THE MODEL

The division of the electorate into five large blocs that could obtain a respectable proportion of the vote in the first round of the 2022 French presidential election requires a methodological approach that allows the simultaneous prediction of multiple outcomes. We mobilized the SUR approach to estimate the vote share of the five blocs in the first round, which then were used to predict the vote share of the two qualified candidates in the second round. The SUR model was used because it considers the possibility that the determinants of voting for one candidate also may determine the vote for other candidates. Contrary to the ordinary least squares approach, the SUR model does not assume that the error term of an equation (i.e., roughly representing the relevant factors omitted from the analysis) is uncorrelated with the error term of the other equations—a highly improbable possibility when using election-results data (Mongrain Reference Mongrain2021; Timm Reference Timm2002; Tomz, Tucker, and Wittenberg Reference Tomz, Tucker and Wittenberg2002).Footnote ¹

Our SUR model of the presidential election was estimated for the 1965–2017 period (i.e., 10 presidential elections) and included five equations. The specifications for these equations were derived from what is referred to as “synthetic” models—first applied to the French forecasting case by Lewis-Beck and Dassonneville (Reference Lewis-Beck and Dassonneville2015)—because they combine both “structural” determinants of the vote (e.g., lagged measures of economic performance and incumbent popularity) and vote-intention data that serve as a catch-all indicator capturing mostly campaign-related events. The first two equations are described as “conventional” because they provide forecasts for the incumbent-party candidate (INC) and the “natural” opposition to the incumbent (OPP).Footnote ² The third equation (CHAL) considers the rivalry between the Traditional Right or Centrist incumbent and the “challenger.” According to our hypotheses, the Republicans’ candidate, representing the Traditional Right, would be Macron’s challenger in 2022. Under the Fifth Republic, the Conservative or Centrist challenger has won only twice against the incumbent: Valéry Giscard d’Estaing against Jacques Chaban-Delmas in 1974 and Jacques Chirac against Édouard Balladur in 1995. The fourth equation (FR) provides a forecast for the RN and the Far Right in line with the voting function created by Jérôme and Jérôme-Speziari (Reference Jérôme and Jérôme-Speziari2003). The fifth model (DR) is a simple equation estimating the vote share of the Diverse Right (divers droite), a constant in French politics that allowed us to estimate the electoral potential of Éric Zemmour.Footnote ³ To summarize, the first three models (i.e., INC, OPP, and CHAL) are based on contrasts between candidates and their “natural” or “traditional” opponents, whereas the fourth and fifth models (FR and DR) oppose “outsiders” (i.e., never-elected candidates) to all other candidates in a given election.

The incumbent’s party model (INC; see table 1, column 1) is faithful to the government-accountability hypothesis (Key Reference Key1966) because it is founded on the assumption that voters look primarily at the incumbent’s record. Therefore, a high level of unemployment (U6) should penalize the incumbent’s party. Similarly, the higher the proportion of people who favorably evaluate the president’s job—as measured by the average popularity of the incumbent president in the quarter before the election (INCPOP3)—the higher the vote share of the outgoing president’s party should be.

Table 1

SUR Model for French Presidential Elections, 1965–2017

Notes: Unstandardized coefficients and t-stats (in parentheses) are reported. Significance levels: *p<0.10; **p<0.05; ***p<0.01 (two-tailed).

The natural opposition (OPP; see table 1, column 2) to the incumbent is the Left when the Center or Conservatives are in power, the Center and Conservatives when the Left is in power (as in 2002 when the Socialist Party, or PS, and its allies held a majority in the National Assembly), and the Conservatives alone against the Left (the PS) in 2017. In the case of the opposition, because it has no record, voters must balance the expected gains (or costs) of keeping the incumbent in power against the potential costs (or benefits) of placing the opposition in charge. Consequently, this calculation should be partly forward looking. This prospective judgment can be approximated by using the vote intentions in favor of the opposition in January of the election year (OPPOLL). The opposition also will be favored by deteriorating economic conditions, which is measured by the change in unemployment (ΔU) between December t–1 and December t–2, where t is the election year. Under the Fifth Republic, the opposition also experienced an electoral boost from cohabitation periods (COHAB) in 1986–1988, 1993–1995, and 1997–2002. It also is noteworthy that Conservatives have never experienced an electoral boost by being in the opposition (CONSOPP).

Challengers (CHAL; see table 1, column 3) also will be judged prospectively using vote intentions in their favor (CHALPOLL). The question for voters is whether they will “do better” than the partners they supported. However, the more popular the incumbent, the more challengers should be penalized (INCPOP3). The challenger equation also includes two dummy variables. The first represents the electoral premium obtained by Macron (as the challenger) after François Hollande’s decision not to seek a second term (DUMHOLL). The second variable marks the loss suffered by Centrist challenger François Bayrou with the disappearance of the Union for French Democracy (UDF) in favor of the creation of the MoDem in 2012 (MODEM).

The Far Right model (FAR; see table 1, column 4)Footnote ⁴ has a notable economic-vote feature. The higher the unemployment rate (U6), the better the RN should do at the polls and vice versa. RN voters also are sensitive to the personality of their candidate because they seek a “strong” leader (Chaillou Reference Chaillou2016). The DUMMLP variable indicates the electoral premium of “personalization” experienced by Marine Le Pen following the 2012–2017 reorganization of the movement through the policy of dédiabolisation (loosely translated as “undemonization”) adopted by the party. Furthermore, three dummy variables were added to the model. The first (DUMTIXIER) noted the candidacy of Far Right candidate Jean-Louis Tixier-Vignancourt in 1965, which is not comparable to the frontiste candidacies. The second dummy variable (DUMLP81) noted the average loss suffered by the Far Right in the period following the inability of Jean-Marie Le Pen to participate in the 1981 election after failing to obtain the required number of signatures from local representatives to run for president (Dubois Reference Dubois2021). Finally, the third dummy variable (COHABRP) noted the electoral premium for Far Right candidates in periods of cohabitation between a right-wing president and a left-wing prime minister, as was the case in 2002.

The fifth model predicts the score of the Diverse Right (DR; see table 1, column 5),Footnote ⁵ which sometimes has undermined the ability of Conservative or Centrist candidates to unite in the first round. This notably was the case of Jean Royer in 1974, Philippe de Villiers in 1995, and Nicolas Dupont-Aignan in 2017. Paradoxically, however, the Diverse Right candidacy also can be positive for right-wing incumbents in that it neutralizes votes that could have gone to the Far Right in the first round. The cost–benefit balance then must be considered. The main variable of the DR model is the vote intentions for the various right-wing candidates as measured by the polls in January of the election year (DRPOLL). An additional dummy variable measures the electoral cost suffered by the Diverse Right when a Conservative is holding power (DUMCONS).