Seven weeks after the re-election of Emmanuel Macron as President of the Republic, the French are called to the polls on June 12 and 19 to elect their deputies. As for the presidential election, it will be necessary to wait for the closing of the last polling stations, at 8 p.m. Sunday, to have the estimates of the results of the first round.
If the estimates of the balance of power between the major political parties, expressed in percentages, are rather reliable, the projections of the number of potential seats for each party are much less so, on the evening of the first round, as the outcome of the second round depends situations specific to each of the 577 constituencies.
Estimates, projections in seats and final results: The Decoders of the World take stock of the various figures circulating on election day.
At 8 p.m. sharp, the last polling stations in major cities close their doors, lifting the ban on the dissemination of results. Most of the media then want to give the French people an idea of the outcome of the ballot. For this, polling institutes make estimates based on the results of a defined number of polling stations that they consider representative of the vote of the French people. Geographic (rural municipalities, small or large towns, etc.) and political (left or right polls, reversal or consolidation of trends observed in the previous election, etc.) particularities are taken into account to build a statistical model.
Like IFOP or Harris Interactive, Ipsos – which produces the Ipsos-Sopra Steria estimate for various media – uses this method. The institute sends its investigators to the 580 sample offices. Each of them attends the counting and transmits:
The institute centralizes all these feedbacks and calculates the estimate of the final result. This is not a simple count: in the case of Ipsos, 70% of the offices in the sample close at 6 p.m., around 10% at 7 p.m. and almost 20% at 8 p.m. Knowing that it takes about an hour to count all the ballots of a polling station (less if abstention is high), the institutes do not have, at 8 p.m., either the partial results of the polling stations that have just closed or even of all the complete results of those of 7 p.m.
Most offices closing at 6 p.m. for these legislative elections – instead of 7 p.m. during the presidential election – the institutes will have more time to refine their estimates.
To produce their figures, the institutes use algorithms that take into account feedback from polling stations the same evening, but also data from previous elections. Each time counts are added, the algorithms refine the estimates.
This system has a weakness since it relies on the voting dynamics observed in the offices which have already closed, to project them on those which close at 8 p.m. However, the same dynamics are not always at work. “Let’s imagine that in the offices that close at 6 and 7 p.m., the voters vote 10 points more than in 2017 for the parties of the Nupes [New People’s Ecological and Social Union] and that the evolution is completely opposite in the large metropolises, the estimate would be wrong,” admits Jean-François Doridot, CEO of Ipsos Public Affairs. While this scenario remains “extremely rare”, it occurred in the first round of the presidential election: estimates of Jean-Luc Mélenchon’s score were around 18% at 8 p.m. before rising during the evening with the results of the big cities.
“Very, very fragile”, “perilous”, “to be taken with great caution”… all the pollsters we interviewed agree on the limits of the projections in seats broadcast on the evening of the first round. “It’s as if on the evening of the first round of the presidential election, we were trying to give the score for the second round,” said Jérôme Fourquet, director of the opinion department at IFOP. He acknowledges that there is “very strong media and political pressure for there to be screenings in seats” despite limited reliability.
Unlike the national estimates in percentages, which are based on the actual results of hundreds of polling stations, the projection in seats uses opinion polls to try to give an image of the future Hemicycle. Harris Interactive surveys 7,000 people on election day to “establish a form of sociology of voting and [to identify] potential voting behaviors in the run-off.”
“We ask people about the second round without having the result of the first, we do not yet know the duels or the triangular ones in their constituency”, comments Jean-Daniel Lévy, director of the political and opinion department at Harris Interactive. Respondents are asked about several second-round hypotheses, which allows the institutes to calculate possible vote transfers.
To give an idea of the number of seats that each political formation could obtain, the pollsters take as a basis the national estimate of the results of the first round and project it in each of the French constituencies. This gives the 577 possible posters of the second round on which they model the results of their opinion polls, to estimate the vote transfers of the unqualified candidates and sketch the possible National Assembly at the end of the second round.
Each party is assigned a range of seats, but the margin of uncertainty remains too large. “When you vary the tendency of a political force by one or two points in the first round, you can tilt thirty to forty seats,” explains Jérôme Fourquet.
The heads of the institutes interviewed all recognize the limits of these figures, which do not take into account the particularities of each constituency, local figures, dissidents and campaign dynamics between the two rounds.
On the evening of the second round, these projections should be more reliable, without being exact to the nearest seat, since the respondents will, this time, have been questioned on known configurations and no longer simply hypothetical.
The Ministry of the Interior, in charge of organizing the elections, begins to broadcast the results by municipality from 8 p.m., then it updates its publications continuously. At the beginning of the evening, only the results of the smallest municipalities, quickly stripped, are available.
After the closing and counting of nearly 70,000 polling stations, the results are sent to the prefectures, which themselves send them to the Ministry of the Interior. These results are then published as they become available. Even if the high rate of abstention usually observed during legislative elections makes the operation faster, the results of all the constituencies are generally not known until late at night from Sunday to Monday.
The results are considered final once validated by the Constitutional Council, after examination of potential cases of irregularities.