The Four-Day Work Week: What the Trials Actually Measured

The Four-Day Work Week

When the largest four-day week trial ended in December 2022, 56 of the 61 participating UK companies kept the policy. That 92% figure is widely quoted. It is accurate, but it answers a narrower question than most readers assume. It shows that companies which volunteered for a trial organized by an advocacy group mostly liked the result. It does not show that a four-day week raises productivity, which the trial did not measure directly.

Read closely, the main studies measured three different things: how workers feel, whether firms kept the policy, and, for a minority of firms, revenue. Separating those three is the only way to see what the evidence supports and where it runs out.

Most “Four-Day Weeks” Were Neither Four Days Nor 32 Hours

The model promoted by 4 Day Week Global is called 100-80-100: 100% pay for 80% of the time, in exchange for 100% of the output. In practice, the trials let each organization design its own version. The UK pilot required only that pay stay at 100% and that staff get a “meaningful” reduction in working time. Companies chose Fridays off, staggered days, annualized hours or conditional schemes tied to performance.

The result was a smaller cut than the name suggests. Employees in the UK pilot worked an average of 34 hours a week during the trial, down from about 38 before it, according to the report by the think tank Autonomy. In the multi-country study published in Nature Human Behaviour in 2025, average hours fell by about five per week. Iceland’s widely cited trials were not four-day weeks at all: workers moved from 40 to 35 or 36 hours spread over five days.

The size of the cut matters because it sets how much more productive each hour has to become for output to stay flat.

How much hourly productivity must rise to keep output flat
Required rise = (old hours ÷ new hours) − 1
Full 100-80-100 model, 40 → 32 hours: 40 ÷ 32 − 1 = 25%
UK pilot average, 38 → 34 hours: 38 ÷ 34 − 1 ≈ 11.8%
Iceland, 40 → 36 hours: 40 ÷ 36 − 1 ≈ 11.1%
Iceland, 40 → 35 hours: 40 ÷ 35 − 1 ≈ 14.3%

A 12% gain in hourly productivity is far more plausible than a 25% one. Our reading, not a tested result, is that cutting meetings, protecting focus time and dropping low-value tasks could close a gap of that size, while a full 20% cut is a much harder test that most trials did not actually run. Headlines about a “four-day week” often describe the hardest version while the data comes from the easier one.

What the UK Pilot Measured, and How

The UK pilot ran from June to December 2022 with 61 organizations and about 2,900 employees. It was coordinated by 4 Day Week Global with the 4 Day Week Campaign and Autonomy, and studied by researchers at the University of Cambridge and Boston College. About two-thirds of participating firms had 25 or fewer employees.

Headline result How it was measured What to keep in mind
56 of 61 firms continued (92%); 18 made it permanent Company decision after the trial Of the 5 that did not continue, 2 extended the trial and 3 paused
Sick days down about 65% Company records, against the same period a year earlier Before-and-after comparison, no control group
Resignations down 57% Company records, against the same period a year earlier Labor market conditions also changed between the two periods
71% of employees reported less burnout Employee surveys at start and end Self-reported, by workers who knew the hoped-for outcome
Revenue up 1.4% during the trial Firm-reported revenue, weighted by company size Based on 23 of the 61 organizations
Revenue up 35% against a comparable earlier period Firm-reported revenue Based on 24 organizations; the comparison cannot separate the effect of hours from changes in business conditions

Two details in the table change how the headline should be read. First, productivity itself was not a common metric, because the firms collected such different data. Revenue stood in for it, and fewer than half the firms supplied revenue figures. A 35% year-on-year jump is large enough that market conditions could explain much of it, but the design gives no way to tell. Second, the employee survey thinned out over time. At baseline, 1,967 of the 2,548 employees who received it responded, or 77%. About 70% of those completed the endpoint survey, which works out to roughly 1,380 people, or 54% of everyone surveyed. If the least satisfied workers were the most likely to skip the second survey, the improvement would look larger than it was. That is a possibility the design cannot rule out, not a finding.

The policy proved durable among participants. A follow-up by Autonomy in February 2024 found that 89% of the firms still had a reduced-hours policy, and 51% had made the four-day week permanent.

The Strongest Evidence So Far: A Comparison Group

The study that moved the debate furthest was published by Wen Fan, Juliet Schor and colleagues at Boston College and University College Dublin in Nature Human Behaviour in July 2025. It pooled six-month trials at 141 organizations in the US, UK, Canada, Ireland, Australia and New Zealand, covering 2,896 employees. Its key addition was a comparison group: 285 employees at 12 companies that had shown interest but kept a five-day week.

Workers in trial companies reported lower burnout, higher job satisfaction and better mental and physical health; the comparison group showed no comparable change. The effects were largest for burnout and job satisfaction and smallest for physical health. The pattern within the data is the most useful part. Employees whose own hours fell by eight or more per week improved the most. Those whose hours fell by one to four, or five to seven, improved less. Larger cuts at the individual level predicted bigger gains, while the size of the company-wide cut did not. The researchers traced the effect through three channels: better self-rated ability to do the job, fewer sleep problems and less fatigue.

The limits are stated by the authors themselves. Companies chose to take part, so the sample is dominated by small and medium-sized firms in English-speaking countries. The comparison companies were few, and every well-being measure came from surveys. The authors call for randomized trials, possibly run by governments, as the next step. In other words, the strongest study so far is evidence that shorter hours improve well-being in firms that want them. It is not a test of productivity, and it is not evidence about firms that would be forced into the change.

Iceland: The Trial People Cite as a Four-Day Week

From 2015 to 2019, Reykjavík City Council and the Icelandic government ran two trials involving about 2,500 workers, more than 1% of the country’s working population. Hospitals, offices, preschools and social services moved from 40 to 35 or 36 hours a week with no pay cut. The 2021 report by Iceland’s Association for Sustainability and Democracy (Alda) and Autonomy found that productivity and service provision stayed the same or improved in most workplaces, and that workers reported less stress and burnout.

The aftermath is the more significant result. In contracts negotiated between 2019 and 2021, unions won shorter hours for their members, and the report states that 86% of Iceland’s workforce had either moved to shorter hours or gained the right to them. What spread was a reduction of a few hours a week inside a five-day schedule, negotiated sector by sector. Calling it a four-day week, as many headlines did, misstates what was tested.

Microsoft Japan’s 40%: One Month, One Ratio

In August 2019, Microsoft Japan closed its offices on Fridays for its roughly 2,300 employees as part of a project called the Work Life Choice Challenge. Sales per employee rose 39.9% compared with August 2018. The company also reported 23.1% less electricity use and 58.7% fewer printed pages, and 92.1% of employees said they liked the arrangement.

The 39.9% figure became the most quoted statistic in the four-day week debate, and it carries the least weight. It covers a single month. It measures sales divided by headcount, a ratio that can swing with the timing of closed deals rather than with how much each hour of work produced. The comparison is with one month a year earlier, with no control for business conditions, and the experiment also changed meeting rules, capping most meetings at 30 minutes. The result is a real data point from one company in one month. It cannot isolate the effect of a day off, and Microsoft Japan presented it as a work-style experiment rather than a permanent policy.

Germany 2024: Physiological Data and Two Dropouts

Germany’s first large pilot ran for six months from February 2024 with 45 organizations, organized by the Berlin consultancy Intraprenör with 4 Day Week Global and studied by Julia Backmann and Felix Hoch at the University of Münster. It went further than earlier trials in what it measured: besides surveys and interviews, researchers used smartwatch data and stress-hormone (cortisol) measurements from hair samples.

More than 70% of the organizations said they wanted to continue, either by extending the test or adopting the four-day week fully. Two large companies dropped out, citing economic pressures or a lack of internal support. Several participants also chose to reduce hours across five days rather than drop a whole day. Backmann’s own conclusion was measured: the model may suit some companies, but the results do not show that every employer in Germany should adopt it.

Trial Scale and length Actual change in time Comparison group Main outcome data
UK pilot, 2022 61 organizations, about 2,900 workers, 6 months About 38 to 34 hours on average None (before and after) Employee surveys, company records, revenue from 23 firms
Fan, Schor et al., 2025 141 organizations, 2,896 workers, 6 countries, 6 months About 5 hours less per week 285 workers at 12 companies Employee surveys
Iceland, 2015 to 2019 About 2,500 public-sector workers 40 to 35 or 36 hours, still five days Not randomized Service provision, well-being surveys
Microsoft Japan, 2019 About 2,300 employees, 1 month Fridays off in August None (prior year’s August) Sales per employee, usage data, staff survey
Germany, 2024 45 organizations, 6 months Varied; some cut hours over five days Not randomized Surveys, smartwatch data, hair cortisol

Four Limits That Run Through Every Trial

1. The companies chose themselves

Every major trial recruited volunteers, usually through an organization that campaigns for the four-day week. As Michael Sanders, a professor of public policy at King’s College London, has pointed out about the UK trials, firms that sign up are likely to be those where staff will embrace the change. That inflates the continuation rate and limits how far the results travel. A firm that joins because its work is already flexible tells you little about a hospital ward or a factory line.

2. Well-being is measured by the people who benefit

Burnout, stress, satisfaction and even “productivity” in several trials were self-rated by employees who knew what the trial hoped to show and who stood to keep a day off if it succeeded. That does not make the answers false, but it makes them hard to separate from enthusiasm. The German pilot’s physiological measures are a step toward outcomes that do not depend on what participants say.

3. Six months is short

Most trials lasted six months, and Microsoft Japan’s lasted one. Workers may push hardest at the start of a visible experiment, a pattern often called a novelty effect. Whether gains persist once the arrangement becomes routine is not something a six-month trial can show; that is an open question, not a measured result. The Autonomy follow-up shows that most UK firms kept the policy for over a year, which speaks to durability of the policy, not to durability of the productivity effect.

4. Output is the weakest measurement

The trials measured well-being carefully and output loosely. Revenue came from a minority of firms, Microsoft’s figure divided sales by headcount, and the largest peer-reviewed study did not measure output at all. The claim that work gets done in fewer hours rests mostly on the fact that firms did not abandon the policy, which is a reasonable signal but not a measurement.

The research also connects to a longer history: the five-day, 40-hour standard itself began as an employer experiment, as described in why the 40-hour week exists, and actual working time already varies widely between countries, as shown in working hours by country. For employers doing the arithmetic, the annual difference is easy to underestimate: moving from 40 to 32 hours removes 416 hours a year at 52 weeks, which is why the baseline for a full-time year deserves a careful look, as covered in how many working hours are in a year.

What a Convincing Productivity Test Would Need

The design that would settle the productivity question is clear, and Fan, Schor and colleagues point toward it. Firms or teams would be randomly assigned to a shorter week or to their usual schedule, rather than volunteering. Output would be measured per hour with objective data the firm already tracks, such as cases closed, units produced or billable work delivered, rather than inferred from revenue or self-ratings. The trial would run long enough, ideally a year or more, to separate a lasting effect from a novelty spike. And it would report results for every firm that started, including those that quit, so that the dropouts count as evidence rather than disappearing from the totals.

 


Sources and references

Fan, Schor, Kelly et al. (2025), Nature Human Behaviour
Work time reduction via a 4-day workweek finds improvements in workers’ well-being: 141 organizations, 2,896 employees and a comparison group
University of Cambridge: Results of the UK Four-Day Week Pilot
Findings from the 61-company UK trial on sick days, resignations, well-being and how employees used their time
University of Münster: Study Results on the German Four-Day Week Pilot
Results from 45 organizations, including dropouts and physiological measurements
CNBC: Microsoft Japan’s Four-Day Work Week Experiment
The August 2019 experiment behind the widely quoted 39.9% rise in sales per employee
↑
✕Close Menu