Methodology

Sources

Deputies, their mandates, their groups and the public votes come from the open data of the National Assembly (data.assemblee-nationale.fr), reused under the Licence Ouverte. They are imported twice a day; the Sources page shows the date and state of each import.

The match between towns and constituencies is derived from the official results of the 2024 legislative elections published by the Ministry of the Interior: each polling station appears there with its town and its candidates, which belong to one constituency. Postal codes come from La Poste’s official base.

Vote positions

For each public vote, the Assembly publishes four nominative lists: for, against, abstention, non-voting. The site shows them as they are, noting a vote by delegation and, for a non-voting deputy, the reason when it is given (member of the Government, President of the Assembly, presiding the sitting).

A deputy sitting on the day of the vote who appears on no list is shown with no position recorded. This is neither a vote against, nor an abstention, nor a proven absence: the data does not allow to say more.

For a motion of censure, only votes for are recorded. Not being on the list does not mean voting against.

After a vote, deputies sometimes declare they meant to vote otherwise (corrections). These declarations are shown apart; they never change the official result.

The group shown for a vote is the one the deputy belonged to on the day of the vote, from their dated memberships, never their current group.

The colour dot beside a group is the colour the Assemblée nationale associates with it in its data (`couleurAssociee`). It helps to spot the group, whose name is always written; when no colour is published, no dot is shown.

Checks

At each import, the official totals of each vote are compared with the nominative lists; any gap is recorded and visible in the back-office. A sample of votes is checked against the National Assembly website; the check report is public in the project’s repository.

A failed import never replaces the data already in place.

Editorial content

Explanations of votes are written by people, reviewed, then published. An explanation is always separate from the official title, states that it was reviewed and says what the vote does not allow to conclude.

The site publishes no judgement on people: no score, no ranking, no comparison between statements and votes.

An AI tool may suggest drafts in the back-office; they are never published without a person rewriting or approving them, and never generated while visitors browse the site.

Filing by topic

Each vote is filed automatically from the title of the text under examination (government or private bill, resolution, motion of censure), never from the content of an amendment or the names of its authors: every vote on a budget bill is filed under taxation and economy, whatever the amendment. A text can belong to several topics.

The rules are public, dated lists of phrases, kept in the project repository. They are applied at every import, and the vote page shows an automatic filing. Procedural votes (suspending or extending a sitting) and texts no rule recognises stay unclassified.

A member of the team can confirm or withdraw a filing in the back-office; that choice prevails over the rules, and withdrawing a published filing is listed among the corrections.

The comparison tool

The comparison tool covers a selection of real votes made by public rules, with no human intervention: solemn public votes on a whole government or private member’s bill, contested (at least 50 votes against, or a rejection), only the latest vote on each text; from the most recent back, a vote is kept when it touches a topic that no vote already kept covers, up to 15 questions. Below 5 votes kept, no questionnaire is offered. Each question repeats the official title word for word, so the measure actually put to the vote. You answer for, against, or skip.

For each deputy, only the questions you answered for or against and on which the deputy voted for or against count. The agreement rate is the number of agreements divided by this number of comparable votes. Abstentions, non-voting, votes outside the mandate and missing positions are excluded and shown as such.

Below five comparable votes, no rate is shown or ranked. The number of comparisons always comes with the rate: 100 % over 5 votes is not 100 % over 15.

For a group, the position used is that of the majority of its members who voted for or against that day, with the numbers shown; in case of a tie, the vote is not comparable for that group.

This result describes proximity on this sample only. It is neither a general political score nor a prediction.

Two deputies are compared on every public vote where both voted for or against. Two groups are compared on every vote where the majority of each one’s members that day voted for or against: a tie in either sets the vote aside, and the numbers of each group are given for the votes where their majorities differ. In both cases, a rate is given only from five votes.

Limits

• Only nominative public votes are published as open data: votes by show of hands do not appear.
• The open data does not say to whom a delegated vote was given.
• The data can lag the sitting by up to a day.
• The match between towns and constituencies follows the boundaries in force in 2024.

Corrections

Any error can be reported from every page. Corrections made to published content are dated, explained and listed on the Corrections page.