical relationships at a national scale. Finally, we have attempted to show that, by disaggregating SOGs in order to construct (slightly) more detailed analytical categories, these locally specified correlations could be further refined, such that if we use multidimensional social spaces combining such factors as age, qualification levels, and employment status so as to shed more light on each city’s electoral spaces, this could produce an analysis of votes by class that, “paradoxically, must distance itself from the premises of the concept’s inventor in order to maintain that concept’s usefulness”. [61]


Average Sociological Profiles of Electoral Categorization Groups

% aged 18-2414.117.016.521.216.411.416.116.7
% aged 25-3924.624.732.830.
% aged 40-5422.921.121.620.622.324.424.122.1
% aged 55-6414.614.012.511.913.715.313.913.5
% aged 65-7915.114.411.110.513.216.712.913.1
% aged 80+
% No qualifications8.
% Primary certificate2.
% National diploma3.
% Vocational qualif.5.313.99.414.820.920.621.916.3
% Baccalaureate14.717.214.616.616.517.713.416.2
% Bac + 2 years12.215.914.515.613.212.57.413.6
% Bac + 3 yrs or more53.231.443,728,416,614,07,225,6
% Farmers, artisans, traders, business owners5.
% MHIP34.120.530.617.510.49.63.516.6
% Prof. inter.% Mid prof.12.818.117.217.316.016.48.516.0
% Employees10.614.
% Laborers2.
% Unemployed6.98.68.510.912.410.618.911.1
% Students interns16.516.614.817.813.711.313.114.9
% Retired incl. early retired3.
% Not working (other)
% Self-employed23.314.814.510.78.711.87.611.9
% Steady salary (civ. serv., perm. contr.)65.470.469.770.374.075.471.371.6
% Unpredict. pay (fixed-term contr., temps)11.314.815.719.017.312.721.116.5
% <2 yrs in nbrhood14.417.916.919.816.712.513.216.5
% 2-4 yrs in nbrhood23.925.525.426.925.422.024.125.1
% 5-9 yrs in nbrhood20.118.319.618.419.319.121.119.2
% 10 yrs+ in nbrhood41.638.338.034.938.746.441.639.2
% Owner-occupiers46.548.339.034.939.550.621.940.2
% Private sector tenants43.239.344.641.631.830.120.936.0
% Social tenants3.59.712.221.226.716.455.921.1
% Freely housed6.


  • [1]
    On the controversy surrounding the promotion of space as a factor to explain voting patterns independently from social factors, see in particular the following two pairs of articles, which respond to each other’s points: Fabrice Ripoll and Jean Rivière, “La ville dense comme seul espace légitime? Analyse critique d’un discours dominant sur le vote et l’urbain∫, and Jacques Lévy, “Regarder, voir: un discours informé par la cartographie∫, Les annales de la recherche urbaine, 102, 2007, 121-30 and 131-40; and Éric Charmes, Lydie Launay, and Stéphanie Vermeersch, “Le périurbain, France du repli?” and Jacques Lévy, “Liens faibles, choix forts: les urbains et l’urbanité”, La vie des idées, 2013 (online). For a recent summary of these debates, see Aurélien Delpirou, “L’élection, la carte et le territoire: le succès en trompe-l’œil de la géographie”, Géoconfluences, 2017 (online).
  • [2]
    “One need only take a look at the list of major political science periodicals to be convinced of this. It has been several decades since voting science lay mainly in the domain of sociology. As a discipline, it features various paradigms and methods from two disciplines, economics and psychology, whose core purpose is no longer that of analyzing voting as a social phenomenon; it now seeks to treat voting as a product of individual reason or psyche”. Céline Braconnier, Une autre sociologie du vote. Les électeurs dans leurs contextes, Paris, LGDJLextenso Éditions, 2010, 1. Our translation. Note that, unless otherwise stated, all translations of cited foreign language material are our own.
  • [3]
    Joël Gombin and Jean Rivière, “Vers des convergences interdisciplinaires dans le champ des études électorales”, L’espace politique, 23, 2014 (online).
  • [4]
    In this respect, the present article uses a comparative dimension to establish some general points, thus extending analyses that were initially offered by way of a thematic paper in which around ten cities were examined in turn: Christophe Batardy, Emmanuel Bellanger, Pierre Gilbert, and Jean Rivière, “Présidentielle 2017: les votes des grandes villes au microscope”, Métropolitiques, 9 May 2017 (online).
  • [5]
    François Goguel, “Structure sociale et opinions politiques à Paris d’après les élections du 17 juin 1951”, Revue française de science politique, 1(3), July-September 1951, 326-33; Jean Ranger, “Droite et gauche dans les élections à Paris: le partage d’un territoire”, Revue française de science politique, 27(6), December 1977, 789-819; Daniel Derivry and Mattei Dogan, “Religion, classe et politique en France: six types de relations causales”, Revue française de science politique, 36(2), April 1986, 157-81.
  • [6]
    Led by the geographer Michel Bussi, the CARTELEC research program (see Laurent Beauguitte and Céline Colange, “Analyser les comportements électoraux à l’échelle du bureau de vote”, a scientific paper produced for the Agence Nationale de la Recherche, 2013, online) was an important tool in disseminating research on this scale. See also, for example: Jean Rivière, “Vote et géographie des inégalités sociales: Paris et sa petite couronne”, Métropolitiques, April 2012, (online); Luana Russo and Laurent Beauguitte, “Aggregation level matters: evidence from French electoral data”, Quality & Quantity, 48(2), 2012, 923-38; Jean Rivière, “Les divisions sociales des métropoles françaises et leurs effets électoraux”, Métropolitiques, 2014, (online); and Laurent Beauguitte and Nicolas Lambert, “L’HyperAtlas électoral parisien (2007-2012): un outil pour l’analyse des dynamiques électorales intra-urbaines”, M@ppemonde, 114(2), 2014 (online).
  • [7]
    For a recent summary, see Julien Audemard and David Gouard, “Le bureau de vote, objet et terrain d’enquêtes électorales”, Pôle Sud, 44, 2016, 5-19.
  • [8]
    Frédéric Girault, “Le vote comme expression territoriale des citadins: contribution à l’étude des segregations urbaines”, Ph.D. diss. (geography), 2000, Université de Rouen.
  • [9]
    Brigitte Le Roux and Frédéric Lebaron, “Chapitre 1. Idées-clefs de l’analyse géométrique des données” in La méthodologie de Pierre Bourdieu en action. Espace culturel, espace social et analyse des données, Paris, Dunod, 2015, pp. 1-20.
  • [10]
    Patrick Lehingue, Le vote. Approches sociologiques de l’institution et des comportements électoraux, Paris, La Découverte, 2011, 254.
  • [11]
    This disaggregation was achieved using R software by Aliette Roux (geography PhD student, spaces and societies (ESO) lab, Université de Nantes) and the script “Reapportion data from one geography to another in R” developed by Joël Gombin (political science PhD student, university research center for public action and politics, epistemology and social sciences (CURAPP-ESS lab), Université de Picardie Jules Verne.
  • [12]
    See the recent summary chapter by Florent Gougou and Nonna Mayer, “Classe, religion et vote” in Yves Déloye and Nonna Mayer (eds), Analyses électorales, Brussels, Larcier, 2017, pp. 175-219.
  • [13]
    This is a so-called reduced statistical information technique, which allows us to draw up non-intercorrelated summary variables referred to as principal components.
  • [14]
    The first design (F1-F2) of this PCA allowed us to take account of just under half the statistical information (43.9% of the variance). To help interpret and read this factorial design, a varimax rotation was used to obtain factors that were more strongly correlated with certain variables. With this rotation, axes 1 and 2 account for 22.6% and 21.3% of the variance respectively.
  • [15]
    In acknowledging that the common words we have used to describe our empirical results carry with them “a petrified philosophy of the social, always ready to spring out” (Pierre Bourdieu, Jean-Claude Chamboredon, and Jean-Claude Passeron, The Craft of Sociology: Epistemological preliminaries, trans. Richard Nice, Berlin and New York, Walter de Gruyter, 1991, 21), we should explain our use of the term “periphery” to describe the upper part of axis 2 of the PCA. Whereas the distinction between “left” and “right” is clear enough in describing the opposition between François Fillon and Jean-Luc Mélenchon on the factorial design’s horizontal axis, a word was needed for the opposite end of the vertical axis to the political “center” ground. The term “periphery”, in conveying the idea of a continuum within the electoral space in relation to the center, was favored over the word “margin”, which, aside from suggesting a discontinuity with the center, carries a significant risk of normative shifts, with an easy transition in our thinking from “margin“ to”marginality”.
  • [16]
    This rallying call between the two rounds had some logic to it–in a manner of speaking–in relation to the shifting value of the Dupont-Aignan variable on the factorial design between 2012 and 2017, which indicates that Dupont-Aignan and Le Pen had become politically closer in 2017.
  • [17]
    For a similar methodological approach (based on MCA using individual data, whereas this article uses a PCA for aggregate data) establishing an electoral space with variables of understanding, see Bruno Cautrès, Flora Chanvril, and Nonna Mayer, “Chapitre 22. Retour sur l’hypothèse de’l’homologie structurale’: les déplacements des catégories sociales dans l’espace politique français depuis La Distinction” in Philippe Coulangeon and Julien Duval (eds), Trente ans après La Distinction de Pierre Bourdieu, Paris, La Découverte, 2013, pp. 327-37.
  • [18]
    This SOG will be referred to as “MHIP” throughout the remainder of the article.
  • [19]
    Pierre Bourdieu, Distinction: A social critique of the judgement of taste, trans. Richard Nice, Cambridge, Massachusetts, Harvard University Press, 1984.
  • [20]
    Bruno Cautrès, “Le vote Mélenchon: le poids de l’héritage communiste et les limites d’une dynamique” in Pascal Perrineau (ed.), Le vote normal. Les élections présidentielle et législatives d’avril-mai-juin 2012, Paris, Presses de Sciences Po, 2012, pp. 110-31.
  • [21]
    Lehingue, Le vote, 244.
  • [22]
    Antoine Jardin, “Le vote intermittent: comment les ségrégations urbaines influencent-elles les comportements électoraux en ◊le-de-France?”, L’espace politique, 23, 2014 (online).
  • [23]
    Céline Braconnier and Nonna Mayer, Les inaudibles. Sociologie politique des précaires, Paris, Presses de Sciences Po, 2015; Jean Rivière, “Jeux d’échelles pour comprendre les déterminants sociaux de l’abstention dans l’espace français” in Déloye and Mayer (eds), Analyses électorales, pp. 59-116.
  • [24]
    Annie Collovald and Olivier Schwartz, “Entretien. Haut, bas, fragile: sociologies du populaire”, Vacarmes, 37, 2006, 50-5.
  • [25]
    For a deconstruction of the common assumptions in this regard, see Gérard Mauger and Willy Pelletier (eds), Les classes populaires et le FN. Explications de vote, Bellecombe-en-Bauges, Éditions du Croquant, 2017.
  • [26]
    HAC is an iterative technique used to produce classifications through a progressive clustering of similar data in increasingly wide classifications, giving a hierarchical representation of these data. The splits in the clusters used here give a variance decomposition with 72% inter-class variance and 28% intra-class variance.
  • [27]
    Translator’s note: “Urban hierarchy” refers to the order of French cities by population.
  • [28]
    Jean-Yves Authier and Renaud Payre, “Lyon, une ville centriste toujours à droite?”, Métropolitiques, 31 May 2017 (online), (accessed 18 September 2018).
  • [29]
    Viviane Le Hay and Gilles Pinson, “Bordeaux, capitale de la Terre du Milieu?”, Métropolitiques, 19 May 2017 (online), (accessed 18 September 2018).
  • [30]
    Rivière, “Vote et géographie”.
  • [31]
    Michel Pinçon and Monique Pinçon-Charlot, Sociologie de la bourgeoisie, Paris, La Découverte, 2007.
  • [32]
    Michel Pinçon and Monique Pinçon-Charlot, Dans les beaux quartiers, Paris, Seuil, 1989.
  • [33]
    Anne Clerval, “The spatial dynamics of gentrification in Paris: a synthesis map”, trans. Angéla Verdier, Cybergeo: European Journal of Geography, 553, 2011 (online), (accessed 18 September 2018); Éric Agrikoliansky, “Paris, 23 avril 2017: un renouveau des clivages socio-politiques?”, Métropolitiques, 5 May 2017 (online), (accessed 18 September 2018).
  • [34]
    Christophe Batardy and Jean Rivière, “Nantes, un bastion socialiste partagé entre les votes Macron et Mélenchon”, Métropolitiques, 28 April 2017 (online), (accessed 18 September 2018).
  • [35]
    See the pioneering work by Frédéric Bon and Jean-Paul Cheylan, La France qui vote, Paris, Hachette, 1988, as well as the more recent article by Joël Gombin, “Marseille, une ville coupée en quatre”, Métropolitiques, 22 May 2017 (online), (accessed 18 September 2018).
  • [36]
    Fabien Escalona, “À Grenoble, Mélenchon capte le vote’rouge-vert’?”, Métropolitiques, 8 June 2017 (online), (accessed 18 September 2018).
  • [37]
    Regarding these cities, which are sometimes described as idéopôles, see Fabien Escalona and Mathieu Vieira, “Les idéopôles, laboratoires de la recomposition de l’électorat socialiste”, Notes de la Fondation Jean-Jaurès, 2015 (online), (accessed 18 September 2018).
  • [38]
    Stéphane Cadiou and Marie-Ange Grégory, “Nice: la fidélité à droite avant tout?”, Métropolitiques, 5 May 2017 (online), (accessed 18 September 2018).
  • [39]
    Céline Braconnier, Jean-Yves Dormagen, Ghislain Gabalda, and Xavier Niel, “Sociologie de la mal-inscription et de ses conséquences sur la participation électorale”, Revue française de sociologie, 57(1), 2016, 17-44.
  • [40]
    Camille Peugny, “Pour une prise en compte des clivages au sein des classes populaires. La participation politique des ouvriers et des employés”, Revue française de science politique, 65(5), October 2015, 735-59.
  • [41]
    Julien Audemard, François Buton, and Nicolas Ferran, “(Dé)mobilisations d’entre-deux-tours: analyser les elections locales montpelliéraines à l’échelle du bureau de vote”, Pôle Sud, 1(44), 2017, 73-95.
  • [42]
    Céline Braconnier and Jean-Yves Dormagen, La démocratie de l’abstention. Aux origines de la demobilization électorale en milieu populaire, Paris, Gallimard, 2007.
  • [43]
    Mauger and Pelletier, Les classes populaires et le FN.
  • [44]
    Joël Gombin, “Contextualiser sans faire de l’espace un facteur autonome: la modélisation multiniveau comme lieu de rencontre entre sociologie et géographie électorales”, L’espace politique, 23, 2014 (online), (accessed 18 September 2018).
  • [45]
    We should note once again that the electoral choice with the strongest statistical link to the presence of laborers is abstention, with r = 0.64.
  • [46]
    Gombin, “Contextualiser sans faire de l’espace un facteur autonome”. This analytical approach, inasmuch as it seeks to identify correlations at the meso level (cities) while still measuring them at a micro level (that of polling districts within each city), is similar to the creation phase of a multilevel model, where the intercepts may vary, and where the trend lines are allowed to freely take on slopes specific to each group to which they apply.
  • [47]
    This idea has already been tested in cities in western France, working over a wider range within the urban hierarchy, with towns of 20,000 or more being taken into account. See Jean Rivière, “Comprendre les configurations électorales internes aux villes de l’Ouest” in Michel Bussi, Christophe Le Digol, and Christophe Voillot (eds), Le tableau politique de la France de l’Ouest d’André Siegfried 100 ans après. Héritages et postérités, Rennes, Presses Universitaires de Rennes, 2016, pp. 183-201.
  • [48]
    It is indeed acknowledged that, once the effects of social composition have been assessed, levels of inequality in the urban space provide one element for understanding the FN vote. See Gombin, “Contextualiser sans faire de l’espace un facteur autonome”.
  • [49]
    For a critical discussion of this divide, see Joël Gombin, “Les trois visages du vote FN”, Le Monde diplomatique, December 2015.
  • [50]
    “In the working-class arrondissements with a high proportion of immigrants (the 3rd, 13th, 14th and 15th), the correlation between the presence of a working-class population and the FN is negative. The working classes there largely come from immigrant backgrounds or share certain sociabilities and characteristics with them, especially the rejection of the far right, which is perceived as racist and dangerous. The FN vote stems rather from the native’ lower middle classes. In those arrondissements with few immigrants, on the other hand, (e.g. the 7th, 8th, and 12th), there is a positive correlation between concentrations of working-class residents and the FN vote, and the native’ workers in those areas make up the Front’s electoral base”. (Gombin, “Marseille, une ville coupée en quatre”.)
  • [51]
    With no better alternative, and conscious of the limits of using this classification, we have consulted INSEE’s guidance, which makes the following definition: “Under the terms of the definition adopted by the High Council for Integration, an immigrant is a person who is born a foreigner and abroad, and resides in France. […] Conversely, certain immigrants may have become French while others remain foreign. The foreign and immigrant populations are therefore not quite the same: an immigrant is not necessarily foreign and certain foreigners were born in France (mainly minors). Immigrant status is permanent: an individual will continue to belong to the immigrant population even if they acquire French nationality”.
  • [52]
    The French capital has regularly been the subject of intra-urban electoral analyses, including: Goguel, “Structure sociale”; Ranger, “Droite et gauche”; Rivière, “Vote et géographie”; Russo and Beauguitte, “Aggregation level matters”; Beauguitte and Lambert, “L’HyperAtlas électoral parisien”; Jardin, “Le vote intermittent”; and Agrikoliansky, “Paris, 23 avril 2017”.
  • [53]
    François de Singly and Claude Thélot, Gens du public, gens du privé. La grande différence, Paris, Dunod, 1988.
  • [54]
    Martine Berger, Les périurbains de Paris. De la ville dense à la métropole éclatée, Paris, CNRS Éditions, 2004; and Edmond Préteceille, “La ségrégation sociale a-t-elle augmenté? La métropole parisienne entre polarisation et mixité”, Sociétés contemporaines, 62(2), 2006, 69-93.
  • [55]
    The CARTELEC research program, which brings together geography and political science researchers, created a database comprising election results from 2005 through 2010 along with social indicators drawn from publicly available statistics (from INSEE, the French tax authorities [the Direction générale des Impôts], and the family allowance fund [Caisses d’allocations familiales]) at the scale of polling districts in French urban areas. The database used for this sub-section was compiled thanks to the collective work of Laurent Beauguitte, Sébastien Bourdin, Michel Bussi, Bruno Cautrès, Céline Colange, Sylviano Freire-Diaz, Anne Jadot, Jean Rivière, and Luana Russo.
  • [56]
    “Analysis is not properly contextual to the extent that the phenomena studied are not presented as a function of social interaction. Some may identify in this a sign of caution in the interpretation of the available data. As we have already noted, in the context of the ecological tradition, given that the analysis is carried out at an aggregated level, the effects of social interaction can only form the subject of hypotheses that are unsupported by direct empirical elements of proof. In this respect, however, if the contextual hypothesis may be considered as empirically unfounded, this applies no less to a hypothesis proposing that the existence of spatial variations in the correlations between demographic characteristics and votes could translate the effects of social position or truly individual evaluation mechanisms. It is not that such a hypothesis should be dismissed, however: it is possible that, if a worker from the north of France votes differently from one from the east or the south, it is simply because the category of worker relates to different statutory or positional realities depending on the area, without the interactional contexts being greatly different or at least electorally decisive. However, as the method remains the same, it cannot be considered to be better founded empirically than the contextual hypothesis“. (Julien Audemard, “De quoi le contexte est-il le nom? Critique de l’usage de la notion de contexte en sociologie électorale”, Revue française de science politique, 67(2), April 2017, 271-89, esp. 284).
  • [57]
    Audemard, “De quoi le contexte est-il le nom?”, 282.
  • [58]
    William Sidney Robinson, “Ecological correlations and the behavior of individuals”, American Sociological Review, 15(3), 1950, 351-7.
  • [59]
    Derivry and Dogan, “Religion, classe et politique en France”.
  • [60]
    Russo and Beauguitte, “Aggregation level matters”.
  • [61]
    Lehingue, Le vote, 254. Lehingue is referring here to the following quotation from Pierre Bourdieu, to be found in English in “The social space and the genesis of groups”, Theory and Society, 14(6), 1985, 736: “The inadequacies of the Marxist theory of classes, in particular its inability to explain the set of objectively observed differences, stems from the fact that, in reducing the social world to the economic field alone, it […] ignores […] all the oppositions that structure the social field, which are irreducible to the opposition between owners and non-owners of the means of economic production. It thereby secures a one-dimensional social world, simply organized around the opposition between two blocs”. I take this opportunity to thank Patrick Lehingue for his careful proofreading of this article.
Jean Rivière
Jean Rivière is assistant professor at the Institute of Geography and Regional Planning of the Université de Nantes (IGARUN). Part of the university’s Spaces and Societies team (CNRS unit no. UMR ESO 6590), his social geography research primarily investigates the spatial dimension of power inequalities and relationships. His work describes the ongoing sociological developments in urban and periurban areas, along with the electoral behaviors that are expressed there. He recently coordinated two sets of thematic papers: “Élections présidentielles: les votes des grandes villes françaises au microscope”, Métropolitiques, 2017, and “Géographie et sociologie électorales: duel ou duo?”, L’espace politique, 2014. More widely, he works to build bridges between geography, sociology, and political science in the domain of electoral analysis. (IGARUN, Université de Nantes, Campus du Tertre, BP 81 227, 44312 Nantes cedex 3.
This is the latest publication of the author on cairn.
Uploaded on on 25/07/2019
Distribution électronique pour Presses de Sciences Po © Presses de Sciences Po. Tous droits réservés pour tous pays. Il est interdit, sauf accord préalable et écrit de l’éditeur, de reproduire (notamment par photocopie) partiellement ou totalement le présent article, de le stocker dans une banque de données ou de le communiquer au public sous quelque forme et de quelque manière que ce soit.
Loading... Please wait