France’s Citizens’ Convention for Climate (2019 to 2021)

France’s Citizens’ Convention for Climate (Convention citoyenne pour le climat) was a panel of 150 residents selected by lot that met from October 2019 into 2020 and produced 149 proposals for cutting greenhouse gas emissions at least 40% by 2030 compared with 1990. President Emmanuel Macron had promised to pass its proposals on “without filter”; he set aside three, the government reworked most of the rest into the Climate and Resilience Law of August 22, 2021, and a promised constitutional referendum never took place. The case is the clearest modern example of what happens when a deliberative panel is given a strong promise and no defined path for keeping it.

France asked ordinary people, chosen like a jury, to spend months learning about climate policy and then write proposals. They did. Most of their ideas went to Parliament in some form. But many were changed along the way, a few were rejected at once, and the members themselves were disappointed with the result. Meanwhile, many people outside the panel did not know much about it or did not trust a group chosen by lot to decide hard questions.

For an overview of other deliberative and direct-democracy cases, see comparative cases. For Ireland’s citizens’ assemblies, which fed into referendums, see the Ireland case.

Defined terms

The Citizens’ Convention for Climate case uses several terms with specific meanings.

• Citizens’ convention (citizens’ assembly, mini-public): a body of residents selected by lot, usually with quotas, that hears evidence, deliberates and makes recommendations. See the glossary.
• Sortition: selection by lot. In practice, modern panels combine random contact with quotas so the final group resembles the population on chosen characteristics.
• Quota sampling: filling fixed shares of a sample by category (for example gender, age, education) rather than accepting whoever answers first.
• CESE (Conseil économique, social et environnemental): France’s Economic, Social and Environmental Council, a constitutional advisory body. It hosted and organized the Convention.⁠²
• “Without filter” (sans filtre): the President’s commitment that the Convention’s proposals would go to referendum, to regulation, or to Parliament without being altered.⁠¹
• Unrelated amendment (cavalier législatif): an amendment added to a bill with no link to its original subject. France’s Constitutional Council strikes these down under Article 45 of the Constitution.⁠⁴
• Haut Conseil pour le Climat (HCC): France’s independent High Council on Climate, which reviews whether laws and policies fit the country’s climate targets.⁠⁵

What happened

The Citizens’ Convention for Climate followed a sequence that can be dated from official and academic sources.

• 2019, origin. A 30-author study in Humanities and Social Sciences Communications states that the President launched the Convention in 2019 “in response to the Gilets Jaunes crisis,” the protest movement that began over fuel taxation.⁠¹
• October 4 to 6, 2019, installation. The Convention’s official site states it was installed on those dates and would meet seven times at the CESE.⁠²
• Mandate. The Prime Minister’s engagement letter asked the Convention to define measures for a reduction of greenhouse gas emissions “of at least 40% by 2030 compared to 1990,” in a spirit of social justice.⁠^1,2
• 2020, proposals. After about nine months of work disrupted by strikes and the COVID-19 lockdown, the Convention adopted 149 proposals.⁠¹ The study reports a final budget of 6.7 million euros.⁠¹
• One week later, the President’s response. According to the same study, the President said he would support 146 of the 149 proposals and used three “veto cards” to set aside a change to the Preamble of the Constitution, a 4% tax on corporate dividends, and a lower motorway speed limit.⁠¹
• February 2021, the bill and a review session. The government drafted a bill in February 2021, and the Convention met for an eighth session in February 2021 to assess the government’s response.⁠¹ On February 23, 2021, the HCC published an opinion on the bill.⁠⁵
• May to July 2021, the referendum stalls. A constitutional amendment to add environmental protection to Article 1, which the President had linked to a referendum, required both chambers to pass identical text. On May 10, 2021, the Senate majority preferred different wording.⁠⁷ On July 5, 2021, the Ministry of Justice stated that the Senate had given the referendum proposal a “fin de non-recevoir définitive,” a final rejection.⁠⁶
• August 13 and 22, 2021, the law. The Constitutional Council reviewed the “Loi portant lutte contre le dérèglement climatique et renforcement de la résilience face à ses effets” in Decision 2021-825 DC of August 13, 2021, and struck down 14 articles and parts of two others.⁠⁴ The law, n° 2021-1104 of August 22, 2021, was published in the Journal officiel on August 24, 2021.⁠¹⁰
• 2022 to 2023, a second convention. The CESE later organized a Citizens’ Convention on the End of Life, whose 184 members met in nine three-day sessions between December 2022 and April 2023 and presented their report to the President on April 3, 2023.⁠⁸

How it worked

The Citizens’ Convention for Climate combined random contact, quotas, expert input and staged voting.

Recruitment. The Convention’s site states that the polling firm Harris Interactive called “more than 255,000 telephone numbers,” 85% mobile and 15% landline, to find 150 volunteers.⁠³ A governance committee set six quota criteria: gender (52% women, 48% men), six age groups from age 16, six education levels with attention to people without a degree, socio-professional category including people in extreme poverty, type of residential area including priority urban districts, and region including overseas territories.⁠³ Attitudes toward climate change were not a selection criterion.⁠¹ Because some members left and were replaced, the Giraudet study reports 159 official participants after 178 people were summoned.⁠¹

Support structures. A governance committee steered the work and three guarantors were charged with the neutrality and sincerity of debates.⁠² Experts and a support team helped draft measures; the academic observers describe the steering bodies’ influence as substantial.⁠¹

Voting. Proposals were adopted by the members. The observers criticized two features: voting rules were shared only hours before votes, and block voting on groups of measures made it harder to reject individual measures.⁠¹ Members also voted on whether to send proposals to a referendum and, by the study’s account, rejected referendums for technical proposals by 60% to 80%.⁠¹

Transmission. The “without filter” commitment named three channels: referendum, regulation, or Parliament.⁠¹ It did not specify what would happen if Parliament amended a proposal, if a measure conflicted with the Constitution, or if the two chambers disagreed. Each of those later occurred.

Evidence: measured results

The Citizens’ Convention for Climate produced measurable outputs, some of which can be stated precisely and some of which remain disputed.

Output. 149 proposals, 146 accepted in principle by the President, 3 set aside at once.⁠¹

Deliberative quality. The Giraudet study, written by 30 social scientists who observed the process, concludes that expert and steering-body input was substantial but did not impair citizens’ “agency, creativity, and freedom of choice,” and that the process built strong consensus and trust among members.⁠¹ It reports that members could and did remove a carbon tax from the agenda.⁠¹

Public opinion outside the room. The same study reports survey results showing over 65% public support for each measure proposed for referendum, but only 28% to 32% of respondents trusted randomly chosen citizens to deliberate productively on complex issues.⁠¹ The authors describe “mutual skepticism” between members and the wider public.⁠¹

Members’ own assessment. In the February 2021 session, members rated the government’s response low, roughly 2 to 3 out of 10 on the study’s report.⁠¹

Climate effect of the law. On February 24, 2021, HCC president Corinne Le Quéré told deputies that the roughly 40% target for 2030 was “pas garantis par l’appareil législatif,” not guaranteed by the legislation, and that building measures would have a “très marginal” effect relative to the national low-carbon strategy; she also said the bill should strengthen the coherence of public action.⁠⁹ The Giraudet study adds that the combined effect of the Convention’s own proposals on the 40% target was never comprehensively assessed.⁠¹

Legal review. The Constitutional Council struck 14 full articles because they were added by amendment with no link, “même indirect,” to the bill as tabled, in breach of Article 45, and censured two passages on delegated-legislation grounds.⁠⁴ It rejected a challenge to the law as a whole under the Environmental Charter, stating it “ne dispose pas d’un pouvoir général d’injonction à l’égard du législateur,” it has no general power to order Parliament to act.⁠⁴

How many proposals were adopted “as written” is disputed. Counts circulated by advocacy groups and the press use different methods and dates, and none was verified for this page; that question is listed below as unresolved.

Failures and criticisms

The Citizens’ Convention for Climate drew criticism from several directions, and the criticisms do not all point the same way.

From members and climate advocates: the promise was broken. The academic observers conclude that the “without filter” commitment was not kept, citing the three vetoes and the government’s reworking of proposals.⁠¹ The HCC’s president said the bill did not guarantee the 2030 target.⁠⁹ The members’ own low rating in February 2021 is the most direct evidence of their disappointment.⁠¹

From Parliament’s defenders: elected bodies must keep the last word. The Senate majority argued that the proposed constitutional wording, “garantit la préservation de l’environnement,” would create, in the words of a Conseil d’État opinion they relied on, a “quasi-obligation de résultat,” a near obligation to achieve results, and would invite litigation; they preferred “préserve,” which a senator described as less legally binding.⁠⁷ On this view, amending or declining a panel’s text is not a betrayal but the ordinary work of a legislature that answers to voters. The Ministry of Justice, for its part, said the Senate’s alternative “n’aurait eu aucun effet juridique,” would have had no legal effect.⁠⁶ Both positions are on the record; neither is adopted here.

From process critics: steering and framing. The observers report that steering bodies were “not strictly neutral,” that some experts went beyond their role, that members got little training in deliberative methods, and that block voting limited choice.⁠¹ A critic skeptical of the outcome can read this as evidence that the panel’s conclusions reflected its organizers; a supporter can point to the same study’s finding that members exercised real agency, including dropping the carbon tax.⁠¹

From the public: legitimacy deficit. With only about three in ten respondents trusting randomly chosen citizens on complex issues,⁠¹ the Convention lacked a broad public mandate even where its individual measures polled well. This cuts against both the claim that the panel spoke for France and the claim that politicians could safely ignore it.

From constitutional design: no path for disagreement. The referendum required identical text from both chambers.⁠⁷ A promise made by the executive could not bind the Senate, and nothing in the Convention’s design said what would happen if it disagreed.

Competing interpretations

Observers read the Citizens’ Convention for Climate in at least three ways.

A deliberative success undone by politics. On this reading, the members did what was asked, produced coherent proposals, and were let down by an executive that overpromised and a Parliament that diluted the result. The observers’ finding on member agency and the HCC’s comments on the bill support parts of this view.⁠^1,9

A legitimacy experiment that confused advice with authority. On this reading, the error was the promise itself. Randomly selected panels can inform elected bodies, but a pledge to pass proposals “without filter” implied an authority the panel did not have and the public had not granted. The low public trust figures and the Senate’s insistence on its own role support this view.⁠^1,7

A useful institution in progress. On this reading, the Convention’s lasting effect is institutional: France ran a second convention, on end-of-life care, in which 76% of voting members supported opening access to “aide active à mourir” while about a quarter opposed, and both positions were reported in the members’ own presentation.⁠⁸ Reporting minority positions alongside majority ones is a design feature many deliberation scholars recommend.

What transferred and what did not

The Citizens’ Convention for Climate left several lessons that later processes adopted and some that did not travel.

Transferred. The CESE reused the format for the End of Life convention, again with members selected by lot, nine sessions, more than 60 expert hearings, and a report handed to the President.⁠⁸ That second convention’s presentation explicitly set out the arguments of the roughly one quarter who opposed the majority view,⁠⁸ which partly answers the criticism that block voting hid dissent.

Did not transfer cleanly. The “without filter” formula was not repeated in the same form in the End of Life process; the members reported that the Prime Minister told them “soyez libres,” be free, which is an invitation to deliberate, not a promise of enactment.⁠⁸ Whether that change was deliberate is not established by the sources consulted here.

Not solved anywhere in this case. No mechanism was created to tell the public, proposal by proposal, what happened and why. The disagreement over counts of adopted proposals is a symptom of that gap.

Application to this framework

The Citizens’ Convention for Climate bears on several parts of the My Vote Forward framework.

• Authority must be stated before participation, not after. The framework asks every proposal to name the body with power to adopt it and what that body is obliged to do; see proposals. The Convention’s promise named three channels but not the obligations of each.
• Advisory results need a reply, not a pledge. A design that publishes a structured response to each proposal (adopted, amended with reasons, rejected with reasons) gives members and the public something checkable. See representation and published releases.
• Deliberating members and the wider public are different populations. The framework distinguishes participants from constituents; see measurement. A panel can be well selected and still lack public recognition.
• Report dissent. The End of Life convention’s reporting of a 76% majority alongside a quarter opposed⁠⁸ is the kind of distribution the framework asks for in collective priorities.
• Framing effects are real. The observers’ finding that steering bodies shaped the agenda⁠¹ is consistent with the question-design concerns in question design and the agenda-control rules in governance.

Unresolved questions

• How many of the 149 proposals were enacted substantially as written? Published counts differ by method and date, and none was independently verified for this page.
• Would a narrower, written commitment (for example, “every proposal receives a reasoned public response within six months”) have preserved trust better than “without filter”?
• What did members’ views look like after the law took effect, rather than in February 2021?
• Can a panel whose individual proposals poll well overcome public skepticism about selection by lot itself, and if so, how?
• How much did the governance committee and experts shape the 149 proposals, and how could that influence be measured rather than described?