1François Fillon’s surprise victory in the French primary elections of the political right and center, which were held on 20 and 27 November 2016 to select a common candidate for the 2017 French presidential election, was the first upset in a presidential campaign whose unusualness stunned France. The primary was organized by Les Républicains (LR, The Republicans)—the main French conservative party—the Parti Chrétien-Démocrate (PCD, the Christian-Democrat Party), and the Centre National des Indépendants et Paysans (CNIP, National Center of Independents and Farmers). In fact, it led to the selection of a candidate who was not one of the two favorites—the media and pollsters all having predicted that the second round would be between Nicolas Sarkozy (former French President from 2007 to 2012) and Alain Juppé (former French Prime Minister from 1995 to 1997). [1] In response to criticism of their inability to predict this result, a number of polling companies explained that their predictions had failed because of a practically unforeseeable, last-minute voter mobilization in Fillon’s favor. But this claim does not hold up to empirical evidence. The shifts were too large to be the result of such a very late dynamic, particularly in the absence of any major disruption during the campaign. Their most probable cause was instead the very nature of the primary, which the French right wing was using for the very first time in order to resolve a triple crisis—of leadership, ideology, and finances—which had weakened it since Sarkozy’s defeat in 2012. Moreover, because the right had not used primaries before, and because the turnout and social makeup of potential voters was unknown, it proved very difficult to develop trustworthy samples which would allow pollsters to gauge opinion precisely.
2Alongside traditional surveys of voting intentions, new tools have recently been used to analyze voter opinions and to understand the forces driving electoral campaigns. These include Voting Advice Applications (VAAs). VAAs compare the positions of their users on a variety of issues with those of the parties or candidates for a given election. They have two main uses. First, they provide voters, in a centralized and interactive way, with a relatively large amount of information about candidates’ positions on key issues in the campaign. Second, researchers can collect extensive empirical data spanning the entire electoral campaign. Voters who use VAAs must express positions on a list of issues if they are to receive personalized “voting advice”; in addition, they are also asked to answer optional questions about their socio-demographic situation and more general political attitudes.
3Interest in such tools led to the development of the first VAA for a primary election in France. This article presents the results of this experiment, which was conducted using the website La Boussole Présidentielle, developed in partnership with the website 20minutes.fr at Cevipof in the fall of 2016 by a research team of specialists in electoral sociology, discourse analysis, and the history of ideas, who provided qualitative and quantitative methods. [2] Our discussion is based on data collected by La Boussole Présidentielle in its “Primaire de la droite et du centre” (primary of the right and center) survey. The VAA was launched online on 24 October 2016, almost four weeks before the first round.
4By looking at the campaign trajectories of the three leading candidates in the polls, the data allow us to test two hypotheses which form the core of this article. The first concerns the political positioning of candidates within their own camp during primaries. It is frequently claimed that more radical candidates stand more chances in primary elections because the smaller electorate voting for such elections is considered as more impassioned and more distant from the center of the political spectrum. [3] In this perspective, the fall 2016 primaries helped polarize the right and center electorate, leading candidates to take firmer positions in order to win over the most radical fringes of their supporters, to the neglect of the average and more centrist voters. We will examine this claim, and ask whether Fillon’s selection was the result of such a polarizing logic.
5The second hypothesis concerns the nature of the primary. Much was made of its policy-heavy character, which was supposedly due primarily to a clash of ideas between two opposing positions—that of Juppé, which was seen as more centrist, and those of Sarkozy and Fillon, which were more “unrestrained” and were in line with an electorate the former president had already helped shift rightwards in 2004, when he took control of the UMP. The “clash of ideas” was taken particularly seriously by those competing in the primary. Most of the candidates in the primary extensively focused their campaign on policies, producing a large number of texts and publicly adopting a large number of policy positions. Their aim was to present their vision of France to voters in their political camp who were attentive to such questions. [4] Three televised debates, which pitted the candidates’ positions against each other, apparently reinforced the claim that ideas would play a decisive role in the election. Again, we will examine this claim, and ask whether this primary was actually issues-based. Whatever the answer, our hypotheses raise the question of whether primaries (particularly those of the right and center) are elections like any other.
Literature Review
6In recent years, primaries have spread internationally as a way of choosing candidates in a context of weakening parties (provoked by crises and/or deficiencies of leadership, shrinking memberships, and popular mistrust), of presidentialization, and of the personalization of political life. [5] France joined this trend with debates on a draft bill on primaries (which did not become law) proposed by Charles Pasqua on 27 June 1994. The trend continued to François Hollande’s nomination [as presidential candidate] on 16 October 2011, by almost three million Socialist voters; and culminated in the period preceding the 2017 presidential election, which involved a succession of semi-open primaries, first among the ecologists of the EELV (19 October and 7 November 2016), then with an open primary of the right and center (20 and 27 November 2016), and finally with an open primary of the “Belle Alliance Populaire” organized by the Socialist Party and its allies (22 and 29 January 2017). Such elections are multiplying to the extent that Rémi Lefebvre and Éric Treille have asked whether we are seeing “a primarization of French political life”. [6] But assessments of the democratic issues and political consequences of primaries are, for the most part, contradictory. [7] Some see them as weakening party programs and party-political activism, [8] as a factor in the “depresidentialization” of government, [9] and as an unfortunate victory of communication and public relations over deliberation by well-informed activists. [10] Others believe primaries help revitalize parties, [11] reinforce the role of the president, [12] and revitalize democratic debate. [13] Similarly, procedural and organizational details—the timetable adopted, the conditions of candidacy, the number of polling stations, the boundaries of the electorate—are the object of lively debate in which the law sometimes intervenes. [14] Finally, interpretations of primary effects are split between those who argue for the “primary penalty thesis”, [15] and those who argue for a “primary bonus”. [16] The former emphasizes the importance of hardcore partisans, [17] that is radical supporters who increase the election’s ideological polarization and encourage the selection of candidates who are more radical than the official party line. [18] The latter emphasizes the electoral gains the chosen candidate enjoys, [19] the internal regulation of dissent as a strategy to preserve party unity, [20] and the legitimacy bonus the successful candidate receives that goes beyond the core electoral base. [21] Whatever the case may be, open primaries “seem to establish the primacy of the party s’upporter’” [22] at the expense of the party activist. They therefore raise questions about how to mobilize and identify this selectorate, which is broader than just the party activists, and push us to examine the distinctive campaign dynamics which arise during primaries.
7Interest in campaign dynamics—which allow us to track voter behaviors in the long term—has been unevenly adopted and operationalized within electoral sociology. [23] This is explained by the growing instability of voting behaviors, the ever-greater range of choice available, and the gradual erosion of the links between political affiliation and voting decisions. Such a situation demands longitudinal survey protocols like sample groups or “rolling cross-sections” using robust voter samples. These survey protocols have only recently been adopted in French political science, unlike among its English-speaking counterparts. [24] Such methods allow us to reconstruct electoral itineraries in order to better understand the logic behind them. [25] This interest in campaign dynamics is linked to research that emphasizes the role of issues, [26] the importance of television and other media in politics, [27] and the process of “personalization”. [28] As primary elections in France are characterized by competing policy visions, the presidency as the end goal, majority suffrage, two-round voting, media attention, and citizen participation, they provide an ideal setting for the inquiry of the effect of campaign dynamics in voters’ decision-making.
8During the late 1960s, issue voting emerged as an alternative model to those based on sociological determinism and/or partisan identity, which were beginning to show their limits. [29] Researchers have been interested in the congruence between voters’ preferences or priorities and those of the candidates as expressed in their public policy platforms. The replacement of “party voting” by “issue voting” has found a place within the growing body of work on changing values and the erosion of traditional political cleavages in the United States and Europe. [30] French researchers contributed to this line of research, in spite of powerful criticism both of its premises and its reliance on issue voting as an analytical framework. [31] Some studies have shown that concerns about the environment and immigration are sufficient to explain the emergence of the Verts (the French Green Party, founded in 1984) and the Front National (the main French extreme-right party, created in 1972 and which took off electorally in the 1980s). [32] But others have indicated that the explanatory power of issue voting was rather limited in 2007, where “strategic considerations were paramount as early as the first round of the presidential election. The candidates’ relative chances of qualifying for the second round took precedence over more substantial questions”. [33] The difficulty of observing the impact made specifically by policy issues may be more marked during the fall 2016 primaries, because candidates come from the same political family and therefore share, in theory, relatively similar positions and outlooks on key policy issues.
9Despite the growing opportunities offered by the internet, and particularly by social networks, for political communication, [34] television has remained “the main source of information for voters”. [35] This continued to be true for the 2017 presidential campaign. Thus, even if there are methodological problems with measuring the impact on voter choice of televised debates between candidates, recent research has nonetheless shown that television played a clear role in changes in viewers’ voting intentions, particularly after the inter-round debate between Emmanuel Macron and Marine Le Pen which took place on 4 May 2017. [36] As several television debates between candidates were organized during the campaign of the primary of the right and center, in this article we should therefore try to identify the influence of televised debates during the primaries, particularly because the media coverage of these campaign events is likely to have impacted the candidates’ images.
10By emphasizing the idea of voters’ “affective intelligence” (defined as a mix of reflex, emotional, and reasoned processes), George Marcus’s work has demonstrated the role of emotional factors in electoral choice, setting rational choice analyses in a broader context. [37] Among indicators which may allow us to understand voters’ emotional intelligence, some researchers have used image-based approaches, hypothesizing that “cognition and emotion can go together”. [38] Researchers interested in the effects of candidates’ images have looked at French presidential elections between 1965 and 2007, [39] but other types of elections have not been studied systematically. [40] This includes primaries, which only appeared recently in the French political landscape. Daniel Boy and Jean Chiche demonstrated that during the 2012 election, images had significant explanatory power—of the same order of magnitude as voters’ position on the left-right scale and their political affiliation—for voting intentions and the moment of voting itself. Furthermore, “images are not constants, established at the beginning of the campaign and then remaining attached to a candidate for its duration. For good or bad reasons, the images of candidates vary with their pronouncements, their performance in debates, and their positions”. [41] The ability to follow the dynamics of the campaign which La Boussole Présidentielle offers us is a valuable resource for understanding the converging or competing logics of issues and images within primaries.
The Boussole Présidentielle Sample
11The structure of VAA data is very similar to that produced by rolling cross-section surveys in which voters are polled daily. To understand the campaign dynamics in the primary of the right and center, we apply methods typically used in rolling crosssection analysis to VAA data. The frequent (ideally daily) repetition of the same questionnaire among micro-samples allows researchers to approach “true causal inference” [42]: rolling crosssections enable the establishment of “links to debates, news coverage, and campaign advertising, as well as identification of the social and psychological mechanisms that mediate the potential impact from external forces”. [43] They are therefore particularly well-suited to analyzing campaign effects. [44] As made clear by Bruno Cautrès and Anne Muxel, the small size of traditional rolling cross-section samples limits the scope of the possible inference. [45] By contrast, VAA data provide large daily samples which allow us to strengthen statistical inference of the decision-making process of voters and the impact of campaign dynamics on it.
12We use a non-probabilistic sample of volunteer respondents collected during the fall 2016 primary of the right and center. La Boussole Présidentielle was online from 24 October to 27 November, the day of the second round. Around 90,000 people expressed positions on the statements presented by the application, and received a personalized “voting advice”. At the beginning of the series of issue-related statements which users had to respond to in order receive their voting advice, they could answer questions about their socio-demographic situation (gender, age, education level, socio-professional category, and so on), and—to avoid “contamination” of expressed issue positions—at the end about their political attitudes and behaviors (certainty that they would vote in the primary, evaluation of candidates’ images, likelihood of voting for each candidate, self-positioning on the left-right scale, and voting intention).
13In order to obtain a sample of reliable users, we eliminated those who did not respond to all questions, and those who responded identically to all issue statements. [46] We then removed from our sample any users under 18 years old (who therefore could not vote), those who answered the questionnaire in the period between the two primary rounds, and those who were unsure they would vote in the first round. The sample of users from La Boussole Présidentielle used in this study is made up of 5,304 voters (see Appendix for the socio-demographic composition of the sample). [47]
14Figure 1 shows the distribution of this sample of 5,304 users during the four weeks before the first round. The VAA recorded new users every day, but we observe a far higher number of respondents at four particular moments: when the VAA first went live, during the second and third debates (the first took place on 13 October and therefore does not fall within the scope of our study), and during the weekend of the first round.
Chronological distribution of the sample of users of La Boussole Présidentielle (N = 5,304)
Chronological distribution of the sample of users of La Boussole Présidentielle (N = 5,304)
Results
François Fillon, a Median Candidate
15La Boussole Présidentielle compares users’ positions with those of the candidates on the major issues of an electoral campaign. We can thereby observe which primary candidates users most agree with. Figure 2 shows that, for the entire month of campaigning during which the VAA was online, user opinions matched those of Nathalie Kosciusko-Morizet and Jean-François Copé most closely. These two candidates’ political platforms were both very clear and very distinct. Kosciusko-Morizet was the most “centrist” candidate socio-economically, and the most progressive on social and institutional questions. [48] Copé was the most conservative on issues of authority and on social questions, and was very far to the right economically. [49] The fact that user opinions are so close to these two candidates demonstrates a polarization within our sample between two coherent ideological profiles. On the one hand, we observe a profile shared by users who are socially progressive and socio-economically centrist, while on the other hand, we see a profile of users who are socially conservative and economically very far to the right. This polarization of platforms and voters explains why, in Figure 2, Juppé receives the lowest score for agreement between his positions and those of users. More than any other candidate, he tried to combine right-wing socio-economic positions (abolishing the Impôt de Solidarité sur la Fortune [ISF, solidarity tax on wealth], following a balanced budget policy, raising VAT, tax cuts for businesses, and so on) with socially progressive positions, centered on the “happy identity” he promoted during the election (defending the integration of immigrants, refusing to abolish birthright citizenship, opposing a ban on Islamic veils in public spaces, and so on). Although Juppé was expected to win, La Boussole Présidentielle shows that users who were certain of voting in the first round agreed with him the least on the seventeen issues it surveyed. In sum, there was a substantial gap between Juppé’s issue positions and those held on average by primary voters.
Agreement, in percent, between the positions of candidates and those of users on issues included in La Boussole Présidentielle
Agreement, in percent, between the positions of candidates and those of users on issues included in La Boussole Présidentielle
16Ideological markers like left-right split determine voting patterns to a lesser degree during primaries. Nonetheless, they confirm that platforms and voters were polarized during the primary of the right and center. [50] Figure 3 and Table 1 compare voting intentions in the first round with voters’ self-positioning on the left-right scale. Users who intended to vote for Kosciusko-Morizet placed themselves at the center of the scale (an average of 4.98). Those intending to vote for Copé were mostly to the right (an average of 7.17). The two candidates seen by pollsters as the likely finalists in the primary also represented electorates which were very distant from each other on the left-right scale [51]: Sarkozy’s electorate positioned itself farthest to the right (an average of 7.43), and Juppé’s close to the center (an average of 5.51). The polarization between the two candidates who dominated the primary campaign left an empty political space for alternative candidacies to emerge—that is, a candidate farther to the right than Juppé, but not as far right as Sarkozy. Indeed, the electorate’s center of gravity during this primary was located precisely between Juppé and Sarkozy: the average for users who were certain of voting was 6.39. Situating themselves between the “centrist” Juppé and the “right-wing” Sarkozy, both Fillon and Bruno Le Maire tried to present alternative candidacies, more aligned with the supposed selectorate of the primary. Users who said they intended to vote for Fillon or Le Maire positioned themselves, on average, at 6.89 and 6.6 respectively on the left-right scale.
Self-positioning on the left-right scale of users of La Boussole Présidentielle as a function of their voting intentions in the first round of the primary
Self-positioning on the left-right scale of users of La Boussole Présidentielle as a function of their voting intentions in the first round of the primary
Descriptive statistics on self-positioning on the left-right scale as a function of voting intentions in the first round of the primary
| Kosciusko-Morizet | Juppé | Le Maire | Fillon | Poisson | Sarkozy | Copé | Sample | |
|---|---|---|---|---|---|---|---|---|
| Average | 4.98 | 5.51 | 6.60 | 6.89 | 7.06 | 7.43 | 7.17 | 6.39 |
| Standard deviation | 1.53 | 1.48 | 1.68 | 1.34 | 2.01 | 1.35 | 1.58 | 1.67 |
| N | 253 | 1,111 | 246 | 1,148 | 87 | 582 | 41 | 3,468 |
Descriptive statistics on self-positioning on the left-right scale as a function of voting intentions in the first round of the primary
17Figure 4 shows the distribution and development of voting intentions within the sample of users, from the VAA launch date (24 October) to the first round (20 November). Two lessons emerge. First, Fillon was felt to be a credible candidate by the primary electorate. Almost a month before the first round, La Boussole’s users expected him to reach the second round—to the disadvantage of the other “third man”, Le Maire, and especially of Sarkozy, whom the polls expected to be the ultimate victor. Fillon subsequently enjoyed a strong swing in his favor. In our sample, Fillon overtook Juppé on 10 November; polling, however, only identified this as taking place on 18 November, two days before the vote. Furthermore, 40% of users in our sample intended to vote for Fillon in the last week of the campaign, and he ultimately won the first round with 44% of the vote. Finally, between a candidate like Juppé who was felt to be too centrist and one like Sarkozy who was felt to be too right-wing, an outsider like Fillon could mobilize the primary electorate and generate real momentum around his candidacy. This finding challenges the polarization hypothesis: outside observers may have felt that the eventual winner of the primary ran a very right-wing campaign, but he actually won the primary election by being perceived by conservative and centrist voters as a candidate of compromise within the different factions of these political families. His victory paints a picture of an opposition party seeking an average candidate to unite its members and supporters, while also being attentive to image, and aiming to maximize its chances of electoral victory in the presidential race. Rather than affirming radicalism, Fillon’s victory demonstrates a search for internal equilibrium around a candidate that is not the most extreme.
Voting intentions of users who were certain to vote in the first round – La Boussole Présidentielle (weighted data and smoothed averages over seven days)
Voting intentions of users who were certain to vote in the first round – La Boussole Présidentielle (weighted data and smoothed averages over seven days)
The Predominance of Image over Issues
A Dynamic Fillon Driven by His “Presidentiality”
18Richard Johnston and Henry Brady have argued that, in order to draw conclusions about the effects of a campaign, it is not enough to test for an association between predictive variables and electoral preferences. Such methods must be supplemented by a graphical representation of key variables. [52] Such visualizations allow us to identify the effects of explanatory variables on voting chronologically. We have already discussed the level of agreement on issues between the sample of users and the candidates (see Figure 2). We have observed the coexistence in our sample of two ideologically polarized user profiles. Beyond this finding, we also see that levels of agreement between users and candidates hardly changed during the campaign, with none of the curves reflecting the trends shown in Figure 4. Paradoxically, the intensity of policy production during this primary did not lead to any particular mobilization on these issues. This was in spite of intense media coverage, particularly around the debates on 13 October, 3 November, and 17 November. As these debates were used by the candidates primarily as a means of introducing themselves to the voting body, we must analyze the shaping of candidates’ images over the campaign, and whether it influenced voting dynamics.
19Figure 5 shows, across the campaign, changes within our sample in the average score for the propensity to vote (PTV) and image traits (presidentiality and empathy) for Fillon, Juppé, Le Maire, and Sarkozy. [53] We have decided to use PTVs as an indicator to measure users’ electoral preferences. This choice allows us “to measure voters’ doubts” and “to reveal the complex map of voters’ political preferences for several candidates”, and is therefore particularly suited to a primary where (almost) all the candidates belonged to the same political party. [54]
20Our first observation is that Fillon’s scores clearly increased as the campaign progressed, while they did not rise for Juppé, rose only very little for Sarkozy, and fell for Le Maire. The aggregate values from the sample also confirm the existence of a swing in favor of Fillon: the average likelihood of voting for Fillon rose from 4 to 7 (on a scale from 0 to 10) between 24-26 October and 18-20 November.
21Secondly, Fillon’s “presidentiality” (+1.8 points) rose more than his “empathy” (+1.2 points). His presidentiality reached a definitive high during 2-4 November, when the second debate took place. [55] We note that Fillon’s rise in presidentiality occurred before the marked growth in his PTV score, which also began (though to a lesser degree) on 2-4 November. The rise in Fillon’s PTV score continued to the end of the campaign, eventually far exceeding his empathy score and almost equaling his presidentiality score. We suggest that it was Fillon’s presidentiality that caused his PTV score to rise over the course of the campaign.
22Juppé’s image traits’ scores barely changed during the campaign. They never rose in a way that might have had a positive effect on his PTV score. We also note that from 2-4 November onward, his empathy and PTV score developed parallel to each other. Similarly, Sarkozy’s image traits and the likelihood of voting for him developed in concert. This suggests a strong link between these variables, but does not allow us to distinguish between the two image traits. Finally, Le Maire’s campaign seems to have had no effect: between our first data point and the first round, his PTV score declined (-0.8 points), as did his presidentiality (-0.4 points) and empathy scores (-0.2 points).
Longitudinal Campaign Effects: Image Effects and the Persistence of Polarization between Candidates
23In order to assess the relative weight of the different explanatory variables for the likelihoods of voting for the main candidates, we used a conditional change model. [56] To understand campaign effects, such a statistical model estimates the likelihood of voting for a candidate as a function of the explanatory variables measured on the day the voters filled out the questionnaire, controlled for the likelihood of voting for a candidate measured before the campaign. In a VAA, this pre-campaign reference value is of course missing from the data. However, it can be estimated by including in the model centered-values for our key explanatory variables: candidates’ image traits (presidentiality and empathy) and issue agreement between each candidate and our respondents. That is, we subtract the average value for all respondents on a day d from the value given by a respondent on the same day. [57] This process creates (pre-campaign) reference values, or control values, for the key explanatory variables. [58] It is based on the hypothesis that, from one day to another, the average variation in a value (for instance, Fillon’s presidentiality) is the same for all respondents. This is a strong hypothesis but, because we do not have a real reference or control measure, it is a necessary one. We can model the conditional change model with the following equation:
25In this equation, {Yit} represents the likelihood of voting for a candidate, {Bit} the value of the independent variable measured during the campaign, and {(Bit - B*i)} the parameterization of the independent variable which gives its pre-campaign value. Within this parameterization, {B*i} represents the average of the values recorded for this variable for all respondents in the sample over a three-day period. [59]
26This model contains four dependent variables: the likelihoods of voting for Fillon, Juppé, Le Maire, and Sarkozy. Our hypothesis is that voting preference depends on a voter’s position on the left-right scale (an important variable to take into account in this primary, because a section of Juppé’s potential voters are located on the center left or left); on agreement on issues (the basic statements of La Boussole Présidentielle); and on candidates’ image traits (presidentiality and empathy). As control variables, we use socio-demographic variables and reference variables for image traits and issue agreement. For each candidate we constructed a series of three linear regression models, going from the simplest (left-right scale alone, with socio-demographic control variables) to the most complex (left-right scale, issues, and images, with socio-demographic variables and reference variables as controls). The socio-demographic control variables are gender, age (in five categories), and education level (in three categories: lower than baccalauréat (French high school degree), baccalauréat or limited higher education, and university degree). Finally, the left-right scale was divided into five categories. When we insert these variables into the conditional change model, we obtain the following equation:
28+ Control variablesit + εitFor each of the four predicted likelihoods of voting, we note first of all that the introduction of candidate image variables leads to a sharp rise in the variance explained by the models. Whatever the predicted likelihood, the values of the adjusted R2s are multiplied by a factor of at least 2.5 when we go from a model based only on the left-right scale to the one which includes image effects. [60] Similarly, adjusted R2 values increase significantly if we compare models which only include issue agreement effects with those which also include image effects.
29Examining the effects of the explanatory variables shows, first of all, that the left-right scale remains highly statistically significant no matter what model we use, with the exception of Le Maire in Model 3. [61] This result is particularly remarkable because we are analyzing likelihoods of voting within a primary, where left-right range is by definition far more limited than it would be in a presidential election. The three main candidates we are making predictions about no doubt strongly emphasized their left-right positioning with their potential electorates: Fillon and Sarkozy saw their voting likelihoods grow increasingly the further to the right their potential voters were situated, and the opposite was true for Juppé.
Linear regressions (OLS) for likelihood of voting for candidates [62]
| ALAIN JUPPÉ | NICOLAS SARKOZY | |||||
| Model 1 | Model 2 | Model 3 | Model 1 | Model 2 | Model 3 | |
| Female | .647*** (.148) | .524*** (.148) | .003 (.106) | -.743*** (.147) | -.737*** (.144) | -.391*** (.101) |
| Age | .030 (.063) | .034 (.063) | -.015 (.072) | .060 (.063) | -.016 (.062) | .021 (.044) |
| Diploma | .302** (.100) | .199* (.101) | -.010 (.045) | -.639*** (.099) | -.553*** (.097) | -.372*** (.068) |
| Left-right scale | -1.034*** (.069) | -.965*** (.070) | -.771*** (.049) | 1.674*** (.069) | 1.403*** (.073) | .185*** (.056) |
| Agreement on issues | .257* (.112) | .180* (.082) | .248 (.132) | -.156 (.107) | ||
| Reference - issues | -.220* (.050) | -.177* (.082) | -.181 (.132) | .184 (.107) | ||
| Empathy | 1.172*** (.348) | 2.532*** (.698) | ||||
| Reference - empathy | -.725* (.348) | -1.981** (.698) | ||||
| Presidentiality | -.725* (.348) | -.978* (.457) | ||||
| Reference - presidentiality | 1.030* (.419) | 1.328** (.457) | ||||
| Constant | 8.468*** (.475) | 1.041 (3.288) | .565 (3.493) | -1.175* (.475) | -10.532 (5.572) | 5.363 (4.176) |
| Adjusted R2 | .089 | .102 | .550 | .211 | .243 | .632 |
| N | 2599 | 2599 | 2599 | 2621 | 2621 | 2621 |
| FRANÇOIS FILLON | BRUNO LE MAIRE | |||||
| Model 1 | Model 2 | Model 3 | Model 1 | Model 2 | Model 3 | |
| Femala | -.012 (.145) | -.035 (.141) | -.058 (.095) | -.037 (.140) | -.050 (.14) | -.132 (.10) |
| Age | .289*** (.062) | .185** (.061) | .017 (.041) | -.324*** (.059) | -.308*** (.059) | -.274*** (.042) |
| Diploma | .279** (.097) | .218* (.094) | .057 (.063) | -.284** (.094) | -.285** (.093) | -.188** (.067) |
| Left-right scale | 1.126*** (.069) | .884*** (.072) | .215*** (.050) | .362*** (.066) | .373*** (.065) | -.085 (.048) |
| Agreement on issues | -.748*** (.062) | -.040 (.061) | -.326** (.107) | -.03 (.102) | ||
| Reference - issues | .797*** (.062) | .050 (.061) | .345*** (.107) | .033 (.102) | ||
| Empathy | -.108 (.515) | .107 (.498) | ||||
| Reference - empathy | .516 (.515) | .181 (.498) | ||||
| Presidentiality | 1.555*** (.394) | 1.224** (.394) | ||||
| Reference-presidentiality | -.962* (.395) | -.591 (.394) | ||||
| Constant | -.377 (.469) | 33.852*** (2.754) | -3.295 (2.852) | 3.69*** (.446) | 17.295*** (4.536) | .501 (5.096) |
| Adjusted R2 | .101 | .171 | .625 | .024 | .031 | .510 |
| N | 2626 | 2626 | 2626 | 2482 | 2482 | 2482 |
Linear regressions (OLS) for likelihood of voting for candidates [62]
30This result directly addresses our first hypothesis regarding the polarization of electorates during primary elections. The Boussole Présidentielle data confirm a polarization effect, but one arising in a context of high primary turnout where, on the one hand, the space occupied by Juppé’s potential voters was not located at the center of gravity of the right-wing electorate, and, on the other hand, the roughly 150,000 LR members who elected Sarkozy head of the party in November 2014 were lost in a crowd of more than four million primary voters. We will now turn to issues and image in order to understand the forces that led to Fillon’s victory.
31The variables which measure the effect of issues (Model 2) had a far greater influence on the likelihood of voting for Fillon than for Juppé or Le Maire, and had no effect at all on that of voting for Sarkozy. Fillon won the issues vote in this primary, but our conditional change model shows that this issue voting pattern was not due to campaign dynamics (the “agreement on issues” variable), but was in place before the campaign began (the “reference—issues” variable). [63] The campaign decreased the importance of issue voting, which worked in Fillon’s favor. Paradoxically, during the campaign, agreement with Fillon did not mean one was more likely to vote for him—indeed, the contrary was true. Instead, the change in the Fillon vote during the campaign lay with the candidate’s image. When we introduce the image variables into the regression model (Model 3), their effects reduce the significance of the issues variable and the socio-demographic variables to almost nothing. In Model 3, it is Fillon’s presidentiality that counts, and not his empathy image. Before the campaign, viewing Fillon as presidential did not lead to a higher likelihood of voting for him (the “reference—presidentiality” variable); during the campaign, however, a one-point rise in his presidentiality led to a 1.5-point rise in the likelihood of voting for him. This also means that, before 24 October (the launch date for the VAA), users of La Boussole Présidentielle who did not intend to vote for Fillon nonetheless felt he was presidential. It was during the campaign that Fillon was able to turn this into voting momentum in his favor.
32Juppé’s campaign dynamic was affected by both issue voting and his image. We can observe positive, significant coefficients for the variables for agreement on issues and empathy: a one-point rise in agreement on issues translates to a 0.18-point rise in the likelihood of voting for him. Agreement with Juppé involves liberal views on economic and social issues, positions which correspond to the ideological profile of Macron’s electorate—a sign of ideological and electoral convergence between the future president and part of the center-right electorate. The negative coefficient of the presidentiality variable indicates that, during the campaign, even those respondents in our sample who said they were unlikely to vote for Juppé felt he was presidential.
33Finally, two things stand out when we examine the result of our regression models for the likelihood of voting for Sarkozy. Firstly, the issues variables are never significant, indicating that his campaign dynamic was affected solely by image variables, and particularly empathy: a one-point rise in empathy led to a 2.5-point rise in the likelihood of voting for him. He was by far the candidate whose image played the most important role in voting. While our respondents accorded him a degree of presidentiality (see Figure 6), this assessment did not translate into a vote in his favor (as shown by the negative coefficient of the “presidentiality” variable). Secondly, among the main candidates, he was the only one for whom some of the socio-demographic control variables remained significant in Model 3, which was also the case for Le Maire. The two of them attracted minority socio-demographic groups within the primary selectorate: less educated voters for Sarkozy, and young voters for Le Maire.
34In this article, we have tried to understand the electoral dynamics in play during the primary of the right and center in November 2016 using results based on data from La Boussole Présidentielle. They show firstly that, contrary to a very dominant narrative, when faced with the very polarized candidacies of Kosciusko-Morizet and Juppé on the left and Copé and Sarkozy on the right, the primary electorate saw Fillon as a compromise candidate. Secondly, our results allow us to show that issues were far less important than expected in the outcome of the election. By contrast, images—particularly Fillon’s—were fundamental for selecting a candidate. In other words, Fillon won the election less because of his positions than because his candidate image matched with the one voters had of the ideal right-wing candidate at the moment he was chosen. This conclusion invalidates the hypothesis that Fillon won because of a debate about ideas, which suggested that his campaign was centered around his policy positions—particularly given the success of his book, Faire, which was published in 2015 and sold more than 100,000 copies.
35Fillon’s success reveals a paradox. He was portrayed in the presidential campaign as a radical. His rivals attacked him, in particular, for his support from the Catholic activist group Sens Commun (Common Sense), an offshoot of the Manif Pour Tous movement. [64] His radicalization was largely a result of the collapse of his legitimacy within LR following the “Fillon affair” that started early February 2017. Within his political camp, Fillon was viewed as belonging to the center, if not as a centrist—as a moderate personality who could at least temporarily quell the ideological crisis LR went through in the wake of Sarkozy’s presidential defeat in 2012. The VAA data analyses do not entirely invalidate the hypothesis of voter polarization during the primary. Such polarization existed before the primary, and reappeared in the presidential election after Fillon’s defeat. Fillon, the average candidate, may have been nominated, but this polarization continued with the resurgence of a radicalized right wing around Laurent Wauquiez, who was elected president of LR in December 2017. Wauquiez re-established the divisions within his political camp between centrists and a “true right wing” based on strong positions on national identity, borders, and immigration. The results of our study—the relative unimportance of issue voting during the primary, the importance of image and presidentiality, the search for a compromise with the expectation of winning the presidential election—allow us to see how important political psychology and the construction of social representations of political actors are for understanding primaries, a type of election which is still relatively new in French political life, particularly on the right.
Socio-demographic composition of the sample and comparisons
| Boussole sample (N = 5,304) | ELABE “election day” poll (N = 1,892) | |
|---|---|---|
| Gender | ||
| Male | 66% | 59% |
| Female | 34% | 41% |
| Age | ||
| 18-24 | 5% | 6% |
| 25-34 | 8% | 10% |
| 35-49 | 22% | 22% |
| 50-64 | 31% | 23% |
| 64 < | 34% | 39% |
| Educational level | ||
| Lower than baccalauréat | 17% | |
| Baccalauréat or some higher education | 33% | |
| University degree | 50% | |
| Left-right scale | Préférence partisane | |
| Extreme left | 2% | 14% (none) |
| Left | 9% | 15% (Left) |
| Centre | 20% | 14% (Modem-UDI) |
| Right | 42% | 46% (LR) |
| Extreme right | 27% | 11% (FN-DLF) |
| Socio-professional categories | ||
| Farmers | 1% | |
| Artisans, traders, business owners | 12% | |
| Managers and liberal professions | 54% | |
| Intermediary professions | 14% | |
| Salaried workers | 12% | |
| Workers | 2% | |
| Economically inactive | 5% | |
| Professional status | Social category | |
| Public sector employees | 13% | 32% (Middle and upper class) |
| Private sector employees | 33% | 18% (Working class) |
| Self-employed | 12% | |
| Retired | 32% | 43% (Retired) |
| Other inactive | 10% | 7% (Other inactive) |
Notes
-
[1]
For instance, one week out from the first round, pollsters predicted that Juppé would receive 31-37% in the first round, compared to 25-30% for Sarkozy and 18-27% for Fillon. In a survey conducted on 13 November, BVA found that 37% of respondents intended to vote for Juppé, 29% for Sarkozy, and 19% for Fillon.
-
[2]
Cevipof has developed two other VAAs: one for the 2012 presidential election, and one for the 2014 European elections. The 2017 Boussole Présidentielle was developed by researchers from the Centre de recherche politique de Sciences Po (Cevipof)–Bruno Cautrès, Vincent Martigny, Sylvie Strudel, and Thomas Vitiello–in partnership with the free daily 20Minutes and Kieskompas (Election Compass), a Dutch research institute which helped introduce VAAs to Europe. For a presentation of the first VAA set up by French researchers, see Thomas Vitiello, “Les études électorales à l’ère d’Internet. Les systèmes d’information et d’aide au vote: le cas de ‘La Boussole présidentielle’” in Jacques Gerstlé and Raul Magni-Berton (eds), 2012. La campagne présidentielle. Observer les médias, les électeurs, les candidats, Paris, Pepper/L’Harmattan, 2014, pp. 137-51. For a review of the literature on VAAs, see especially Diego Garzia and Stefan Marschall (eds), Matching Voters with Parties and Candidates: Voting advice applications in a comparative perspective, Essex, ECPR, 2014, and the special issue edited by Patrick Dumont and Raphaël Kies, “Les systèmes d’aide au vote: défis et potentialities”, Revue internationale de politique comparée, 22, 2015.
-
[3]
Barbara Norrander, “Ideological representativeness of presidential primary voters”, American Journal of Political Science, 33, 1989, 570-87.
-
[4]
Jean-François Copé, Le sursaut français, Paris, Stock, 2016; François Fillon, Faire, Paris, Albin Michel, 2015; Alain Juppé, Mes chemins pour l’école, Paris, Jean-Claude Lattés, 2015, Pour un État fort, Paris, Jean-Claude Lattés, 2016, and Cinq ans pour l’emploi, Paris, Jean-Claude Lattés, 2016; Nathalie Kosciusko-Morizet, Nous avons changé de monde, Paris, Albin Michel, 2016; Bruno Le Maire, Ne vous résignez pas!, Paris, Albin Michel, 2016; Nicolas Sarkozy, La France pour la vie, Paris, Plon, 2016, and Tout pour la France, Paris, Plon, 2016.
-
[5]
Giulia Sandri, Antonella Seddone, and Fulvio Venturino (eds), Party Primaries in Comparative Perspective, Farnham, Ashgate, 2015; Pouvoirs, 154, special issue “Les primaires”, 2015; Olivier Duhamel, Les primaires pour les nuls, Paris, Éditions First, 2016.
-
[6]
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, Rennes, Presses Universitaires de Rennes, 2016, pp. 11-40, here p. 11.
-
[7]
Luca Bernardi, Giulia Sandri, and Antonella Seddone, “Challenges of political participation and intra-party democracy: bittersweet symphony from party membership and primary elections in Italy”, Acta politica, 52(2), 2016, 218-40; William Cross and Jean-Benoît Pilet (eds), The Selection of Political Party Leaders in Contemporary Parliamentary Democracies, London, Routledge, 2013; David Peterson and Paul Djupe, “When primary campaigns go negative: the determinants of campaign negativity”, Political Research Quarterly, 58(1), 2005, 45-54; Amber Wichowsky and Sarah Niebler, “Narrow victories and hard games: revisiting the primary divisiveness hypothesis”, American Politics Research, 38(6), 2010, 1052-71.
-
[8]
Jonathan Hopkin, “Bringing the members back in? democratizing candidate selection in Britain and Spain”, Party Politics, 7(3), 2001, 343-61; Rémi Lefebvre, Les primaires socialistes, La fin du parti militant, Paris, Raisons d’Agir, 2011.
-
[9]
Bernard Dolez, “Le Parti socialiste, les primaires et la (dé)présidentialisation du regime”, Pouvoirs, 154, 2015, 67-79.
-
[10]
Pierre Avril, “Les primaires: un affaiblissement de la démocratie?”, Pouvoirs, 154, 2015, 133-42.
-
[11]
Gérard Grunberg, “Les primaires présidentielles ouvertes: mort ou renaissance des partis?”, Pouvoirs, 154, 2015, 55-65. Note a shift in the thesis which Lefebvre and Treille defend compared to their 2011 book, where the authors argue that “primaries remain a ‘party affair’”: see Lefebvre and Treille, “Introduction”, p.15ff.
-
[12]
Vincent Martigny, “La mort du débat institutionnel”, Le 1, 96, 2016.
-
[13]
Dominique Rousseau, “Les primaires: un sens et un projet démocratiques”, Pouvoirs, 154, 2015, 125-31.
-
[14]
Alexandre Lemarié, “Primaire à droite: Sarkozy veut diminuer le nombre de bureaux de vote”, Le Monde, 13 October 2015; Anne Levade, “Le droit des primaires: règles, contrôle, finances, sanctions”, Pouvoirs, 154, 2015, 99-109.
-
[15]
Peterson and Djupe, “When primary campaigns go negative”; Wichowsky and Niebler, “Narrow victories and hard games”.
-
[16]
John Carey and John Polga-Hecimovich, “Primary elections and candidate strength in Latin America”, The Journal of Politics, 68(3), 2006, 530-43.
-
[17]
Norrander, “Ideological representativeness of presidential primary voters”.
-
[18]
Nelson Polsby and Aaron Wildavsky, Presidential Elections, New York, Scribner and Sons, 1978; David W. Brady, Hahrie Han, and Jeremy C. Pope, “Primary elections and candidate ideology: out of step with the primary electorate?”, Legislative Studies Quarterly, 32(1), February 2007, 79-105.
-
[19]
Gilles Serra, “Primaries, conventions, and other methods for nominating candidates: how do they matter?”, CIDE Working Papers, DTEP-285, September 2016.
-
[20]
Rafael Hortala-Vallve and Hannes Mueller, “Primaries: the unifying force”, Public Choice, 163(3-4), 2015, 289-305; Vanessa Jérome, “Des primaires pour (ré)unir les écologistes?” in Lefebvre and Treille (eds), Les primaires ouvertes en France, pp. 105-24; Rémi Lefebvre, “Les primaires à droite” in ibid., pp. 65-104; Éric Treille, “La fabrication partisane des primaires socialistes ou la codification d’une nouvelle règle du jeu electoral” in ibid., pp. 41-64.
-
[21]
Éric Montigny and Charles Tessier, “Vers des primaires ouvertes: des partis en quête de légitimité populaire. Les cas du PQ et du PLC”, Politique et Sociétés, 36(2), 2017, 119-41.
-
[22]
Lefebvre and Treille, “Introduction”, p. 27.
-
[23]
Bruno Cautrès and Anne Muxel, “Introduction: the dynamics of electoral choice” in Bruno Cautrès and Anne Muxel (eds), The New Voter in Western Europe: France and beyond, Basingstoke, Palgrave Macmillan, 2011, pp. 1-9.
-
[24]
These methods remained marginal in spite of a vibrant argument in their favor during the 1960s: François Chazel, “La méthode du panel et ses possibilités d’application à la structure politique française”, Revue française de sociologie, 7, 1966, 684-99. For a useful account of longitudinal approaches in French electoral sociology, see Viviane Le Hay, “The ‘Panel Électoral Français’ 2007: methodological issues” in Cautrès and Muxel (eds), The New Voter in Western Europe, pp. 225-49. Such techniques were used very early on by, among others, Paul Lazarsfeld, “‘Panel’ studies”, The Public Opinion Quarterly, 2(4), 1938, 596-612.
-
[25]
Kathleen Knight and Michael Marsh, “Varieties of election studies” in Mark Franklin and Christopher Wlezien (eds), The Future of Election Studies, Oxford, Pergamon, 2002, pp. 13-31.
-
[26]
Particularly following Valdimar O. Key’s pioneering book, The Responsible Electorate: Rationality in presidential voting, 1936-1960, Cambridge, Belknap, 1966.
-
[27]
Beginning with the work of Jay G. Blumler, “Three ages of political communication”, Revue de l’Institut de sociologie, 47(1-2), 1999, 31-45, and continuing to the most recent instances which show the relationships between political behavior and televisual practices. See Thierry Vedel, “La télévision, un vieux média qui ne meurt pas” in Pascal Perrineau (ed.), Le vote disruptif: Les élections présidentielles et législatives de 2017, Paris, Presses de Sciences Po, 2017, pp. 101-14.
-
[28]
Ian McAllister, “The personalization of politics” in Russell J. Dalton and Hans-Dieter Klingemann (eds), The Oxford Handbook of Political Behavior, Oxford, Oxford University Press, 2007, pp. 571-88. See also Alan Renwick and Jean-Benoît Pilet, Faces on the Ballot: The personalization of electoral systems in Europe, Oxford, Oxford University Press, 2016.
-
[29]
David Butler and Donald Stokes, Political Change in Britain: The evolution of electoral choice, Basingstoke, Macmillan, 1970; Norman Nie, Sidney Verba, and John Petrocik, The Changing American Voter, Cambridge, Harvard University Press, 1976.
-
[30]
Russell J. Dalton and Martin P. Wattenberg (eds), Parties without Partisans: Political change in advanced industrial democracies, Oxford, Oxford University Press, 2000; Mark Franklin, “Demographic and political components in the decline of British class voting 1964-1979”, Electoral Studies, 1(2), 1982, 195-220; Ronald Inglehart and Scott Flanagan, “Value change in industrial societies”, American Political Science Review, 81(4), 1987, 1289-319.
-
[31]
Loïc Blondiaux, “Mort et résurrection de l’électeur rationnel”, Revue française de science politique, 46(5), October 1996, 753-91; Patrick Lehingue, Le vote, Paris, La Découverte, 2011.
-
[32]
Jean Chiche and Nonna Mayer, “Les enjeux de l’élection” in Daniel Boy and Nonna Mayer (eds), L’électeur a ses raisons, Paris, Presses de Sciences Po, 1997, pp. 219-37.
-
[33]
Nicolas Sauger, “Electoral agenda and issue voting” in Cautrès and Muxel (eds), The New Voter in Western Europe, pp. 153-70, here p. 168.
-
[34]
Rachel Gibson, Karolina Koc-Michalska, and Thierry Vedel, “Online campaigning in France 2007-2012: political actors and citizens in the aftermath of the Web 2.0 evolution”, Journal of Information Technology and Politics, 11(2), April 2014, 220-44.
-
[35]
Vedel, “La television”, 102.
-
[36]
Particularly because of what was felt to be a poor performance by Le Pen. “In total, around 14% of viewers who watched at least part of the debate changed their voting intention following it” (ibid., 112).
-
[37]
George Marcus, The Sentimental Citizen: Emotion in democratic politics, Philadelphia, Penn State University Press, 2002. See also Philippe Braud, L’émotion en politique, Paris, Presses de Sciences Po, 1996.
-
[38]
Daniel Boy and Jean Chiche, “The decisive influence of images” in Cautrès and Muxel (eds), The New Voter in Western Europe, pp. 79-100, here p. 80. For a presentation of debates about the different traits used to evaluate candidates, see Sylvain Brouard and Éric Kerrouche, “Candidates matter in parliamentary elections: France as a least likely case study”, Revue française de science politique, 63, 2013, 83-104, here 93.
-
[39]
Daniel Boy and Jean Chiche, “L’image des candidats dans la décision électorale”, Revue française de science politique, 57, 2007, 329-42; Roland Cayrol, “Le rôle des campagnes électorales” in Daniel Gaxie, Explication du vote, Paris, Presses de Sciences Po, 1985, pp. 385-417; Roy Pierce, “Candidate evaluations and presidential electoral choices in France” in Anthony King (ed.), Leaders’ Personalities and the Outcomes of Democratic Elections, Oxford, Oxford University Press, 2002, pp. 96-126.
-
[40]
Michael S. Lewis-Beck et al., French Presidential Elections, New York, Palgrave Macmillan, 2012.
-
[41]
Boy and Chiche, “L’image des candidats”, 340.
-
[42]
Richard Johnston and Henry Brady, “The rolling cross-section design”, Electoral Studies, 21(2), 2002, 283-95.
-
[43]
Ibid.
-
[44]
Henry Brady and Richard Johnston (eds), Capturing Campaign Effects, Ann Arbor, University of Michigan Press, 2006.
-
[45]
Bruno Cautrès and Anne Muxel, “Conclusion: the benefits of panel studies” in Cautrès and Muxel (eds), The New Voter in Western Europe, pp. 251-60, here pp. 255-6.
-
[46]
This filter allows us to eliminate from the sample users who visited La Boussole Présidentielle several times, either using the same computer or different computers, because it is extremely unlikely that users would have responded twice to all of the optional extra questions (on socio-demographics, candidate images, propensity to vote scores, etc.) in order to receive “voting advice”.
-
[47]
In all the longitudinal analyses included in this article, we weighted subsamples of users (each subsample was a sample made out of respondents over a period of three consecutive days) on gender, age, and education level in order to keep the socio-demographic structure of our sample constant across the campaign. This procedure neutralizes effects arising from socio-demographic variations in the sample from one observation period to another.
-
[48]
On socio-economic issues, during the campaign Kosciusko-Morizet spoke in favor of only modifying the ISF, while the other candidates wanted to abolish it. She also proposed establishing an unconditional basic income of€470 that would have incurred a temporary budget deficit. On social issues, she defended a cautious approach to environmental issues, supported equal rights for heterosexual and homosexual couples, and called for the introduction of proportional voting in electing députés, i.e. the members of the National Assembly.
-
[49]
The primary feature of Copé’s campaign was re-establishing authority in society. This expressed itself through strong positions on institutional issues, including government by decree and the reintroduction of a compulsory three months’ military service for all. Alongside this re-establishment of authority within the institutional sphere, he wanted to “bring order back to society” by banning clothing or symbols openly showing membership of a religion within public places, and by abolishing birthright citizenship. On economic issues, Copé proposed raising VAT by 3% in order to finance a cut in business taxes, and supported a reduction in unemployment benefits.
-
[50]
Larry M. Bartels, Presidential Primaries and the Dynamics of Public Choice, Princeton, Princeton University Press, 1988.
-
[51]
Until 16 November, no poll predicted any outcome except for a second round between Juppé and Sarkozy.
-
[52]
Johnston and Brady, “The rolling cross-section design”, 286-7.
-
[53]
Candidates’ image traits (presidentiality and empathy) were measured by the following questions: “In your opinion, how well do the following phrases apply to each of the following candidates: he/she is made of the right stuff to be a President of the Republic; he/she understands the problems of people like you” (answers on a scale from 0, “not well at all”, to 10, “very well”).
-
[54]
Bernard Denni, Olivier Lê Van Truoc, and Sandrine Astor, “Probabilités de vote: comportements et préférences des électeurs pourla présidentiellede2012”, proceedings ofSondages2012, studygroupfor the Société FrançaisedeStatistique, https://sites.google.com/a/iepg.fr/trielec/publications/le-vote/probabilite-du-vote, accessed 20 September 2013.
-
[55]
The second debate, which took place on 3 November, gave substantial room to a discussion of the candidates’ political positions and of strategies for alliances with centrist political forces, and particularly with François Bayrou’s Mouvement Démocrate (Democrat Movement, Modem) (10:15pm-10:45 pm during the debate). Juppé called for a large right and centrist alliance including LR, the UDI, Modem, and even “those disappointed with Hollandism”. In this he opposed Sarkozy, who called for a political alliance which excluded Bayrou and his party, who did not support him in the second round of the 2012 presidential election that Sarkozy lost to Hollande. While all the other candidates participated in the discussion, Fillon refused to join in “this punch-up”, which he saw as belonging to “the Republic of parties”. During this part of the debate, all of his contributions dealt with his desire to draw up “a contract, not with Bayrou or the center, but with the French people”, arguing that “in order to get France back on the right track, we need an energetic, radical program”, and that “once we’ve won the confidence of the French people, neither Bayrou nor anyone else will have anything to say”. By drawing the discussion away from electoral alliances, he aimed to take the moral high ground that he himself qualified as “a Gaullist stance above political parties”. Political commentators noticed Fillon’s posture, and in the days after the debate claimed that he had shown his presidentiality during the debate, and that he had become the third candidate in the primary, at Le Maire’s expense, who also battled for outsider status in this primary election.
-
[56]
Johnston and Brady, “The rolling cross-section design”.
-
[57]
In order to guarantee that a sufficient number of respondents were included in each observation period, the centering of the reference variables was carried out over three-day observation periods. Because our data covers a period of twenty-eight days, one of the observation periods includes four days, rather than three.
-
[58]
The alternative to this conditional change model would have been a linear regression model which did not include these pre-campaign reference variables. We tested this alternative model. The results show that the degree of variance explained is of the same scale as in the conditional change model, and that the significance of the explanatory variables is consequently the same for the two models. But the direction of the coefficients for the image variables is sometimes different because, without the pre-campaign reference variables used in the conditional change model, it is impossible to distinguish between the results of campaign effects and attitudes which were already established before the campaign. The absence of these variables would lead to incorrect conclusions about campaign dynamics during the primary.
-
[59]
A more detailed methodological presentation of the statistical model is available on the website of the Revue française de science politique.
-
[60]
These results support the work of Chiche and Boy (“The decisive influence of images”).
-
[61]
Although we know that, for large-scale data sets like the current one, statistical tests are typically significant. Among the candidates, Le Maire was closest to the average position on the left-right scale for our sample. It is therefore unsurprising that he managed to attract voters on both the center and the right.
-
[62]
The sample was made up of voters who were sure of voting in the first round of the primary of the right and center, with each three-day period of the data weighted based on the average values of the sample for gender, age, and education level. * p < 0.05; ** p < 0.01; *** p < 0.001.
-
[63]
Recall that La Boussole Présidentielle was put online on 24 October. The data collected, therefore, only measure the effects of the campaign during its final four weeks.
-
[64]
La Manif Pour Tous (Demonstration for All) is the name of the social movement that has emerged out of the protests against the same sex-marriage bill adopted under Hollande’s presidency in the spring of 2013.
