Iceland has been proving for ten years that a four-day week works without production collapse

Nine out of ten Icelandic employees now work a reduced week. Six years after the system’s generalization, the country’s GDP per capita has not declined. This result, achieved in a country of 370,000 inhabitants that has conducted the world’s most rigorously documented experiment on working time reduction, changes the nature of the debate: we no longer discuss principle, we discuss method.

The four-day week is no longer a campaign promise from a politician. In Iceland, it is a stabilized social norm, built on ten years of experimentation, evaluation, and institutional adjustment. The track record exists. The data exists. The question now is what to do with it.

The essentials

  • According to data published by Alda and Autonomy, 86% of Icelandic workers had, as of July 2021, access to the right to request reduced hours, and 59% had actually been offered reduced hours during 2021-2022.
  • Iceland’s GDP per capita continued to progress over the 2019-2024 period, with no negative correlation with working time reduction.
  • The pilot trial conducted between 2015 and 2019 covered 2,500 workers across very different sectors: Reykjavik municipal administrations, central state services, hospitals, childcare centers, offices.
  • The success of the model rests on prior investment in work reorganization, not merely the decision to reduce hours.
  • The forward-looking issue: if productivity gains linked to AI free up time on a large scale in the coming years, Iceland offers the only solid empirical framework for deciding how to redistribute this time.

Four years of rigorous experiments before generalization

The story begins in 2015 at Reykjavik City Hall. Not with a law, not with a national sectoral agreement: with a pilot experiment designed as a scientific study. Volunteer teams test the transition from 40 to 35 or 36 weekly hours, at constant salary. Researchers measure productivity, well-being, absenteeism, quality of service delivery. They document difficulties, adjustments, resistances.

The trial expands in 2017. It then covers 2,500 employees, or roughly 1% of Iceland’s working population. Hospitals, childcare centers, administrative offices, social services enter the scheme. The scope is deliberately heterogeneous: the objective is not to validate the concept under ideal conditions, but to understand how it behaves in constraining environments, particularly in healthcare where time presence is directly linked to service quality.

The results published in 2021 by Alda and Autonomy are clear: maintenance of service quality, significant improvement in worker well-being, reduction in stress and burnout, without measurable degradation of productivity. At several sites, productivity advances, because the reduction in working time forced teams to identify and eliminate unnecessary meetings, redundant tasks, inherited processes that persisted out of habit.

This is the central mechanism. The four-day week did not work because employees worked “faster.” It worked because work organization was rethought. Managers at each site received time, resources, and an explicit mandate to map processes, identify waste, test alternatives. Investment in this reorganization precedes the reduction in hours, not the reverse.

How generalization was built

After the publication of results, Icelandic unions negotiated. The two main federations, VR (commerce and services) and BSRB (civil service), integrated the right to a reduced week into collective agreements. Not as a discretionary option left to employer discretion: as an enforceable right inscribed in the collective contract.

The Icelandic government, for its part, extended the scheme to state civil servants. According to the report Going Public published by Alda and Autonomy in July 2021, 86% of the working population then had access to the right to request reduced hours within a union or collective agreement — this figure designating workers who had gained this right to negotiate, not necessarily those who had actually adopted it. The follow-up report published in October 2024 (2021-2022 data) establishes that 59% of workers had actually been offered reduced hours during this period.

This transition from experimentation to norm reveals something important about labor policy. Generalization was not decreed from above. It followed the classical path of durable social reforms in Nordic countries: documented experimentation, agreement of social partners, integration into collective agreements, gradual extension. This sequencing explains why the model held without creating massive employer resistance, including in the public sector where continuous service constraints are strongest.

There are sectors where the transition to four days remains more complex. Hotels and restaurants, certain high-continuity healthcare services, industrial activities with continuously operating production lines: in these cases, the adopted model is not necessarily a strict four-day week, but a reduction of weekly hours below the 40-hour threshold, distributed differently depending on teams. The result is a palette of regimes, not a single imposed format. This is what the notion of “equivalent scheme” summarizes in Icelandic statistics.

A GDP that did not collapse — and why this is the right indicator to watch

The most frequently raised counterargument to the four-day week is simple: if people work less, production drops. Iceland now offers ten years of data to test this proposition. The answer is negative.

Iceland’s GDP per capita was approximately 66,000 to 67,000 dollars in 2019, the year the pilot trial ended. In 2023, it exceeds 84,000 dollars according to International Monetary Fund data. Iceland went through the same covid disruption as the rest of Europe, and emerged from it faster than the average of OECD countries.

This figure does not prove that the four-day week caused growth. Iceland’s economy benefits from a particular structure (tourism, fishing, geothermal energy, finance) and a small size that facilitates adaptation. But it refutes the hypothesis that reduced working hours mechanically trigger production contraction. The conditions for this refutation are known: prior reorganization, maintenance of wages, negotiated framework. These are conditions, not miracles.

Absenteeism, meanwhile, has declined. Data from Icelandic social insurance show a reduction in sick leave in sectors covered by new collective agreements. The effect is not negligible: in France, according to the National Health Insurance Fund, the cost of work stoppages exceeded 15 billion euros in 2023. If part of this absenteeism is linked to burnout and chronic stress, reduction of working time is a public health policy as much as a social policy.

What experiments currently underway in the rest of the world teach differently

Iceland is not alone. Similar experiments have been conducted in the United Kingdom (61 companies, 2022), Germany (45 companies, 2023-2024), Japan (several large groups since 2021), Spain (public pilot since 2023). Their results converge on the principle: companies testing the four-day week generally maintain their production level and reduce turnover.

But these experiments differ from Iceland on one crucial point: they remain voluntary, centered on companies that volunteer, hence self-selected. Organizations entering a reduced-week pilot are, by definition, those thinking they have the means and organization for it. Selection bias is structural.

Iceland solved this problem because its experimentation was designed from the start to cover constraining sectors, particularly care and social services. It is precisely in these sectors that employers were most skeptical, and it is there that adaptations have been richest in lessons. A home care service transitioning to four days without recruiting additional staff must map its routes, rationalize its travel, delegate certain tasks. This organizational work benefits it beyond the sole question of working time.

Comparison with foreign experiments also invites caution regarding direct transposition of the Icelandic model. The institutional density of Nordic countries, the strength of union bargaining power, high social trust between employers and employees, the small country size facilitating consensus: these conditions are not reproducible identically in France or Germany. What the Icelandic model transmits is less a format than a method. Experiment seriously, measure rigorously, negotiate collectively, generalize progressively.

AI is arriving: the Icelandic model becomes decision infrastructure

The forward-looking question shifts dimensions when read in light of productivity gains linked to artificial intelligence. In the acceleration of language model capabilities, economists agree on one point: certain categories of repetitive cognitive tasks will see their processing time drastically reduced in coming years. The question of where the freed time will go is open.

There are two possible answers. The first: time freed by AI allows producing more with the same workforce, translating into productivity gains captured as profit. The second: this time is redistributed as free time, translating into reduced working duration at constant pay. Acemoglu and Johnson, in Power and Progress, emphasize that this choice is not neutral: the gains of technical progress are redistributed according to institutional power relations in place when they materialize, not automatically according to social needs.

Iceland offers, in this context, something no one else possesses: an experimental framework tested over ten years for redistributing productivity gains as free time, with documented results on GDP, health, absenteeism, and job satisfaction. This is not a turnkey exportable model. It is a large-scale proof of concept that allows other countries to move from an ideological debate to a public policy debate.

The horizon is short. Companies deploying generative AI tools on a large scale are beginning to measure significant time gains on functions like writing, synthesis, data processing, customer support. Goldman Sachs estimated in 2023 that 300 million jobs in advanced economies could be partially automated. Even if this projection remains speculative in its precise contours, the intermediate time gains are already visible in data from companies that have documented their deployments.

The question of distributing these gains between capital and labor will arise in the next five to ten years with an intensity that social negotiations of the 2000s did not know. Countries that will have built, by then, an institutional framework to negotiate this redistribution will be better equipped than those approaching the debate without decision infrastructure. Iceland has this infrastructure. It took ten years to build.

What remains difficult to generalize

The Icelandic model does not answer all questions. Three blind spots deserve naming.

The first concerns self-employed workers and freelancers. The four-day week as negotiated in Iceland rests on collective employment contracts. The 15 to 20% of workers who, in European economies, work outside salaried employment benefit from no equivalent mechanism. Working time reduction risks widening a fracture between protected employees and precarious workers if not accompanied by broader reflection on the status of new forms of work.

The second concerns sectoral inequality. In an advanced service economy, it is easier to reorganize the work of a financial analyst or communications officer than that of a care assistant or bus driver. Iceland made a notable effort to include care in its experimentation, but results in these sectors are more heterogeneous and required additional hiring in several cases. The promise of a reduced week without additional cost does not hold equally.

The third touches international competitiveness. Iceland is a small open economy whose main sectors (tourism, fishing, geothermal energy) are not exposed to direct labor cost competition with low-wage countries. For industrial economies competing with Asian producers on volume markets, the question of unit labor cost cannot be dismissed as easily. This is not an argument against working time reduction, it is a constraint to integrate into device design.

The next step belongs to negotiators

Iceland did the hard part: demonstrate that it is possible on a large scale, in constraining sectors, over a sufficient duration to speak of results rather than promises. The next step is not Icelandic. It will be played out in countries beginning to seriously address the question of sharing productivity gains linked to AI.

The European Commission launched in 2024 a consultation on the future of work in the context of automation. Several Member States, including Spain, Belgium, and Portugal, have introduced or tested reduced-week schemes at different scales. The United Kingdom published in February 2023 the results of its largest national pilot, covering 61 companies and 2,900 employees: 92% of participating companies maintained the scheme immediately following the pilot, according to Autonomy and 4 Day Week Global data. A follow-up report published a year later, in February 2024, established that 89% of them had maintained it by that point.

These figures accumulate. They do not yet build a coherent movement, because institutional conditions differ too much from one country to another. But they converge on one lesson that Iceland stabilized first: the four-day week holds when treated as a serious work organization policy, not as an electoral promise or a startup HR benefit.

The question that remains open is timing. If productivity gains linked to AI accelerate in service sectors over the coming years, businesses and states that have not built a negotiation framework beforehand will find themselves having to decide in urgency. Iceland had ten years. Other countries will probably have less.

Sources

  1. 4jours.work — Tracking experiments by country, Iceland
  2. Alda and Autonomy, Going Public: Iceland’s Journey to a Shorter Working Week, 2021 report (available on autonomy.work)
  3. International Monetary Fund, GDP per capita data Iceland 2019-2023 — imf.org
  4. Goldman Sachs, The Potentially Large Effects of Artificial Intelligence on Economic Growth, March 2023
  5. 4 Day Week Global and Autonomy, UK pilot results 2022-2024 — 4dayweek.com
  6. National Health Insurance Fund, work stoppage data France 2023 — ameli.fr
  7. Alda/Autonomy – Going Public (primary report, July 2021)
  8. Alda/Autonomy – On Firmer Ground (follow-up report, October 2024)
  9. Autonomy – UK Four-Day Week Pilot Results (February 2023)
  10. Autonomy – Making It Stick (one-year follow-up, February 2024)
  11. BSRB – A shorter work week (Icelandic primary union source)
  12. Macrotrends – Iceland GDP Per Capita (World Bank data)