Just three weeks after Emmanuel Macron dissolved the Assemblée Nationale after his party’s debacle in the European elections, the French are once again being called to the polls on Sunday, June 30.
While estimates of the general power balance between the major political parties − expressed in percentages − are generally close to reality, projections of the number of potential seats for each party are much less accurate on the evening of the first round, because the outcome of the second round depends so much on the specific situation in each of the 577 constituencies.
Le Monde‘s Les Décodeurs team take a look at the different figures you’ll see on election day, from estimates and seat projections to voter turnout.
Voting estimates at 8 pm
At precisely 8 pm on the day of the vote, the last polling stations in major cities close their doors, and the ban on broadcasting results is lifted. At this moment, most of the media want to give their readers the results of the election, with the national percentages obtained by each party or coalition. To do this, polling organizations make estimates based on the results of a defined number of polling stations they consider representative of the French vote.
The polling institutes IFOP, Harris Interactive and Ipsos − which produces the Ipsos-Talan estimate for various media − all use this method. Ipsos sends its surveyors out to the 600 polling stations that make up their sample. Each of them takes part in the count and notes:
- Information on the number of voters when the polling station closes, to estimate turnout;
- A partial result with the scores obtained for each candidate, at the end of the first two hundred ballots counted; and
- The full results at the end of the polling station count.
The institute centralizes all this feedback and calculates the estimated final result based on a statistical model that takes into account geographical (rural areas, small or large towns, etc.) and political (traditionally left- or right-leaning posts, reversal or consolidation of trends observed at the previous poll, etc.) particularities.
It is not a simple count: In the case of Ipsos, around 70% of the sample polling stations close at 6 pm, 5-10% at 7 pm and the remaining 20-25% at 8 pm. Given that it takes around an hour to count all the ballots in a polling station (less, if the turnout is low), at 8 pm the institutes have neither the partial results of polling stations that have just closed, nor even the complete results of those that closed at 7 pm. Each time new counts are added, the algorithms refine the estimates.
The weakness of this system is that it relies on the voting dynamics observed in polling stations that have already closed, and projects them onto those closing at 8 pm. If the results of polling stations closing at 6 pm and 7 pm turn out to be 10 points higher for a party or candidate than in previous elections, but the dynamics are different in large cities closing later, the first estimate given at 8 pm is likely to be wrong before being refined later. This scenario is rare, but it did occur in the first round of the 2022 presidential election: Estimates of Jean-Luc Mélenchon’s score were around 18% at 8 pm, before rising over the course of the evening to approach 22% as results from major cities, where he scored higher, were taken into account.
Specifics linked to an expected increase in turnout
The method described above provides an accurate estimate of the national balance of power between the various parties and coalitions. However, it is only a very imperfect guide to the makeup of the future Assemblée, which will depend on 577 separate elections.
What’s more, the expected rise in voter turnout is likely to complicate matters considerably for pollsters. According to a June 27 Ipsos survey, between 61% and 65% of voters could go to the polls, compared with 47.5% in 2022.
Partner service
Learn French with Gymglish
Thanks to a daily lesson, an original story and a personalized correction, in 15 minutes per day.
Try for free
Mathematically, the higher the turnout, the more likely it is that three − exceptionally, four − candidates will be able to stand in the second round. The two leading candidates are qualified, as are those who have obtained a number of votes greater than 12.5% of registered voters. In 2022, due to low voter turnout, only eight three-way runoffs were possible in the 577 elections.
“The uncertainty of seat simulations increases on the evening of the first round with the number of three-way runoffs. For example, with the 2022 results, a turnout of 60% would have led to 120 three-way runoffs, and up to 200 with 65% turnout,” details Jean-François Doridot, director general of Ipsos Public Affairs.
Polling institutes have to incorporate these parameters into their sample of polling stations. “When you prepare for a first round, you’re already preparing for a second round,” said Doridot. “As we don’t know the results of the first round, we need to have polling stations that will reflect all possible three-way configurations in the second round, to be able to anticipate vote transfers.”
“Very, very fragile” seat projections
“Very, very fragile,” “perilous,” “to be taken with great caution”… All the pollsters we interviewed agree on the limits of the seat projections broadcast on the evening of the first round. “It’s as if, just after the first round of the presidential election, we were trying to predict the score of the second round,” explained Jérôme Fourquet, director of the IFOP opinion department, back in 2022. He acknowledged that there was “very strong media and political pressure to make seat projections,” despite their limited reliability.
Unlike national estimates in percentages, which are based on actual results from hundreds of polling stations, the seat projections are based on opinion polls. Harris Interactive interviews 6,000 to 7,000 people on polling day to “understand the sociology of the vote, voters’ motivations and potential electoral behavior in the second round.”
“We’re asking people about the second round without having the results of the first, so we don’t yet know the two-way or three-way contests in their constituency,” said Jean-Daniel Lévy, director of the politics and opinion department at Harris Interactive. Respondents are questioned on several second-round hypotheses, enabling the institutes to calculate possible vote carryovers.
“The fragility of the simulation in seats comes from the fact that it is based on information that is measured before the first round,” agreed Doridot of Ipsos.
To give an idea of the number of seats each political party could win, polling agencies take the national estimate of first-round results and project it onto each of France’s constituencies. This gives the 577 possible second-round lists of candidates on which they base the results of their opinion polls, to estimate the carryover of votes from unqualified candidates and sketch out the possible Assemblée Nationale at the end of the second round.
Each party is allocated a range of seats, but the margin of uncertainty remains too large. A change of one or two points in the tendency of a political force in the first round can swing 30 to 40 seats in the projected hemicycle.
The heads of the institutes interviewed all recognize the limitations of these figures, which fail to take into account the particularities of each constituency, local figures, dissidents and the campaign dynamics of the inter-tournament period.
Despite these limitations, all three polling firms published seat projections on the evening of the first round two years ago. Doridot points out that while Ipsos will provide seat-based projections on the evening of the first round, they will not offer any beforehand, “as the fragilities prior to the vote are too great.”
On the evening of the second round, these projections should, on the contrary, be more reliable, although not precise to the nearest seat, because this time respondents will have been asked about known configurations rather than merely hypothetical ones.
Official results continue late into the night
The Interior Ministry, which is responsible for organizing the elections, starts releasing results town-by-town at 8 pm, then updates its website continuously. In the early evening, only the results for the smallest towns, which have been quickly counted, are available.
Once nearly 70,000 polling stations have been closed and counted, the results are forwarded to the prefectures, who in turn send them to the Interior Ministry. They are then published progressively. The results for all constituencies are generally not known until late on Sunday night.
The results are considered definitive once they have been validated by the Constitutional Council, after examining potential cases of irregularity.