Methodology

Full transparency: this is how we calculate your political match.

1. Which candidates are in the test

Until the official list of candidates is published, the test includes the people who have declared their candidacy or been designated by their party and who meet at least one of these two conditions:

• They are backed by a party that has a group in the National Assembly.
• They have at least 2% for the first round in the published polls of two different polling institutes over the last four weeks.

We review the list with each new wave of polls, and the same criterion applies to everyone. Candidates who declare later are added as soon as they meet it.

The official list of candidates, drawn up by the Constitutional Council once the 500 sponsorships (parrainages) have been checked, is expected on 16 March 2027. From then on, the test will include all the official candidates, and only them.

After the first round, on 18 April 2027, the test will keep only the two candidates in the second round, on 2 May 2027.

2. Sources of information

The candidates’ positions are documented on the basis of public and verifiable sources, giving priority to the candidates’ own documents:

• The candidate’s programme for the 2027 presidential election, as each candidate publishes it
• Until then, the programme of the candidate’s party for the 2022 elections, marked as such on the candidate’s profile
• Votes of the candidate’s party group in the National Assembly, as recorded on its website (assemblee-nationale.fr), or in the European Parliament
• Public statements by the candidate, cited via the media outlet that reported them when there is no campaign document

3. Question design

Every question in the test meets these criteria:

• Neutrality: It does not mention candidates, parties or political figures
• Clarity: Accessible language, no jargon
• Relevance: Issues currently debated in French politics and in the presidential campaign
• Differentiation: It helps distinguish between the candidates’ positions

Response scale

We use a 5-point Likert scale:

4. Candidate positions

For each question, each candidate is assigned a position (1–5) based on those sources. Currently all positions have the same confidence weight (3) in the calculation: no candidate is treated differently through weighting.

• Positions with a linked source: They include the specific document that supports them, with its date
• Estimated positions: Inferred from the candidate’s statements, record or programme as a whole when we have not found a document that states them explicitly. They are marked as such on the candidate’s profile

Documenting all 506 positions with a source is a work in progress. Marking what does not yet have one seems more honest to us than presenting everything with the same appearance of precision.

If a candidate is missing a position

Currently all 11 candidates have a position on all 46 questions; those not backed by an explicit document are marked as estimated. If a candidate were missing a position (when they are added or when a new question is added), it would not be made up: that question would be left out of their calculation.

The number of sourced and estimated positions for each candidate is shown on their profile.

The same applies to comparisons between two candidates: only the questions on which both have a position are compared.

5. Calculation algorithm

Your match with each candidate is calculated using an algorithm that takes three factors into account:

• Quadratic scale: Large differences of opinion are penalised more than small ones
• Conviction factor: Strong opinions (strongly agree/strongly disagree) carry more weight than moderate ones
• Reduced weight for neutral answers: “Neutral” answers have less impact on the final result

You decide the importance: in the full test you rate from 1 to 5 how much each question matters to you (if you leave it unchanged, it counts as 3); in the quick test they all count as 3. The confidence weight is currently the same (3) for all positions. Your age, your sex and the candidate you would vote for, if you choose to give them, play no part in calculating your match.

Conviction factor

Your firmest opinions carry more weight in the result:

• Strongly agree / Strongly disagree → Factor 1.0 (maximum weight)
• Agree / Disagree → Factor 0.75
• Neutral → Factor 0.5 (lowest weight)

Example:

• Your answer: 5 (Strongly agree)
• Position of candidate X: 4 (Agree)
• Confidence weight: 3, Importance: 4
• Difference: |5 - 4| = 1
• Base score: (4 - 1)² = 9
• Conviction factor: 0.5 + (2 × 0.25) = 1.0
• Total weight: 3 × 4 × 1.0 = 12
• Score: 9 × 12 = 108 points
• Maximum possible: 16 × 12 = 192 points
• Match on this question: 108/192 = 56%

Tie-break

The second and sometimes even the third candidate closest to your answers may end up 3 points or less behind the first: one or two different answers would be enough to change the order. Often these are candidates who take almost the same positions on the test questions. When this happens, your result offers you a tie-break with a few more questions.

The questions come from a separate bank, with each candidate’s position documented in the same way as for the test questions, and they are not part of the main test. For each tie, up to 8 are chosen: those that most separate the tied candidates, only from sourced positions (never estimated ones), and only if at least one of them separates the candidates by 2 points or more. Your answers are added to the usual calculation for all candidates, with normal importance, and the result shows where the candidates stood before and where they stand now, with what each one stands for. You can only do the tie-break once. The tie-break does not change your compass or the topic-by-topic comparison, and it does not count in the barometer or the reports, which are still based only on the test questions. When candidates take almost the same positions on almost everything, few questions separate them, and they may remain tied: we would rather say so than invent differences.

Why not a longer test? Adding more questions of the same kind would barely separate those candidates, because two candidates only differ on the questions where they think differently, and each pair needs its own. With the tie-break, each person answers only the questions that separate their candidates.

6. Political compass

The political compass places your position on two axes:

The test categories are grouped as follows to calculate your position:

• Economic axis: The topics “Economy, taxes and debt”, “Pensions”, “Work and cost of living” and “Health”
• Social axis: The topics “Immigration”, “Security and justice”, “Education and youth”, “Energy, climate and farming” and “Society, secularism and regions”

Topics that do not count towards the compass

Three topics, “Europe”, “Foreign affairs and defence” and “Institutions and democracy”, are not assigned to either axis because they do not fit well into the left–right or the progressive–conservative scheme: their questions count towards your match with each candidate, but do not move your position on the compass. That is why we stress that the compass is a simplification.

How we calculate your position

To ensure consistency between your match with the candidates and your position on the compass, we use a system of automatic polarity:

1. Analysis of each question: For each question, we analyse the positions of reference candidates, which differ on each axis: some on the left and some on the right for the economic axis, and some progressive and some conservative for the social axis. The reference candidates are Jean-Luc Mélenchon and Fabien Roussel (left) and David Lisnard and Édouard Philippe (right) for the economic axis, and Marine Tondelier and Raphaël Glucksmann (progressive) and Éric Zemmour and Bruno Retailleau (conservative) for the social axis. The economic axis covers the economy, pensions, work and health; the social axis covers immigration, security, education, energy and climate, and society. Questions on Europe, defence and institutions count towards your match but not towards the compass.
2. Determining polarity: If the left-wing (or progressive) candidates have higher positions on a question, answering “strongly agree” moves you towards the left (or the progressive pole). If it is the right-wing (or conservative) candidates, it moves you towards the right (or the conservative pole). If the two groups answer a question almost the same way (their averages differ by less than half a point on the 1–5 scale), the direction is unclear: that question does not count towards the compass, only towards your match with the candidates.
3. Calculating the score: Your answer is converted to a scale from -100 (left/progressive) to +100 (right/conservative), adjusted for the polarity of each question.
4. Average per axis: Your final position on each axis is the average of all the questions in the corresponding categories.

This system ensures that if you have a high match with a left-wing candidate, the compass will consistently place you on the left, and vice versa.

7. Limitations

We want to be transparent about the limitations of this test:

• It cannot capture all the nuances of politics
• Candidates’ positions may be simplified
• Campaign programmes do not always reflect what candidates do once elected
• Some questions may distinguish between candidates better than others
• What a president can do also depends on having a majority in the National Assembly, which the test does not take into account
• The test does not take into account candidates’ and parties’ track records in keeping their promises

8. What is published

Everything needed to audit the test is available on the site itself:

• The full list of questions grouped by topic, with an explanation of each one.
• Each candidate’s position on each question, with its justification and, where available, the document that supports it.
• The calculation algorithm, detailed on this page and explained with a step-by-step numerical example.
• The aggregated response data, with its methodology note and the minimum threshold applied.

Of the 506 positions, 436 currently link to the document that supports them and 70 are marked as estimated.

Revision log

Significant changes to the database of questions and positions:

• 19 October 2026 (evening): seven signed analyses are published (the candidates explained, the compass, where the candidates agree, pensions, how the presidential election works, how your match is calculated and why the barometer is not an opinion poll). Each candidate page now has a reading of their positions and six frequently asked questions by topic; 17 comparisons between two candidates have a written summary, based only on sourced positions and taken down automatically if one of them changes. Each topic has a short introduction, and your profile by topic in the results shows what each end of the scale means. Candidate pages no longer quote poll figures. The methodology now lists the topics that count on each compass axis, and the election guide now explains that the campaign spending caps set by law are multiplied by a coefficient of 1.23 (€16.851 million and €22.509 million in 2022).
• 19 October 2026: the website is published for the presidential election of 18 April and 2 May 2027, in French and English. The test has 46 statements (20 in the quick test) and 17 tie-break questions; the positions of the 11 candidates are based on their manifestos, their public statements and the votes of their group in Parliament. 562 of the 693 positions link to the document they come from; the other 131 are marked as estimated.

9. Contact and corrections

If you spot an error in the candidates’ positions or have suggestions for improving the methodology, contact us at contact@affinitepolitique.fr.

We particularly value comments that include verifiable sources.