Presidential Election 2027Probabilities of winning, recalculated every day
FRRUEN

First- and second-round polls

Each line links to its source. Gaps between pollsters (house effects) are estimated by the model and removed before averaging.

For each poll: the pollster, fieldwork dates, sample size, client, method and the notice of the Commission des sondages, when this information exists in our sources. "Not given" means the source consulted does not provide it; the original record, reachable through the link, prevails. The model's figures are not polls.

Polls over time

One dot per first-round poll (raw published value). Line: the model's median estimate, after correcting for house effects. Candidates in alphabetical order.

The chart shows all first-round polls; the filters below apply only to the table.

How to read this chart. Each dot is a first-round poll, at the value published by the pollster; the line of the same colour is the model's median estimate for that candidate. Hover over a dot to see the pollster, the date and the sample.

Learn more

The gap between a dot and the line comes partly from house effects, which the model estimates and corrects, and from sampling error. A single dot is not a trend: only the line summarises all the polls. Candidates are chosen with the buttons, in alphabetical order.

Polls

229 polls · 31 Mar 2023 → 22 Oct 2026

Reset

How to read this table. Each line is a published poll: pollster, fieldwork dates, sample size, then the score of each candidate tested. A dot (·) means the candidate was not tested.

Learn more

The client, the method and the link to the notice filed with the Commission des sondages are shown on a poll when the source gives them; otherwise one "not given" line above the list says so, instead of repeating it on every row. Second-round head-to-heads are listed apart, one card per pollster, with every pair tested. The filters let you keep only one pollster, one round or one period.

Client, Method, Notice of the Commission des sondages: not given.

1st round

PollsterFieldworkSample Le PenGlucksmannAttalMélenchonRetailleauZemmourTondelierDupont-AignanSource
Ifop 30 Mar–31 Mar 2023 1105 31.0··22.0·6.0·3.0 notice table
Ifop 30 Mar–31 Mar 2023 1105 32.0··20.0·6.01.03.0 notice table
Ifop 30 Mar–31 Mar 2023 1105 35.0··20.0·7.02.04.0 notice table
Ifop 30 Mar–31 Mar 2023 1105 36.0··20.0·7.01.03.0 notice table

House effects

Average gap of each pollster from the consensus, in points, after regularisation. Positive: the pollster gives the candidate more than the average.

How to read this table. Each line is a pollster; each cell, the average gap between its results for a candidate and the weighted consensus, in points. A plus sign means a score usually higher than the consensus.

Learn more

This gap (pollster bias) can come from the method, the mode of interviewing or the weighting; it says nothing about the pollster's intentions. The model corrects part of it, with caution when the pollster has few polls.

PollsterLe PenGlucksmannAttalMélenchonRetailleauZemmourTondelierDupont-Aignan
Cluster17 −1.9+0.10.0+2.10.0−0.3−0.6+1.0
Harris Interactive −0.1−0.70.0+1.10.0+1.2−0.5−0.2
Ifop +0.8+0.90.0−0.80.0−1.0−0.3−0.5
Ipsos −1.1+0.20.0−0.60.0+0.2+0.4−0.2
Odoxa +1.1−0.70.0+1.10.0+0.2+0.1−0.2
OpinionWay +0.5+0.60.0−1.60.0−0.2−0.1−0.2
YouGov +0.6−0.30.0−1.30.00.0+1.0+0.4