Methodology

Every number in the Radar has a source, a calculation and a collection date. This page explains all three, and the limits of each number.

Sources

How we credit each source is on the licences and credits page.

Collection

• In plenary weeks, the data are updated every morning from Tuesday to Friday, so each result appears the day after the vote. Outside plenary weeks, a full collection runs every Sunday night, with dossiers, the agenda, written questions, news and Rising.
• Each edition is written from the collection named at its top; daily updates change the data pages, not the text of a published edition.
• The collection date is shown at the top of every piece.
• We respect the limits of each source: one request per second to the Parliament and to HowTheyVote, one request every ten seconds to GDELT.
• The Parliament completes and corrects its data in the days after a vote. Sessions from the last 21 days are collected again every time, and any change follows the correction rules below.
• Each collection is kept as a dated copy, so any number can be traced back to the data it came from.

What goes to a vote

From the Parliament's plenary agenda (the activities foreseen for each sitting day), the Radar lists the laws under the ordinary legislative procedure and the annual budget. For each one, it shows the rapporteur in the committee responsible, with their group, and the political groups that follow the file through a rapporteur or a shadow rapporteur, from the Parliament's data on each procedure. The agenda is a draft until the session opens and can change. Who follows a file says nothing about how a group will vote.

What was voted

The Radar lists every item voted in plenary (a report, a resolution, an agreement or a procedural request) with its official title, the reference of the document and the result.

One item can have dozens of votes: amendments, split votes, the text as a whole. We report the vote that decides the item, the final vote, chosen by these rules, in this order. Lapsed or withdrawn votes never count.

1. Second reading. If a motion to reject is adopted, the Council position is rejected. If an amendment is adopted (this needs an absolute majority), the Council position is amended. If nothing is adopted, the Council position is approved. The vote shown is the motion to reject, if one was put to the vote, so a vote "for" is a vote to reject the Council position; otherwise, the last amendment adopted or, failing that, the last valid vote.
2. Motion to reject. If one is adopted, the item is rejected. The vote shown is that motion, so a vote "for" is a vote to reject.
3. The text as a whole. Otherwise, the last valid vote on the whole text, leaving out procedural votes: referral back to committee, postponement, motions to reject and votes on whether to vote on amendments.
4. Provisional agreement. The vote on the agreement reached with the Council.
5. Otherwise, the last valid vote that is not procedural; then an adopted postponement or referral; then the last procedural vote.

The possible results are: adopted, rejected, adopted without a vote, vote postponed and referred back to committee; at second reading, Council position approved, amended or rejected. When the Parliament adopts its amendments and refers the text back to committee to negotiate with the Council, the result is "referred back to committee" and the vote shown is the vote on the amended text.

Every roll-call vote has its own page. When a vote is not the final vote of its item, such as a proposal for rejection that failed, its page says which vote it was and links to the final vote, so the result of one vote is never shown next to the count of another.

How each group voted

• Roll-call votes only. Only a roll-call vote records how each MEP voted. A vote by show of hands gives only the result; an electronic vote without names also gives the count. Neither gives a breakdown by group.
• Which votes. The main votes identified by HowTheyVote, plus every final vote taken by roll call. Votes on single amendments are not in the weekly tables.
• Four counts per group: for, against, abstention and did not vote. Each MEP counts in the group they belonged to on the day of the vote.
• Group position. The option (for, against or abstention) with most votes among the group's members who voted. When two options tie at the top, the group is shown as split. The position is calculated from the votes; it is not the group's official voting instruction.
• Cohesion. The share of the group's voting members who followed the group position. 0.95 means that 95% followed it.
• Agreement index. The largest option minus half the sum of the other two, divided by the number of members who voted (Hix, Noury and Roland, 2005). 1 means that everyone voted the same way; 0 means an even split into three. It allows comparison between groups of different sizes.
• Abstention and absence. An abstention counts as a position. Members who did not vote are left out of both measures and shown as "did not vote". This count does not distinguish absence, justified absence and presence without voting.
• Non-attached members are not a political group: we show their four counts, without a position, cohesion or agreement index. An MEP who had not yet joined a group on the day of the vote is counted apart.

Groups always appear in order of size, from the largest to the smallest, as the Parliament lists them, with non-attached members last. The order is the same in every table and every text, whatever the result of a vote, and it is not a ranking. If a change in seats changes the order, we update it and date the change on this page.

Seats: the Parliament's full list of MEPs, 3 October 2026.

Majorities

For every vote decided in plenary we show which political groups were on the winning side: the groups whose position (the option chosen by most of their members who voted) matched the result. A group is decisive when, without its votes on the winning side, the result would have been the opposite; we calculate this only for votes decided by a simple majority.

Some votes need a majority of all MEPs, such as an objection to a Commission measure or an amendment at second reading. When one of them is rejected with more votes for than against, we say so and give the majority it needed, as the official minutes record it. A few other votes, such as a motion of censure, need a different majority; when the type of vote is not clear from the data, we say only that it needed a qualified majority, as recorded in the official minutes.

The Majorities page shows, for the year and for the last 90 days, how often each group was on the winning side and how often each pair of groups took the same position. Groups appear in the Parliament's order, by size; the page does not rank them by these rates. Non-attached members are not a group and are left out. We use the official names of the groups and no other label.

Closest votes

The closest votes are the main votes decided by a simple majority with the smallest difference between votes for and against. Each edition shows the three closest of the week group by group. The site lists the closest of the last session and of the year, each with its decisive groups and the national delegations of two or more voting members that would have reversed the result had they voted with their own group. Votes that need a majority of all MEPs are left out of these lists and reported apart.

Comparing two MEPs

Choose two MEPs to see in how many votes they voted the same way, out of the votes in which both took part, and the list of votes in which they differed. The comparison gives counts only, with no score.

How each country voted

• For each vote, the four counts for the MEPs elected in each of the 27 member states, and how each of them voted.
• Voted differently from their group. An MEP who voted for, against or abstained, differently from the position of their own group on the day of the vote. Not voting does not count. We do not calculate it when the group is split, nor for non-attached members.
• National delegations. A national delegation is the MEPs of one country within one political group. It voted differently from its group when the option chosen by most of its members who voted is not the group position. If the delegation is itself split, it is not counted. The note in each edition lists first any delegation whose votes changed the result (had it voted with its group, the vote would have gone the other way); then the largest delegations, by members who voted, that voted the opposite of their group (for where the group voted against, or against where it voted for), up to ten in all. Only delegations where at least two members voted count; a single MEP who voted differently from their group appears in the view of their country. The full list is on each vote page and in the data files.
• The count does not judge anyone. Voting differently from the group's majority can follow the line of a national delegation, a free vote or a personal position.

Country and MEP pages

Each of the 27 countries and each MEP has a page with every vote of the year and a few numbers. All of them come from the same roll-call votes as the rest of the Radar.

• Winning side. Of the votes for and against, the share that went the way of the result: for when the text was adopted, against when it was rejected. Abstentions and absences are left out.
• Alignment with a group. The share of votes (for, against or abstention) that matched the position of that group in the vote. Votes in which the group was split do not count, and non-attached members have no position.
• Cohesion of a country. In each vote, the share of the country's MEPs who voted and followed the option chosen by most of them; the page shows the average for the year.
• Majority of a country. The option with most votes among the country's MEPs who voted. A tie at the top is shown as split.
• MEPs of a country. Those who appear in the most recent vote, in the group they belonged to on that date.
• Participation of an MEP. Of the votes the Radar follows in which they were already an MEP, the share in which they voted. It is not attendance at the session, and it does not show justified absences.
• Delegations off the group line. How many times in the year a delegation of the country, with at least two members voting, voted by majority differently from its group.

Each number is shown next to the median of the group, of the country or of all MEPs, as context. Lists of MEPs are in alphabetical order by surname; we do not rank MEPs by these numbers.

Where the votes came from

Each MEP page says how the MEP was elected in June 2024: the party list, its share of the vote and its seats, and the electoral rule of the country. Where we have them, it also gives the constituency, the MEP's place on the list and their personal votes. A map shows where the list won its votes.

The votes for each list come from BLUE_EP, under CC BY 4.0: Hublet, F. (2024). BLUE_EP: A Dataset of Municipality-Level Results of European Parliament Elections. Codebook v1. Paris: Groupe d'études géopolitiques. 10.5281/zenodo.14569325. BLUE_EP gives the votes for each list and none for individual candidates. Personal votes come from the official results of the countries that publish them by area.

Countries elect their MEPs under one of three rules, as summarised by the European Parliamentary Research Service, and each page says which one applies:

• Closed list, in France, Germany, Hungary, Portugal, Romania and Spain. Voters choose a party list, and the party sets the order of its candidates. The map shows the list.
• Preferential vote, in 19 countries: all except the six above, Ireland and Malta. Voters choose a list and can also mark candidates on it; how many, and how much the marks count, varies by country. The map shows the list. Where the country publishes each candidate's votes by area, the map has a second layer, Personal votes: the share of the valid votes in each area that carried a personal vote for the MEP. In Czechia, for instance, voters can mark up to two candidates on the list, and candidates with personal votes from at least 5% of the list's votes move to the top of the list (Act No. 62/2003 Coll.).
• Single transferable vote, in Ireland and Malta. Voters rank candidates in order of preference. BLUE_EP gives first preferences only, added up by party, and the map shows those.

The map colours each area by the list's share of the valid votes cast there, which are the votes cast minus invalid and blank ballots. We sort the areas by that share and cut them into five bands with about the same number of areas, so the darkest band holds the fifth of areas where the list did best compared with the rest of the country. The legend gives the limits of each band, which change from list to list, and each area shows its share and its votes. Each area counts once, whatever its size, so two bands can hold very different numbers of voters. Every party's map uses this method, these colours and this wording. Beside the map we list the five areas where the list did best, among those with at least 1,000 valid votes.

Results are by local administrative unit, as Eurostat defines it: in most countries the municipality, in Portugal the parish. Four countries publish results only for larger units. Greece reports its 392 municipalities (dimoi), each of which groups several of the local communities used for statistics. Slovenia reports 89 electoral districts and Ireland its 3 constituencies, and Malta votes as a single constituency. There is no map for Malta, or for Denmark, where we do not yet have municipal boundaries we can reuse; the pages of their MEPs keep the account of how they were elected. Votes with no place on the map, such as those cast abroad, are left out of it, so the total on a map can be lower than the national result.

In Portugal, results by parish exist only in the provisional count published on election night by the Ministry of Internal Administration. The official result, published by the National Elections Commission (Mapa Oficial n.º 4/2024), differs from it by less than 0.2% for each of the eight largest lists, which include all the lists that won a seat.

An automatic test compares the votes on each map with the official result of the country, and any difference is stated under the map.

Dossiers

• One page per procedure, built from the Parliament's data, with how the groups voted from HowTheyVote: official title, type, stage, committee responsible and committees for opinion, rapporteurs and shadow rapporteurs with group and country, events, documents, plenary dates, the items voted, how the groups voted and a link to the Legislative Observatory.
• Which dossiers: procedures opened or voted in the current year, those on the plenary agenda from one week back to five weeks ahead, and reports tabled in the current year but not yet voted.

Rising

Rising shows which dossiers moved fastest on their way to plenary in the past week. Plenary debates and votes are not counted: they have their own sections in the Radar. It is calculated once a week, for the last full week from Monday to Sunday, so during a plenary session the site shows the previous week's Rising.

For each dossier and each week, from Monday to Sunday, three components are compared with the dossier's own average over the previous four weeks:

• Procedure steps (Parliament data). Steps that bring a decision closer count more: a report adopted in committee, a provisional agreement approved in committee, an interinstitutional negotiation, a mandate to negotiate, a report tabled for plenary. Other steps, such as a referral to committee, a draft report or a committee opinion, count less. Plenary debates and votes, signature and publication in the Official Journal do not count. The same type of step on the same day counts once.
• Written questions linked to the dossier (Parliament data). The Parliament's data do not say what a question is about, so we link a question to a dossier only when its title contains the distinctive words of the dossier's title. This links few questions (2.6% of those asked in 2026), and 13 of 15 links checked by hand were right. The Parliament publishes questions with a delay, so this component looks at an earlier week.
• News coverage. The number of distinct outlets in the dossier's story in In the news (GDELT). Weeks in which GDELT or our own collection covered less than 60% of a normal week are left out. In such a week, Rising uses only the Parliament components and carries the beta label.

The components are combined into a score from 0 to 100. A dossier enters only if something happened in the week and it moved faster than its own recent average. Dossiers with the same score share a rank.

The score is absolute: it does not depend on the top dossier, so a quiet week looks quiet and scores can be compared from one week to the next.

Until the news component has four weeks of collection in which GDELT worked normally, Rising is in beta: it uses only the Parliament components and carries the beta label.

We tested Rising on the nine weeks from 10 August to 5 October 2026. Of the dossiers in the weekly top ten, 46% went to plenary (a debate, a vote or a place on the agenda) in the following four weeks, against 22% of all the dossiers that moved. Counting plenary debates and votes did not improve this, and it filled the top ten with the items just voted, so they were left out. The news component has not been tested yet; it will be before Rising leaves beta.

Rising has limits. It counts steps and says nothing about importance: a large dossier that does not move does not appear. It compares a week with the four before, so a dossier that moves at an even pace does not rise. A dossier that goes from committee to plenary vote in the same week appears only if it had committee steps that week. The Parliament publishes its data in stages; a published Rising is not changed when late data arrive, and the data file is updated with the next collection. The news component measures volume only and does not read tone.

In the news

In the news lists the stories about the Parliament's votes, dossiers and MEPs that were most covered by news outlets in the week. The list comes from the count below; we add no story to it and choose none by hand.

• We use articles found by GDELT, which reads outlets in 65 languages, that match a dossier through search terms written by hand from its official title, or that name an MEP together with the European Parliament. The terms describe the subject of the dossier, never a position on it, and they are reviewed every week.
• All the articles linked to the same dossier or MEP from Monday to Sunday form one story. The same article published by several outlets within 72 hours counts once.
• To find what is hot, each story is scored from the outlets that covered it, the EU countries of those outlets and how much its coverage grew compared with the four previous weeks in which GDELT worked normally. A story enters only with at least 3 outlets from at least 2 EU countries. The Radar shows the 10 highest scores and the newsletter the top 3. Ties are broken by the number of countries, then of outlets.
• Each story shows a short summary in English that we write, saying what the coverage is about, and for each article the outlet, its country, the language, the date and the link. We never show the text, the headline or the image of an article: the small square next to each story is drawn by the Radar from the count, with the number of articles on each day of the week. We draft the sentence with AI tools; before it goes out, a person opens the linked articles and checks it, as with every other part of the Radar.
• Parliament press releases appear apart and are not part of the count. A release can show its own photo only when the photo is credited to the European Union, and then with its credit; photos from any other source are never shown.
• We remove only a wrong match: an article that is not about the dossier or the MEP it was linked to. An article that mentions the dossier only in passing is not about it. After the removals the count is done again, so a story can drop below the minimum or change place, and every removal is logged. Outlets under EU sanctions (Council Regulation (EU) No 833/2014, Annex XV) are never listed. No outlet is added or removed for its political line.
• When GDELT does not respond, or when GDELT or our own collection covers less than 60% of a normal week, the list for that week is not ranked, and the page says why. A normal week is the average of the last four weeks in which GDELT worked normally. We do not complete the list by hand.

GDELT does not index every country and language equally, so countries with fewer indexed outlets weigh less. A large volume of coverage says nothing about importance, and we do not measure tone. Matches made by search terms can be wrong; we check a sample by hand every month and publish the share that was right.

Vote of the day on social media

In plenary weeks, on the day after each voting day, we publish one short video on social media with the result of one vote of that day. Each MEP is a dot in the hemicycle, placed as in the Parliament's seating plan, and the bars by group follow the order by size used across the Radar. The vote comes from these steps, in order, among the main votes of that day:

1. The vote on the dossier of the top story in the most recent In the news ranking published before the video.
2. If no vote of that day is linked to that story, the votes on the subject most watched that week. Every Monday we search TikTok and Instagram for the most-watched short videos about the Parliament and EU politics. Each video is assigned one HowTheyVote topic through a published keyword list, and the topic with the most views qualifies the main votes of that day tagged with it.
3. If several votes qualify, the closest of them; if none does, the closest main vote of that day. Closest means decided by a simple majority with the smallest gap between votes for and against, as under Closest votes.

Checks before publishing

• An automatic test compares the total of every vote between the Parliament and HowTheyVote and checks that the group counts add up to the total. If any test fails, nothing is published.
• When the two sources differ, that count is not published until the difference is explained. On the day of a vote, the Parliament's count can arrive after HowTheyVote's; until it does, the count is marked as provisional.
• The numbers on country and MEP pages are checked against a second calculation made by another route, and the country counts against each vote.
• A person checks the numbers against the data tables, the names and the facts before publishing.

Checks run before launch, in October 2026:

• 331 of 331 votes in our 2026 set (the main votes on HowTheyVote and the final votes by roll call), up to the session of 5 to 8 October, had the same total at the Parliament and at HowTheyVote, and in all of them the group counts added up to the total.
• The 26 main votes of the session of 5 to 8 October on HowTheyVote matched our final vote for each item.
• The counts of 3 votes matched the official summaries in the Legislative Observatory, and 3 dossiers matched their Observatory files on committees, rapporteurs, shadow rapporteurs and dates, with the differences described under limits.

Limits

• The breakdown by group exists only for roll-call votes.
• The weekly tables do not cover committee votes, votes on single amendments or secret votes.
• On the day of a vote, the Parliament publishes its data in stages, and the title of an item sometimes appears first only in French. Until the official English title arrives, we use the official English title of the dossier or, failing that, the title used by HowTheyVote, marked as provisional.
• A resolution can be based on several motions tabled by different groups. In lists, the reference shown may be the first motion in the Parliament's list, which is not always the one adopted; the dossier page lists them all.
• The stage of a dossier in the Parliament's data is the reading stage, not its status: a finished dossier can still show its first reading. We show the last step and its date.
• A committee that declined to give an opinion can still appear as a committee for opinion.
• Some dossiers arrive without a committee responsible, and rapporteurs who have left the Parliament can appear without a group. Some plenary documents have no procedure in the Parliament's data, so a tabled report can be missing from the dossier list.
• Some new dossiers arrive without an English title. We show the title in the language available and mark it as provisional.
• HowTheyVote is an independent project. If it stops, the breakdown by group comes from the Parliament's own roll-call records, and the piece says so.

Corrections

We correct errors within 48 hours, with a dated note at the top of the piece, and list every correction on the corrections page. When the Parliament corrects its own data after we publish, we update the piece and add a dated note. Write to radar@parleon.eu.

Data to download

The data behind the Radar are available in CSV and JSON. Each file states its sources, its licence and its collection date at the top. For Rising and In the news, the files give what the site shows for each entry. See licences and credits.