1In just a few years, primary elections have become established in French politics as a legitimate way of selecting candidates for the presidential election.  The November 2016 primary of the right and center, the fruit of several internal power struggles, was largely inspired by the primary election held in October 2011 by the Socialist Party (PS) and the Radical Party of the Left (PRG).  Like the 2011 primary election, the one in 2016 was organized like a typical open primary.  Held on 20 and 27 November 2016, it was a success in terms of mobilization, with more than four million voters in each of the two rounds. The election resulted in a very clear victory for François Fillon, who received 44.1% of the votes cast in the first round and almost two thirds in the second.
2Primary elections can be seen as the driving force of a more direct form of democracy and a renewed link between voters and political parties, but also, paradoxically, as symptomatic of a “democracy of opinion” that provides further confirmation of the transformation of partisan organizations into machines designed to tap into the aspirations of a supportive electorate.  Only a few empirical studies provide evidence for such debates. Studies that focus specifically on the sociology of primary voters are even scarcer. However, analyzing these voters’ profiles, their relationships to the election, and the factors that affect their choices is vital if we are to grasp what is really at stake in this type of election. With that in mind, we carried out questionnaire-based exit poll surveys at seven sites in five French départements during the primary of the right and center in November 2016. The survey was an attempt to investigate the social and political factors that mobilized primary voters, and to examine how those factors affected the choices made during the election.
Analyzing the Primary through the Sociology of Voters
3French debates on primaries, which focus on questions similar to those raised by international studies, have until now had little support from empirical research on mobilization and electoral choice.  As American studies have shown, electoral sociology is a major part of the discussion about the effects of primaries on political life and its actors. In this article, we would like to answer two main questions.
4The first line of questioning is to ask who primary voters are. So far, researchers have not provided unanimous answers. Some American studies posit a significant difference between the entire potential electorate and those who vote in primaries, who tend to be older and more educated,  while others emphasize the absence of major sociological divides in a country where, let us not forget, participation in most elections is particularly weak in working-class neighborhoods.  The Anglophone literature also expands the question to include that of the ideological representativeness of primary voters. Here too, while some studies do not identify anything particularly distinctive about these voters,  others show that primary voters are more radical than the average voter, but still more moderate than the members and activists of the party or parties holding the elections.  In France, localized studies of the left primaries in October 2011 provided some answers: they point to the lack of mobilization among significant electoral segments—young people, working-class neighborhoods—and show that opening the primary to allow people who are not party members to participate has encouraged the (relative) proliferation of left-wing (Arnaud Montebourg) and right-wing (Manuel Valls) candidates in the PS. 
5The primary of the right and center clearly lends itself to this type of questioning, especially as numerous commentators have maintained that the traditional electorate of the governing right has been infiltrated by groups from other parties—whether from the left, in order to support the candidature of Alain Juppé for strategic reasons, or from the extreme right, attracted by the conservative programs of François Fillon or Nicolas Sarkozy. Moreover, looking beyond what unites the various primary participants, should it not be possible to identify different political and sociological profiles within the electorate?
6The second line of questioning concerns the role of party networks in the dynamics of mobilization. Do different levels of proximity between voters and party networks affect voting habits? Surveys carried out to date in France show that local party networks should indeed be taken into account when explaining how people vote at primaries. Here, awareness of the characteristics of the regions where the vote is held seems to be the determining factor. Candidates seem to achieve their best results in their main supporters’ home turfs.  Moreover, in the context of the primary of the right and center, the role of activism is a question worth pondering for a party that has, in contrast to many left-wing parties,  never been known for it, especially at a time when existing practices are already being disrupted by internet usage. 
7We will attempt to use these two lines of questioning to investigate how the dynamics of mobilization at work during the November 2016 primary may have impacted electoral choices. Several localized surveys will be used to help us answer these questions.
A Localized, Multisite Survey
8To gather data about these voters, we conducted a questionnaire-based exit poll survey at several polling stations. The questionnaire contained questions about politicization, political position, the respondent’s electoral choices at the primary and at the presidential election in 2012, the amount of contact the respondent had had with partisan networks, and to what extent the respondent had followed the electoral campaign in the media. We also included various questions about the socioeconomic and sociocultural characteristics of the voters.
9We selected seven primary polling station sites  according to three main criteria: the sociodemographic characteristics of local residents, voting results at previous elections, and which of the primary candidates received local support (Tables 1 and 2). 
Sociodemographic Characteristics of the Selected Polling Stations
|Polling station||Commune||Département||Proportion of population over 65||Proportion of workforce employed in non-managerial office work or manual labor||Proportion of university graduates||Proportion of households having lived in the area for more than ten years|
|540203, 540204, 540205, 540206||Nancy||Meurthe-et-Moselle||15.5%||10.3%||51.7%||32.1%|
Sociodemographic Characteristics of the Selected Polling StationsThe figures were calculated using the data for the 2013 IRIS units corresponding to the primary polling stations. The proportion of the population over 65 was calculated in relation to the total population of the relevant IRIS units. The proportion of the workforce employed in non-managerial office work or manual labor (“employés et ouvriers”) was calculated in relation to the total workforce of people over 15. The proportion of university graduates was calculated in relation to the number of people over 15 not in education. The proportion of households having lived in the area for more than ten years was calculated in relation to the total number of households.
Electoral History of the Selected Polling Stations
|Polling station||Nicolas Sarkozy 1st round 2012||Participation in the 1st round of the primary||François Fillon 1st round of the primary||Juppé 1st round of the primary||Nicolas Sarkozy 1st round of the primary||Participation in the 2nd round of the primary||François Fillon 2nd round of the primary||François Fillon 1st round of the presidential election 2017|
Electoral History of the Selected Polling StationsParticipation rates calculated in relation to the number of registered voters in 2017.
10The questionnaire was distributed at all seven sites during the first round of the primary, at five of the sites during the second round (Avignon, Castelnau-le-Lez, La Grande-Motte, Nancy, and Nantes), and at two of the sites during the first round of the presidential election (Avignon and La Grande-Motte).  The decision to expand the survey to include the presidential election gave us the opportunity to compare primary voters with people who voted in the presidential election, in particular those who voted for François Fillon. This enabled us to consolidate our observations about the former group.
11The variety of the surveyed locations, which were selected from among a number of rightleaning areas, provides a sufficiently numerous and diverse population to ensure that our findings can be at least somewhat generalized. Furthermore, the inclusion of a spatial variable means we can integrate certain differences, such as the sociopolitical characteristics of the different locations or the role played by local party networks. There is considerable sociological and political diversity among the sites chosen. In 2012, Nicolas Sarkozy fared worse than his national average (27.2%) in just two of the polling districts: Avignon and Roubaix. The three polling districts where support for the governing right was highest, Castelnau-le-Lez, La Grande-Motte, and Nantes, are all extremely dissimilar. Located in the outskirts of Montpellier, Castelnau-le-Lez is a rapidly expanding town. The area covered by the “primary” polling station is inhabited by a small middle class alongside people of higher socioeconomic status. The town mayor is a member of the Republicans (LR) and has long been known for his anti-Sarkozy attitude. He was one of Alain Juppé’s sponsors during the primary. La Grande-Motte, a seaside resort on the border between Hérault and Gard, is inhabited predominantly by middle-class retired people. Its mayor is close to Fillon and was one of his sponsors. It was there that Fillon held his last campaign rally before the first round. Although the governing right usually achieves very good results there, the Front National (FN) also received a large proportion of the vote. Finally, the polling district in the center of Nantes is home to a significant number of bourgeois intellectuals. Political leaders in Nantes have tended to support Juppé.
12Also located in the city center, the Nancy polling district is similar to the one in Nantes: its inhabitants are predominantly young and educated; compared to the FN, the right and center received a relatively large share of the vote there. In Nancy, a traditional stronghold of the Radical Party where the prominent elected representatives are centrists rather than members of the Republicans, most local political leaders supported Juppé. The Béziers district is its mirror image. Located in a town in demographic decline, its inhabitants are mostly elderly working-class people. The Republicans received a large share of the vote there, although, as in Avignon, they had to compete with the FN’s strong performance. Élie Aboud, Deputy for the 6th electoral district of Hérault in 2016, is a friend of Jean-François Copé and was one of his sponsors during the primary. The Avignon and Roubaix districts  are largely made up of working-class neighborhoods. In both districts, however, the population is younger and less right-wing than that of Béziers: the left received the majority of votes in both.
13The sites’ electoral results reflect these different sociological profiles. During the primary, participation was particularly high in the polling stations serving older and more educated populations, as well as, unsurprisingly, in areas where support for the governing right was higher. Juppé obtained his best results in places with the lowest levels of support for the governing right, in other words in the Nancy polling station. Sarkozy fared best in rightleaning working-class districts with a small middle class, a constituency that he often has to share with the FN, as in Béziers and La Grande-Motte.  Fillon’s results in the primary followed the opposite pattern to Juppé’s: he did well everywhere, but tended to do less well in the most working-class areas (Table 2).
14Despite variable response rates at the different sites and during different election rounds, our method produced samples large enough to enable meaningful statistical analysis. We obtained three samples during the three questionnaire phases: one sample for the first round of the primary (N = 3,015, response rate (RR) = 43.8%); one sample for the second round of the primary (N = 1,832, RR = 33.3%); and one sample for the first round of the presidential election (N = 1,433, RR = 36.5%).  Alongside the questionnaire responses and the data from the districts surveyed, we also obtained the electoral rolls for the polling stations serving the seven sites. This meant we could compare the age and sex structures of the actual electorates with the age and sex structures of our samples.
The Primary, Mediator in the Power Relationship between Three Electorates
15To answer our research questions, we subjected our data to two multiple correspondence analyses (MCA).  For the primary elections, often presented as the mobilization of sociologically homogenous units, we treated the political data (political positions, voting decisions, proximity to the party, level of politicization) as the active variables, while the sociological variables are used as illustrative variables (Graphs 1 and 2). Based on the structure of the two axes, we can then identify three main voter groups distinguished by their different mobilization characteristics.
MCA, First Round of the Primary
MCA, First Round of the Primary
MCA, 2nd Round of the Primary
MCA, 2nd Round of the Primary
Forms of Politicization and Their Relationship to Partisan Bodies
16Our multiple correspondence analyses revealed two dimensions that structure the samples from the first and second rounds in a similar way. On the horizontal axis, the first dimension is ordered according to the respondents’ degree of proximity to the party. It goes from the core LR electorate (members or former members of the parties of the governing right, such as the UMP [Union for a Popular Movement], the RPR [Rally for the Republic], or the UDF [Union for French Democracy]; voters who expressed support for the party) at one end to the groups with the least contact with the party (supporters of the left and people who voted for the left) at the other. The projection of socioeconomic variables shows, unsurprisingly, that this dimension opposes specific categories of voters. On the right-hand side of the graph, clustered around the LR hard core, are older voters, people who work in the private sector or did so before retirement, and people whose children attended private school for at least some of their education. In contrast, the halo of left-wing voters contains people who are younger, more likely to be government employees, and whose children have never attended private school. This first dimension is responsible for the greatest amount of inertia (8.98%).
17On the vertical axis, the second dimension opposes the most politicized voters (party members, voters who expressed a very strong interest in politics, voters who said they had had contact with activists during the campaign) at the bottom of the graph to the least politicized voters (voters who said they had no particular political allegiance, voters who expressed a low level of interest in politics) at the top of the graph. The projection of socioeconomic variables reveals a band of more politicized voters containing the oldest voters and those with the highest income (in the bottom-right section of the graph), but also government employees (in the bottom-left section of the graph). The least politicized voters, on the other hand, included people living in rented accommodation, and more generally the poorest and youngest voters. This second dimension is less clearly structured than the first because the vast majority of voters surveyed were unusually highly politicized.
18More specifically, 90% of voters during the first round said they were interested in politics.  Our study shows that the general profile of people who voted in the primary of the right and center is in line with the conclusions of numerous previous studies regarding the social unrepresentativeness of voters at this type of election. It also confirms the findings of several pre-election polls:  50-75-year-olds made up half of the sample at the first round of the primary, despite representing just slightly more than a third of voters registered in the electoral rolls of the regular (as opposed to primary) polling stations involved in our study. At the polling stations where we were able to specify voter ages more precisely,  we found that voters aged between 50 and 75 were six times more likely to vote in the election than voters under 35. In short, retired people exerted an outsize influence on the outcome of the primary.
19Primary voters are also distinguished from the rest of the electorate by their socioeconomic characteristics. Based on a whole range of socioeconomic traits, it is clear that the participants in the study are economically better off than the population as a whole. Although some differences—such as whether or not voters own their primary residence—may be partly due to variation in mobilization patterns between different age groups, these generational contrasts only make the variation in terms of socio-professional category or education level even more striking.
20The participants in our study have another significant characteristic: the prominence of Catholic voters. In our sample from the second round, only a quarter of respondents described themselves as non-religious. A clear majority said they were Catholic, with only 6% of believers belonging to other religions. All levels of practice were equally well represented: 19% of respondents were regularly practicing Catholics, 26% were occasionally practicing Catholics, and 25% were non-practicing Catholics.  The fact that 87% of voters who completed the questionnaire during the first round said that they routinely participated in all elections  is surely the result of both high levels of politicization—with a number of different root causes—and the way that the norm of participation is more likely to be thoroughly internalized by older people, particularly in areas where Catholic culture continues to be influential. 
21The composition of the sample from the second round of the primary shows signs of a minor broadening of the electoral base in terms of both age and socioeconomic characteristics of voters—as happens in any election where participation increases from one round to the next, as it did at most of the polling stations surveyed. Moreover, a comparison of primary voters with the profile of voters who actually voted for the right, and particularly for Fillon, in the first round of the presidential election provides more confirmation that primary voters are not representative of the electorate as a whole. The specific features of the social profile of primary voters, therefore, cannot be due solely to a social structure rooted in the socioeconomic characteristics of the right-wing electorate.  From this perspective, the way the election was organized, in terms of the geographical distribution of polling stations, contributed to the quantitative over-representation of one specific electoral target group. 
22Nevertheless, the second dimension of our MCAs adds an important nuance to this observation. This becomes clear when we consider the second dimension alongside the first, which represents the proximity of voters to LR. The intersection of the two dimensions reveals three electoral blocs distinguished not just by party—or ideological—identification, but also by different forms of politicization. The first bloc is the LR hard core, made up of members and former members of the former parties of the governing right and an inner circle of supporters. This first group of very highly politicized voters seems to be united more by proximity to partisan networks than by the sociocultural traits of its members. These members, who represent about a third of the total sample, are characterized by the greater frequency with which they claim to have discussed the election with their friends and family, the greater frequency of their encounters with party activists, or the greater probability compared to other voters that members of their family belong to the party. On the other hand, they tend to be less educated than the rest of the sample. Within the group, party members (about 8% of the total sample) align most fully with these characteristics, as their position on the graph shows.
23The second bloc contains voters who said they were “supporters of the right”. They were the largest group at just over a third of the sample. They have less contact with party networks and are less politicized than the first bloc, which may be due to lower levels of political investment, or at least to weaker party identification. In contrast, although this bloc is undoubtedly the most representative of the right-wing electorate as a whole, it is still a fringe group with unusually strong ties to politics. The high level of politicization of this group compared to the wider electorate is without doubt connected to the social characteristics of its members, who are slightly younger, and also on average more educated, than voters in the first bloc. These traits are certainly responsible not just for the ideological stance of this group of voters, but also for the fact that they are less susceptible to the activities of partisan networks.
24Finally, the third bloc consists of the voters with the least connection to LR, or in other words, voters who said they supported the center or the left. They are just as politicized as LR supporters, probably because of their sociocultural characteristics—as a group, they are extremely well educated. Their other characteristics differ significantly from those of the voters in the first two blocs. Two variables are relevant here: their relationship to religious belief and their relationship to the public sector. While a large majority of voters in the first two blocs said they were Catholic, most in the last bloc said they were not religious. These voters also tend to have close ties to the public sector: few of those who had children had sent them to private schools, while in the first two blocs most children were sent to private schools. Similarly, a significant number of this group are employed in the public sector, while among voters with stronger ties to LR the proportion of public sector employees is generally small. On that basis, we assume that these voters are mobilized in the main by their very high level of politicization. This is especially the case for the left-wing voters for whom the election was not a priori intended. Among supporters of the left, 5% said they were a member of a political party. This remarkably high proportion provides further evidence that a significant number of left-wing voters participated in the primary for strategic reasons. 
25This division of voters into three blocs is important insofar as each of the blocs made specific choices during the primary. The final result of the election was, in a sense, the product of those choices, and of all the strategies implemented by the different candidates in order to mobilize one or another of the electoral blocs.
From Specific Forms of Politicization to Voting Decision-Making Processes
26The votes cast in the first round seem to follow clearly identifiable political patterns. Votes cast for Nicolas Sarkozy are primarily determined by the first dimension and correlate closely with the LR hard core. Votes for Sarkozy become less frequent the further one goes from the cluster of LR members, the only category of voters among whom the former President of the Republic outdid his competitors. Beyond this small group of party members, Sarkozy also did relatively well (receiving more than double his average share of the vote) among party supporters. He fared significantly worse among supporters of the right (12%), and even less well among supporters of the center or left (1-2%). Table 3 illustrates this continuum. Incidentally, proximity to party networks (an LR member in the family, contact with LR activists during the campaign) increased the probability of voting for Sarkozy in the first three categories (LR members, LR supporters, supporters of the right). Being a former member of the parties of the governing right did not have a similar effect, which shows that it was above all the party networks that were currently active at the time of the election that benefited the former head of state.
Votes Cast During the First Round of the Primary Correlated with Proximity to LR
|N. Sarkozy||F. Fillon||A. Juppé|
|LR members (N = 209)||42.1%||35.9%||16.7%|
|LR supporters (N = 652)||36%||48.1%||12.7%|
|Supporters of the right (N = 932)||11.7%||62%||19.9%|
|Supporters of the center (N = 360)||1.9%||34.4%||52.2%|
|Supporters of the left (N = 204)||1.5%||18.1%||66.2%|
|Total sample (N = 2613)||18.2%||47.8%||26.6%|
|Official results in the polling stations surveyed|
Votes Cast During the First Round of the Primary Correlated with Proximity to LRData obtained from the survey conducted during the first round of the primary. The headcounts refer to respondents who indicated how they voted during the first round of the primary (excluding blank and null votes). The table should be read horizontally. For example, 42.1% of LR members said they voted for Sarkozy in the first round of the primary.
27Votes cast for Alain Juppé, on the contrary, are located at the other end of the first axis. Voting for Juppé is highly correlated with voters saying they were supporters of the left. Indeed, almost a third of people who voted for Juppé during the first round of the primary said they were supporters of the left, and almost half said they had voted for François Hollande in the second round of the presidential election in 2012. Nathalie Kosciusko-Morizet’s electorate was the only other category to include such a high proportion of leftwing voters. Only 8% of people who voted for the five other candidates had voted for Hollande in the second round of the presidential election of 2012. These left-wing voters are highly politicized, and it can be assumed that at least some of them were motivated to vote by strong strategic or ideological reasons. These are who Martial Foucault meant when he referred to “infiltrated” voters. 
28Nevertheless, the mobilization of left-wing voters alone is not enough to explain the sociological positioning of the votes cast for Juppé. After controlling for these voters, people who voted for Juppé are still generally more educated, younger, and have stronger ties to the public sector than the electorate as a whole. This is first and foremost due to centrist voters. Juppé’s electorate, like Kosciusko-Morizet’s, contains three times more supporters of the center than the electorates of other candidates. In our sample, this category of voters has the highest education level (82.2% have a higher education qualification). But even after controlling for these voters, supporters of the right who voted for Juppé turn out to have similar social characteristics to supporters of the center in terms of age, education level, and ties to the public sector.
29Fillon’s electorate is quite different. Unlike Sarkozy’s electorate, which is partly defined by the votes cast by the hard core of LR activists and members, its position in the first dimension is basically that of a typical right-wing electorate. Fillon was able to appeal to a very broad electorate—right-leaning, of course, but extending well beyond the party’s direct sphere of influence (Table 3). Fillon’s victory over his competitors was more convincing among supporters of the right in general than among voters who were active in party networks, although he also did well among LR supporters.
30Voters who described themselves as Catholic are located close to the right-wing electorates of Sarkozy and Fillon, with no distinction between different levels of practice (regular, occasional, non-practicing).  Voters who described themselves as non-religious are located on the other side of the axis, closer to supporters of the center or left and people who voted for Juppé.  During the second round, these oppositions were even more pronounced. Juppé’s electorate moved a little further away from the groups closest to LR, assumingly because of the very slight increase in the number of supporters of the left and center who voted in that round.
31In summary, the votes cast for Fillon are in the center of the sociological mobilization at the primary. Although his unexpected victory was in part thanks to the confrontation between his two principal rivals, of which he was the main beneficiary,  it would be difficult to argue that he deserves no credit for his own effective strategy. Drawing on the lessons of the Socialist experience in 2011, and no doubt also aware of pre-election polls describing potential voters for the primary of the right as older, more religious, wealthier, and more radical in their views on the economy and identity than the average right-wing voter,  Fillon and his campaign team implemented mobilization strategies carefully tailored to that specific electorate. Alongside appeals to Catholic voters—a constant feature of all the candidates’ campaigns —Fillon also decided to use techniques designed to win over other segments of the electorate with high mobilization potential, including retired people and hunters.  Moreover, Fillon’s campaign budget was allocated in a way that reflected the actual media consumption habits of primary voters. Although his budget (1.22 million euros) was one of the highest of the campaign, just behind Juppé’s, Fillon spent the least of all the candidates on the budget item devoted to internet pages and telematics services (7,629 euros). In total, his expenditure on that item came to less than 1% of his budget, compared to an average of 8% of the other candidates’ budgets.  Fillon’s team prioritized more traditional campaign expenses, like organizing public meetings or phone calls.  Ultimately, it seems that by focusing its efforts on media tools selected with baby boomers’ habits in mind, Fillon’s campaign team was able to reach the largest number of voters.
32Indeed, 80% of voters said they had followed the campaign on television, four times more than the number of people who said they had visited one of the candidates’ websites and ten times more than the number of people who had “liked” a candidate’s Facebook account. In that respect, even among party members, voter habits during the primary did not deviate from the norm seen during other elections in France, and particularly during presidential elections.  The clear pre-eminence of television among media sources during the primary is due in large part to the unusually high average age of voters. While television consumption levels are very stable across different age groups, internet usage varies widely. Voters under 35 were twice as likely to have consulted the website of at least one candidate than voters between 50 and 75.
The Role of Local Political Context
33At this stage in the analysis, we can now understand Sarkozy and his political supporters’ hesitations about holding an open primary: it seems he would have fared better in a primary in which only party members were eligible to vote.  Was Sarkozy’s heavy lead among party members simply due to ideological considerations that, for example, induced members to adopt a more radical political line, or was it rather to do with forms of party loyalty that can be undermined by other forms of local support? In other words, did Sarkozy dominate the party-member vote equally in all areas, or did local party networks influence the voting decisions not just of the wider electorate, but also of party members?
34The question is worth asking, because networks of political influence during the campaign were more relevant among circles of acquaintances—family, friends, or neighbors—than among the deterritorialized social networks of the internet. While almost all voters (97%) said they had discussed the primary with members of their family, friends, or even activists, barely a quarter said they had looked at the website, Facebook account, or Twitter page of at least one candidate. More generally, contact with activists played an important role during the campaign. Over a third of respondents (36%) said they had had contact with activists, either at rallies or public meetings (10%), when being given leaflets (5%), or above all in more informal encounters (25%).
35These proportions were higher in areas where the party had a stronger presence. La Grande-Motte provides a good example of this phenomenon. In this town with fewer than 9,000 inhabitants and more than 300 LR members,  18.3% of voters were LR members—three times more than at other locations—while another 31% of voters were LR supporters (compared to 24% at other locations). It is hardly surprising, then, that more than half of the voters at this polling station said they had come into contact with LR activists during the campaign.
36Although it might be expected that the presence of so many party members in La Grande-Motte would result in a clear lead for Sarkozy, Fillon actually far outperformed him in the first round. There is every reason to conclude that party networks worked in favor of the winner of the primary, even among party members. The LR mayor of La Grande-Motte was a fervent supporter of Fillon’s candidacy, and throughout the campaign Fillon and Sarkozy were neck and neck among party members (43% each) in the town. In Avignon, on the other hand, where local politicians did not express support for any of the candidates, Sarkozy enjoyed widespread support from activists and dominated his rivals among party members (53% voted for him, compared to 18% both for Fillon and Juppé). However, when respondents were invited to express their opinion on a range of economic and social issues in the questionnaires handed out during the first round of the presidential election, there was nothing in their answers to suggest any ideological differences between party members in the two towns. It seems likely that the efficacy of party networks in La Grande-Motte was due not just to the density of the networks themselves, but also to the residential stability of the most mobilized voters: retired people, who tend to live in areas where most people know each other.
37Fillon did not just perform very well in places where he benefited directly from the political support of an elected official, as in La Grande-Motte. He also did well in the Nantes polling district; indeed, it was there that he achieved his most crushing victory over his two main rivals (Table 2). He received 57% of the votes cast in the first round, three times more than Juppé and five times more than Sarkozy. Although the gap between Fillon and Sarkozy was reduced among party members (52.4% to 23.8%), his margin of victory was still considerable even there. It is tempting to see this as the effect of the favorable local political climate in the Pays-de-la-Loire region, where Fillon had previously been president of the Regional Council  and where he benefited not just from being a well-known political figure, but also from the help of numerous political supporters.
38* * *
39Although primaries do constitute a significant expansion of the right to participate in the selection of candidates, this expansion must once more be put into perspective in light of these new results. The sociological analysis of voters at the primary of the right and center confirms the exclusionary nature of primaries as a political practice. Whether or not a person votes in a primary is, in effect, largely determined by their sense of their own political competence, which itself is correlated with an interest in politics and a relatively high socioeconomic status.
40Nevertheless, as well as providing evidence for the opening of the election to a specific section of the left-wing electorate—which, incidentally, only confirms the typical sociological profile of “selectors” —the analysis of the electorates and their relationships to the campaign reveals diverse forms of politicization that determined how the different electorates engaged with the election. The Sarkozy, Juppé, and Fillon electorates are all highly politicized, but the former is characterized by politicization through contact with party networks, while politicization in the latter two seems rather to be closely linked to sociological characteristics, and in particular a high level of education. The evidence suggests that the primary election represented a tension between these two types of politicization, with the second ultimately prevailing.
41Fillon’s victory at the primary seems to have been driven by a double logic, both sociological and political: he managed to win over the majority of voters who were motivated to participate by particular sociological characteristics and who were much more deeply rooted in right-wing environments than people who voted for Juppé. He did so thanks to a discourse that resonated almost flawlessly with the policy expectations of the most politicized segments of an electorate which, despite the vagaries of politics, continue to maintain a certain level of party loyalty.
42However, the specific characteristics of electoral mobilization at the primary in November 2016 became an additional challenge to overcome for the LR candidate for the presidential election. More precisely, two problems presented themselves following the primary. First, how to mobilize voters who had not participated in the primary election, and second, how to mobilize centrist voters, who overwhelmingly voted for Fillon’s main rival during the primary. It seems clear that Fillon failed to meet this challenge.  According to our survey results from the first round of the presidential election, the LR candidate was not able to rally either the right-wing voters who did not participate in the primary, or the centrist voters. Among people who voted for Sarkozy in the second round of the presidential election in 2012, 77% of those who had voted for Fillon in the primary also voted for him in the first round of the presidential election in 2017. That percentage falls to 33% of right-wing voters who voted for Juppé during the primary. Although we cannot ignore the effect of the scandals that hit the LR candidate during the presidential campaign, Fillon’s failure is also undoubtedly due in part to the electoral dynamics produced by the decision to hold a primary. Indeed, by encouraging Fillon to radicalize his discourse and platform, the structure of the primary of the right and center opened up a new electoral niche for Emmanuel Macron: the same Macron who, along with Le Pen, benefited the most from the defection of Sarkozy’s former voters—regardless of whether or not they had voted in the primary—in the first round of the presidential election in 2012.
Rémi Lefebvre and Éric Treille, “Introduction: vers une primarisation de la vie politique française” in Rémi Lefebvre and Éric Treille (eds), Les primaires ouvertes en France. Adoption, codification, mobilisation, Rennes, Presses Universitaires de Rennes, 2016, pp. 11-40.
Initially supported by Charles Pasqua, the idea of a primary of the right really dates back to the end of the 1980s. See Florence Haegel, “La primaire à l’UMP: genèse et enjeux”, Pouvoirs, 154(3), 2015, 89-98.
On the legal and organizational aspects see Anne Levade, “Le droit des primaires: règles, contrôle, finances, sanctions”, Pouvoirs, 154(3), 2015, 99-109.
On the consequences of primaries in terms of democratic functioning, see several articles in the special issue of Pouvoirs devoted to primaries in 2015, particularly Yves Mény, “Primaires: vertus (apparentes) et vices (cachés) d’une greffe américaine”, 27-40; Rémi Lefebvre, “Les primaires: triomphe de la démocratie d’opinion?”, 111-23; and Pierre Avril, “Les primaires: un affaiblissement de la démocratie?”, 133-42.
For a review of this research, see Lefebvre and Treille (eds), Les primaires ouvertes en France.
John G. Geer, “Assessing the representativeness of electorates in presidential campaigns”, American Journal of Political Science, 32(4), 1988, 929-45.
Andrew de Nitto and William Smithers, “The representativeness of the direct primary: a further test of V.O. Key’s thesis”, Polity, 5(2), 1972, 209-24; David W. Moore and C. Richard Hofstetter, “The representativeness of primary electorates: Ohio, 1968”, Polity, 6(2), 1973, 197-222.
Austin Ranney, “The representativeness of primary electorates”, Midwest Journal of Political Science, 12(2), 1968, 224-38; Austin Ranney, “Turnout and representation in presidential primary elections”, American Political Science Review, 66(1), 1972, 21-37; Austin Ranney and Leon Epstein, “The two electorates: voters and non-voters in a Wisconsin primary”, Journal of Politics, 28(3), 1966, 598-616.
James Adams and Samuel Merrill III, “Candidate and party strategies in two-stage elections beginning with a primary”, American Political Science Review, 52(2), 2008, 344-59; Karen M. Kaufmann, James G. Gimpel, and Adam H. Hoffman, “A promise fulfilled? open primaries and representation”, Journal of Politics, 65(2), 2003, 457-76; Barbara Norrander, “Ideological representativeness of presidential primary voters”, American Journal of Political Science, 33(3), 1989, 570-87.
Julien Audemard and David Gouard, “Les primaires citoyennes d’octobre 2011: entre logique censitaire et influences partisanes locales”, Revue française de science politique, 64(5), October 2014, 955-72. The electorate at primaries has also been observed to be more educated and more politicized in Italy. See the special issue “Le primarie in Italia” in Osservatorio elettorale della regiona Toscana, 55, 2006.
Audemard and Gouard, “Les primaires citoyennes d’octobre 2011”; Laurent Olivier, “La territorialisation des primaires socialistes en Meurthe-et-Moselle. Appropriations locales et traces temporelles d’une procédure nationale” in Lefebvre and Treille (eds), Les primaires ouvertes en France, pp. 239-54.
Florence Haegel, “La mobilisation partisane de droite: Les logiques organisationnelles et sociales d’adhésion à l’UMP”, Revue française de science politique, 59(1), February 2009, 7-27.
Anaïs Theviot, “Les primaires: terrain d’expérimentation de l’innovation politique? Le cas de la campagne d’A. Juppé en 2016: une mobilisation scientifique orchestrée par les data” in Lefebvre and Treille (eds), Les primaires ouvertes en France, pp. 213-33.
Some sites actually correspond to several primary polling stations grouped together. In some cases, the site configuration made it impossible to distinguish between the voters of different polling stations on election day. In such cases, we conducted the survey indiscriminately. The survey ultimately covers thirteen primary polling stations corresponding to about sixty regular polling stations with a total of 46,063 registered voters.
We had to perform several different processes to obtain data about these primary polling stations, which were formed by grouping traditional polling stations together. After having identified the boundaries of the different sites, we overlaid them on INSEE’s IRIS units to work out the intersecting boundaries. We then used these boundaries to calculate the coefficients for disaggregating the IRIS data at the scale of the primary polling stations.
Our survey was subject to the usual vagaries of exit polls carried out with scarce material resources. On this topic, see issue 44 of the Pôle Sud journal, and in particular Lorenzo Barrault-Stella and Nazli Nozarian, “La mobilisation d’étudiants dans une enquête autour de bureaux de vote: Produire des matériaux quantitatifs dans un contexte de sous-financement de la recherche”, Pôle Sud, 44(1), 2016, 21-34; and Laura Giraud and Christine Pina, “C’est ici qu’on vote? Ou les leçons inattendues d’un questionnaire sortie des urnes”, Pôle Sud, 44(1), 2016, 35-47. We would like to take this opportunity to thank the pollsters who helped us, in particular the students of economic and social administration at the University of Avignon especially Louis Gérard-Bendelé the students of Sciences Po Lille, the University of Lorraine, the department of political science at the University of Montpellier, and finally Marie Neihouser, who coordinated the survey at the Béziers polling station.
The Republican mayor of Roubaix declared his support for Juppé during the primary campaign.
Martial Foucault observed that Front National supporters were most likely to express an intention to vote for Sarkozy at this primary. Martial Foucault, “Les électeurs infiltrés peuvent-ils menacer le résultat de la primaire?”, Note du Cevipof, November 2016.
Although the three questionnaire phases were conducted at the same locations, they do not constitute a panel data set because not all the voters participated in each sample.
The MCAs were carried out using the R software environment and the FactoMineR package. Values were imputed for missing data using the missMDA package. These values did not affect how the dimensions were constructed. Although it can lead to overestimation of the inertia percentages, the imputation of values for missing data makes it easier to read the connections between the observed data.
Our data enable only an imprecise measure of this level of politicization. In particular, we do not have access to markers that could be used to evaluate voter competence. We rely, therefore, on the subjective aspect of politicization; in other words, voters’ capacity to declare their own interest in politics, which reflects how selfempowered they are to express their views on related subjects. On this topic see, among others, Loïc Blondiaux, “Faut-il se débarrasser de la notion de compétence politique? Retour critique sur un concept classique de la science politique”, Revue française de science politique, 57(6), December 2007, 759-74.
Jérôme Jaffré, “Le corps électoral déformé de la primaire de la droite” in “L’enquête électorale française”, Note du Cevipof, 11, round 2, February 2016; Jérôme Fourquet and Hervé le Bras, La guerre des trois. La primaire de la droite et du centre, Paris, Fondation Jean-Jaurés, 2017, 77-83.
At two sites (Castelnau-le-Lez and La Grande-Motte), while we were handing out the questionnaires we were able to record the age and sex of voters who declined to respond to our survey. These observations enabled us to reconstruct the actual age and sex structure of the electorates of these two polling stations. At one of the sites (Avignon), we were able to procure the voter log used at the primary, giving us an even more precise understanding of this structure.
For reference, most surveys and studies of religious belief in France estimate that those who are non-religious constitute between a third and a half of the French population. Obviously, it may be the case that this proportion is genuinely smaller in the places we surveyed. However, our observation that Catholics made up a large proportion of primary voters is confirmed by other analyses. See Claude Dargent, “Les catholiques entre les primaires de la droite et le vote à la présidentielle”, Note du Cevipof, 29, round 9, February 2017.
We should be cautious when interpreting this figure, based as it is on self-declaration.
Pierre Bréchon, “L’abstention: de puissants effets de génération?” in Anne Muxel, La politique au fil de l’âge, Paris, Presses de Sciences Po, 2011, pp. 91-111.
In “L’enquête électorale française”, Jérôme Jaffré pointed out that comparing the social profile of voters who planned to vote in the primary with that of voters who voted for the right during the regional elections of 2015, rather than with the electorate as a whole, does not diminish the sociological contrast. Jaffré, “Le corps électoral déformé de la primaire de la droite”.
See the Fondation Terra Nova note, “Primaires de la droite et du centre: l’effet bocal”, 25 January 2017.
People who voted in the primary first had to sign a charter saying, “I share the Republican values of the right and center and am committed to achieving a change in power in order to aid France’s recovery”.
Foucault, “Les électeurs infiltrés.
Jérôme Fourquet and Hervé Le Bras’s study had already cast doubt on the idea that the votes of the most regularly practicing Catholics were a statistically decisive factor in Fillon’s electoral victory. See Fourquet and Le Bras, La guerre des trois, 91-116.
See also Dargent, “Les catholiques entre les primaires de la droite et le vote à la présidentielle”.
Brice Teinturier, “Chapitre 1: L’inédite (et dernière?) primaire de la droite et du centre” in Pascal Perrineau (ed.), Le vote disruptif, Paris, Presses de Sciences Po, 2017, pp. 23-42.
See Jaffré, “Le corps électoral déformé de la primaire de la droite”.
Yann Raison du Cleuziou, “Définir le vote catholique légitime: Un enjeu de pouvoir religieux et politique”, Études, 7, 2017, 65-76.
Fourquet and Le Bras, La guerre des trois, 53-60. See also Patrick Stefanini and Carole Barjon, Déflagration. Dans le secret d’une élection impossible, Paris, Robert Laffont, 2017.
Bruno Le Maire spent the most on this budget item, devoting a full quarter of his budget to it. See the campaign accounts published on the website of the primary: <http://www.primaire2016.org/actualite/>.
This budget item was completely neglected by the four “smallest” candidates: Nathalie Kosciusko-Morizet, Bruno Le Maire, Jean-Frédéric Poisson, and Jean-François Copé.
Thierry Vedel and Yves-Marie Cann, “Une communication électorale de rupture? L’Internet dans les stratégies des candidats et les pratiques d’information des électeurs” in Pascal Perrineau (ed.), Le vote de rupture. Les élections présidentielle et législatives d’avril-juin 2007, Paris, Presses de Sciences Po, 2008, pp. 51-75; Karolina Koc Michalska and Thierry Vedel, “Les pratiques informationnelles durant la campagne présidentielle” in Pascal Perrineau (ed.), Le vote normal. Les élections présidentielle et législatives d’avril-mai-juin 2012, Paris, Presses de Sciences Po, 2013, pp. 43-61.
Once the basic principle of an open primary had been agreed, there were still various disagreements between the different candidates’ senior campaign staff regarding how the election should be organized and how open it should be (the number and location of polling stations, how French citizens living abroad should be able to vote, etc.). Some candidates had more of a vested interest than others in having a broad electorate.
Figure provided by the office of the mayor of La Grande-Motte.
Jérôme Fourquet and Hervé Le Bras discuss the “fiefdom effect” in this context: La guerre des trois, 61-76.
On the different forms of this decreased intention to vote for Fillon at the presidential election in 2017, see Martial Foucault and Flora Chanvril-Ligneel, “Le vote François Fillon, autopsie d’un naufrage” in Perrineau (ed.), Le vote disruptif, pp. 221-36.