Switzerland has built a safeguard that does not automatically close the door when pressure rises. It first asks whether the pressure is real, whether it is attributable to free movement, and whether an intervention is justified. Then, in August, a Senate committee added an instrument its own government had already examined and declined to endorse economically: a levy on immigration.
To understand why that sequence matters, start with the machine.
The machine
The new package of agreements between Switzerland and the European Union — Bilateral Agreements III, signed in Brussels this March — revises the rules on the free movement of persons, expanding EU citizens’ access to Switzerland. Alongside that expansion, Switzerland secured something it had sought for years: a concretized safeguard clause. If the application of the free-movement agreement causes serious economic or social difficulties, Switzerland may temporarily restrict free movement.
The interesting part is how the clause decides that difficulties exist.
The domestic implementing law defines four hard indicators: net immigration from the EU, growth in the number of cross-border commuters, the rise in unemployment, and the rise in the social-assistance rate. For each, the government will set a numerical threshold by ordinance. The values presented publicly are precise to the decimal: an increase in net EU immigration — “extraordinary,” in the government’s own characterization — equal to 0.74 percent of the resident population; a commuter rate rising 0.34 percent in a year; unemployment up 30 percent year over year; social assistance up 12 percent.
If any one of these values is crossed nationwide, the Federal Council must examine whether to activate the safeguard clause.
Examine. The verb is doing careful work. Crossing a threshold obligates the government to deliberate. Activation remains a separate, discretionary decision. And even declining is procedurally encumbered: if the government intends to let a crossed threshold pass without a safeguard request, it must first consult the competent parliamentary committees, the cantons, and the social partners. The audience that must be brought into the decision is institutional — the classic channels of Swiss concordance. Around the hard thresholds sits a softer layer of indicators covering immigration, the labor market, social security, housing, and transport: a falling stock of vacant apartments or rising congestion hours can also put activation on the agenda, and a canton facing regional strain can request it. The Senate committee now wants the government to run the examination annually, whether or not a wire has tripped.
Beyond the border, the machine continues. Switzerland must bring its case to the joint Switzerland–EU committee. If the two sides cannot agree, an arbitral tribunal can be seized, and its mandate is specific: it must determine both that serious difficulties exist and that they were caused by the application of the free-movement agreement. Switzerland can act on its own even against an adverse ruling, but the EU may then respond with proportionate rebalancing measures — reaching, in the worst case, into other internal-market agreements.
Measure, examine, consult, prove causation, negotiate, intervene proportionately, review. Whatever else one thinks of this design, it is a machine built to demand evidence before restriction.
The dashboard problem
The machine has one structural blind spot, and the government’s own calibration exercise exposed it.
When the State Secretariat for Migration presented the proposed threshold values, it also ran them backward: since free movement began in 2002, the thresholds would have obliged the Federal Council to examine activation eight times. The unemployment wire alone would have tripped in 2002, 2003, and 2009.
Look at those years. The dot-com aftermath. The financial crisis. Unemployment spiked in both for reasons that had nothing to do with European workers. One reconstruction of the exercise also catches 2020, the pandemic year — perhaps the clearest illustration of the attribution problem, since in 2020 immigration was itself collapsing. Yet the treaty standard, the one an arbitral tribunal would eventually apply, requires difficulties caused by free movement.
The domestic dashboard can detect load. The treaty process still has to establish where the load came from. A recession would light up the trigger at precisely the moment when immigration explains the least. The indicators measure strain on the systems that absorb population growth — housing, transport, social assistance, the labor market — and strain has many parents.
One wire deserves separate mention. In twenty-three years of retrospective data, the social-assistance threshold never crossed. It remains one of the four hard alarm wires anyway. Its presence tells us something about what legislators consider important enough to monitor even when the historical record rarely or never activates it; why that particular fear earns permanent representation in the machine is a question the machine itself cannot answer.
The price enters
Into this architecture, the Political Institutions Committee of the Council of States now proposes inserting a new tool: an immigration levy.
The design, in outline. Once the safeguard clause has been activated — the levy is one of the tools the trigger unlocks — employers who hire workers newly arrived from the EU would pay an annual charge, at least 4,000 francs per worker according to the proposal’s author, senator Andrea Caroni. Adults arriving through family reunification would be subject to the levy as well, at a lower rate of at least 2,000 francs; minor children would be exempt. Once the clause is active, the charge would extend to third-country nationals too, closing the obvious substitution route. All proceeds would be redistributed to the population. The committee’s stated purpose: create an incentive to use the domestic labor force.
Caroni has been explicit about the instrument’s appeal. A levy, he argues, is milder than imposing quotas on EU citizens. That single sentence contains an entire theory of restriction. A quota says: this many. A levy says: this much per person. A quota treats migrants as units to be counted and capped; a levy treats migration as a flow to be priced at the margin. One is a wall with a gate; the other is a toll booth. The political attraction is easy to see — a toll preserves openness in principle while taxing it in practice, and it generates revenue to hand back to residents, a visible dividend from a contested phenomenon.
The idea is older than this summer. The economist Reiner Eichenberger proposed an immigration charge in 2014, during the debates over implementing the mass-immigration initiative; the Social Democrats and the think tank Avenir Suisse floated their own models the same year; all were shelved for a simple reason — the free-movement agreement forbids discriminating against EU workers, and a tax on their hiring is discrimination in its plainest fiscal form. Caroni revived the idea with a formal postulate in December 2023. It circulated again as a possible counter-project to the “No to a 10-million Switzerland” initiative, the popular measure that would have capped the country’s population. In June, voters rejected that initiative. The cap lost at the ballot box. The levy — never put to the people, never formally proposed by the government — remained politically available, the last instrument left standing.
The referendum did not create the levy. It cleared the field for it.
The government’s report
Here the story acquires its central tension, because the levy arrives with a paper trail.
Caroni’s 2023 postulate obliged the Federal Council to study the instrument, and on May 6 of this year the government delivered its report. It is a remarkable document to sit behind an adopted committee proposal, because it is, in substance, a rejection.
On the economics, the report is blunt. The case for a levy rests on the premise that the taxed immigration costs the country more than it contributes. For that premise, the government found no empirical evidence. Migrant workers pay taxes and social contributions like residents; they pay their own rent. A levy would amount to a special tax on an economically desired activity, something the report says is generally associated with welfare losses. That is the government’s phrasing of the problem, in a report written by its own economists.
On the international comparison, the report offers one live experiment. Singapore has operated an employer levy on foreign workers for decades, adjusting it repeatedly — most recently to a uniform 650 Singapore dollars per month. The report’s observation: despite regular adjustments, the levy has not produced a decline in the immigration of foreign workers. This proves less than it seems — Singapore is no controlled experiment, and surging demand for labor could mask a genuine braking effect. But it means the government’s own comparative example fails to demonstrate the steering outcome the instrument promises. The committee adopted the levy with that observation sitting in its briefing file.
On the law, the report stacks obstacles. Outside the safeguard clause, any levy on EU workers violates the free-movement agreement’s non-discrimination guarantee. A flat charge on family reunification risks colliding with Article 8 of the European Convention on Human Rights if it burdens weaker households without individual adjustment. Commitments under GATS and a long list of free-trade agreements stand in the way of taxing transferred specialists. And a levy designed to raise money for the treasury would be a tax requiring a new constitutional basis, meaning a mandatory referendum with a double majority.
The strange doorway
And yet the same report, in the same chapters, draws a map of the one narrow passage through all of this.
Swiss public law recognizes a category between fee and tax: the steering levy, the Lenkungsabgabe. Its purpose is behavioral, never fiscal. It is constitutional without amendment only if its proceeds are returned in full to the population and the economy. The report compares this with the CO2 levy, noting that the immigration levy would have to be purer still: the CO2 levy is one-third fiscal, and this one could not be fiscal at all. The redistribution clause in the committee’s proposal is therefore no gesture of political sweetening. It is the legal load-bearing wall. Remove it, and the levy becomes a tax without a constitution.
Then the report goes one step further. Under the updated free-movement agreement, it concludes, it would in principle be conceivable to apply a temporary immigration levy, structured as a steering levy, as a protective measure — provided the safeguard clause has been validly invoked, and provided Parliament writes the instrument into the implementing law during the current legislative deliberations.
Read as a whole, the May report is a rejection containing a doorway: no demonstrated economic benefit, a welfare-loss framing, a human-rights flag, an unflattering comparative example — and, folded inside, the precise legal recipe for building the thing anyway.
August 18
The committee took the recipe and discarded the verdict.
On August 18 the SPK-S voted to anchor the steering levy in the implementing law, exactly where the report said it would have to go: 7 to 0, with 6 abstentions.
The vote records support without consensus — seven yes, no opposition, and six members unwilling to endorse. What the abstentions mean is unknowable from outside; Swiss committee deliberations are confidential, and no abstaining member has spoken. The unresolved issue could be economic effectiveness, treaty compatibility, the family-reunification design, tactical positioning ahead of the plenary, or something else entirely. The autumn session of the Council of States, where positions go on the public record, will tell us which.
The seam
One design detail deserves closer inspection, because it is where the proposal’s two theories pull apart.
The committee’s stated logic is labor-market steering: make the imported worker more expensive, and the employer reaches for the domestic one. Whatever its empirical merits — Singapore counsels modesty — the logic at least has a shape: an employer, a hiring decision, a substitution margin.
Now apply it to a spouse. An adult arriving through family reunification has no employer paying the charge, no position a resident might fill instead, no margin on which any incentive can operate. The steering theory simply has no purchase there. Applied to family members, the levy no longer operates through employer substitution. It begins to look instead like a deterrent or a toll on the exercise of family-reunification rights. The payment architecture concedes the point: for the worker, the employer pays; for the spouse, the household does.
And precisely at this seam, the legal exposure peaks. The Federal Council’s report locates the ECHR risk in flat charges on family reunification that burden weaker households. The place where the economic theory stops working and the place where the human-rights vulnerability concentrates are the same place. Watch what the Senate does with that provision in the autumn. If the family clause is stripped, the levy becomes a cleaner labor-market instrument. If it survives, the proposal can no longer be explained by labor-market steering alone.
The question the machine cannot answer
Step back, and there are two stories nested here.
The outer story is the machine: a safeguard built around repeated evidentiary and procedural gates, made to measure before it examines, examine before it consults, establish causation before it restricts, and restrict only proportionately and for a time. Whatever one thinks of managed migration, this is restriction designed to justify itself continuously — architecture that forces the question “is the pressure real, and is it ours to blame?” at every stage.
The inner story is the toll that a committee now wants to bolt into that machine: an idea circulating in Swiss debate for over a decade, whose economic premise the government’s own report could not substantiate, whose comparative example does not demonstrate the promised effect, and whose most legally fragile clause is also its most economically incoherent one.
The tension between the stories is the point. The safeguard is built to demand evidence before restriction. The proposed restriction arrives carrying, in its own file, the evidence against its economic case.
Legislation, I have argued before, is evidence of how legislators model the world before it is evidence about the world. By that standard, the levy is unusually revealing. It models immigration as a continuous flow with a per-unit cost, to be priced rather than prohibited; it models the resident population as a party owed compensation; and it models the government’s economic objections as a smaller obstacle than the government’s legal permission. In June, voters rejected the cap. In August, a committee proposed a price — gently, conditionally, behind a trigger, with the proceeds returned as a dividend.
Whether the Senate follows, whether the abstainers break, whether Brussels treats a toll as an “appropriate measure” or as discrimination wearing a safeguard’s clothes — all of that is for the autumn.
For now, the sequence stands on its own. The government mapped the doorway while recommending nobody use it. In August, the committee walked through.
WE&P by: EZorrillaMc&Co.
Sources
- Swiss Federal Council — Criteria for applying the safeguard clause, 14 May 2025
- Swiss Federal Council — Report on an immigration levy, 6 May 2026
- Swiss Federal Assembly, Political Institutions Committee of the Council of States — Communiqué on Bilateral Agreements III / immigration law, 18 August 2026
- Watson — Andrea Caroni on an immigration levy instead of quotas, 17 May 2026
- SWI swissinfo / Keystone-SDA — Switzerland and EU sign Bilateral Agreements III, 2 March 2026
- Swiss Federal Assembly — Political Institutions Committee on the “No to a 10-million Switzerland” initiative and counterproposal, 4 November 2025
- SWI swissinfo — Swiss voters reject the “No to a 10-million Switzerland” initiative, 14 June 2026
Travel in Life 