Nine to Nil: To Win the Cantons a Veto on the EU Treaties, a Swiss Committee Would Cancel a Referendum

Switzerland is about to decide how Switzerland decides. The Council of States takes up a committee initiative this autumn that would change the rules for the popular vote on the package of treaties meant to reset the country's relationship with the European Union. It is being reported as a clean federal question: should the small cantons hold a veto over the Switzerland-EU agreements, or should a simple majority of voters settle it? That is the right question. It is not the whole transaction. The committee's own record shows what the cantonal veto would cost, and the cantons are not the ones who would pay.

Start with the text, because the text is why there is an argument. Article 140 of the Federal Constitution lists what must go to the vote of the People and the Cantons, and the list is short: amendments to the Constitution, and accession to organisations for collective security or to supranational communities. Article 142 then says such proposals pass only if a majority of voters and a majority of the cantons approve. The package known as Bilaterale III is engineered to be neither. It is dynamic alignment with EU law across updated and new single-market agreements, with dispute settlement, and no membership. Article 141 catches it instead, as a treaty containing important legislative provisions, and an optional referendum is decided by the popular majority alone. The Federal Council reached that conclusion on 30 April 2025, restated it with its message of 13 March 2026, and restated it on 12 August 2026 while asking the Council of States to leave the matter alone.

The Council of States' State Policy Commission found the route around it. On 6 May 2026, by six votes to six with the president's casting vote, it launched initiative 26.425. On 30 June it adopted the draft by seven votes to six. The device is a transitional provision written into the Constitution itself, instructing the Federal Council to ratify. Put the approval in the Constitution and it becomes a constitutional amendment, and Article 140 does the rest: People and Cantons, double majority, cantonal veto. It is elegant, and the commission is candid that settling a dispute about Article 121a and immigration was part of the appeal. Then comes the decision the reporting has not carried. By nine votes to nil, with four abstentions, the same commission resolved to write into the same constitutional provision that the implementing legislation shall not be subject to referendum. The stated purpose is in its release: the agreements and their implementing legislation should form one package, so that the accompanying measures cannot be brought down in a later popular vote. Read the three decisions in sequence. Six to six on whether to try at all. Seven to six on the draft. Nine to nil, not one vote against, on removing the referendum from the domestic law.

That is a trade, and Article 141a names the machinery. Where ratification goes to a mandatory referendum, Parliament may fold the treaty's constitutional amendments into the ratification decision; where it goes to an optional referendum, it may fold in the statutory amendments. The Federal Council's route leaves the implementing statutes exposed to the ordinary optional referendum, which is the basis for its claim that its own solution is the cleaner one in democratic terms. The commission's route closes that door by writing the closure into the Constitution. So the live question is not whether Switzerland keeps a veto. It is which veto: twenty-three cantonal votes distributed by territory, or fifty thousand signatures against the wage-protection law that made the treaty tolerable in the first place.

It is worth knowing what the cantonal veto actually weighs. There are twenty-three cantonal votes rather than twenty-six, because Article 142 names six cantons that get half a vote each: Obwalden, Nidwalden, Basel-Stadt, Basel-Landschaft, Appenzell Ausserrhoden and Appenzell Innerrhoden. Set those weights against the Federal Statistical Office's 2025 count of Swiss nationals aged eighteen and over, and one vote in Appenzell Innerrhoden carries 39.6 times the weight of one vote in Zurich. Then work out the cheapest way to lose. Fourteen cantons, holding 11.5 votes between them and containing 978,501 Swiss adults, are enough to deny a constitutional amendment its cantonal majority. That is 18.0 per cent of the adult citizenry. Eighteen per cent, correctly distributed, defeats what the other eighty-two want.

The veto is real but not routine, and the record can be computed rather than asserted. Take the Swissvotes dataset of every federal vote since 1848 and keep the legal forms that require both majorities. There have been 486 such votes decided, down to 14 June 2026. The people carried 214. In ten cases the cantons then struck the proposal down. A control confirms the sorting: among votes needing only a popular majority, 127 were carried and none was defeated afterwards, which is what the Constitution requires. The ten cluster late. Two fell on the same day in June 1994, the family-policy article went in 2013, and the most recent was the corporate responsibility initiative of 29 November 2020, carried with 50.73 per cent and killed 14.5 cantons to 8.5. In 1975 the economic-policy article died on a tie, eleven to eleven, because a tie is not a majority.

The case for the commission deserves to be put at its strongest. Dynamic adoption of law written elsewhere, policed by a tribunal, is a larger transfer of authority than an ordinary trade agreement, and if the Constitution's list of double-majority triggers was drafted with accession in mind and does not catch this, that is arguably a gap in the text rather than a reason to accept the lighter vote. Cantons will administer much of the resulting law. And the package logic is not cynical: uncouple the treaty from the wage protection and voters can approve the treaty in one ballot and dismantle its compensations in the next, which is a real way for a negotiated settlement to die. Nor is the geography as clean as the arithmetic suggests. The ten cantons asking for the double majority are the small German-speaking cantons of central and eastern Switzerland, plus Ticino. The fifteen content with a popular majority include the whole of French-speaking Switzerland whatever its size, Jura among them at 51,770 adults. Language and history are in this, not only population.

But the cantons were asked, and the record of the asking is a government document. The Federal Council's consultation results report of 13 March 2026 itemises the positions at section 1.2.5.1. Fifteen cantons backed the optional referendum. Ten asked for a mandatory one. Bern abstained. The fifteen hold fourteen of the twenty-three cantonal votes and 75.3 per cent of the adult citizenry; the ten hold eight votes and 11.3 per cent. The cantons that the cantonal-majority rule most over-weights are the ones demanding it, and the cantons it under-weights are declining a veto they are being offered. The fifteen also gave a reason, recorded in the same document: the package effects no accession to a supranational organisation, the Constitution therefore provides an optional referendum, and the decision of the constitution-maker should be respected.

This series has argued that decentralization is a question about where a decision sits. Switzerland is the case everyone reaches for, and the unusual thing here is that both routes push the decision outward, to two edges that are in competition: the canton, over-weighted on purpose so the small are not governed entirely by the large, and the signature, which belongs to whoever can gather fifty thousand of them. Federalism and direct democracy are usually described as the two halves of the Swiss settlement, as though they pulled the same way. They do not, always. And there is a last detail in Article 141 that should be uncomfortable for both camps. The optional referendum can be demanded by fifty thousand voters or by any eight cantons, so under the Federal Council's route the cantons keep the trigger. What they lose is the weight of their vote once the question is put. A right to call a vote you cannot decide is not nothing, and it is not what is being fought over.