Four-Day Week Trials — What the 92% Success Rate Hides

The short answer to what four-day-week trials found is this: workers who keep their pay while cutting roughly five hours a week report markedly lower burnout and better health, and most of the employers who volunteer to try it choose to keep it ✓ Established [3] [1]. What the trials have not shown, despite the way they are often reported, is that productivity rises across the economy when hours fall. The largest and most cited study, the 2022 British pilot run by the advocacy group 4 Day Week Global with the think tank Autonomy and researchers at the University of Cambridge and Boston College, enrolled 61 companies and around 2,900 workers between June and December 2022 [1]. At the end, 56 of the 61 companies (92%) said they would continue and 18 made the change permanent [4]. Staff reported a 71% reduction in burnout and 39% said they were less stressed; resignations fell from 2.0 to 0.8 per 100 employees a month and sick days from 2.0 to 0.7 per employee ✓ Established [1] [5]. Those numbers travelled around the world within days. The caveats printed in the same report travelled far less.

The fine print matters because it changes what the evidence can support. The UK report’s revenue figure, a 1.4% rise over the trial period, was calculated “across those 23 organisations who gave the data” — 23 of 61 ✓ Established [1]. A separate year-on-year comparison showing revenue up 35% drew on 24 firms. On staff turnover, sick days and hiring, the authors were explicit: “we are unable to say that these three trends are statistically significant,” partly because of small numbers and partly because of the volatile 2022 labour market [1]. Christopher Deubert, an employment lawyer who worked through the report line by line, noted that between 44 and 51 of the 61 organisations provided data depending on the measure, that 58% of employees answered the final survey, and that the results had not been peer reviewed when published [6]. Anthony Veal, a work-time scholar at the University of Technology Sydney, added that revenue was measured rather than output, in a year when prices were rising fast, so a 1.4% nominal increase could conceal a real-terms fall [7]. None of this makes the pilot worthless. It makes it a pilot.

The study that moved the debate onto firmer ground arrived in July 2025. Writing in Nature Human Behaviour, the Boston College sociologists Wen Fan and Juliet Schor, with Orla Kelly and Guolin Gu, analysed 2,896 employees at 141 organisations across six countries — the United States, Canada, the United Kingdom, Ireland, Australia and New Zealand — and compared them with 285 employees at 12 companies that did not change their hours ✓ Established [3] [8]. Over six months, burnout on a five-point scale fell from 2.83 to 2.38 in the trial firms while it drifted from 2.90 to 2.94 in the comparison group; job satisfaction rose from 7.07 to 7.59 against a decline from 6.66 to 6.47 [2]. The authors also found improvements in mental and physical health, and their central conclusion was that “income-preserving 4-day workweeks are an effective organizational intervention for enhancing workers’ well-being” [3]. This is the strongest wellbeing evidence yet published, and it comes with its own warning label: the same authors write that their findings “likely overestimate the true effect” ◈ Strong Evidence [2].

The Nature Human Behaviour paper is candid about why its estimates are probably too high. Its participating firms chose to take part, and “because those companies may be more supportive of flexibility and wellbeing initiatives, the findings likely overestimate the true effect” [2]. It was not randomised, because “random assignment was not feasible for an organization-wide change at mostly small firms.” Most importantly for anyone reading it as a productivity study, “every outcome here, including hours, is self-reported: no biomarker, productivity, revenue, or service-level measure was collected” ✓ Established [2]. The comparison group was small and drawn only from the United States, its survey response rate at baseline was 58% against 80% in the trial firms, and the sample was, in the authors’ words, “disproportionately high-income and Anglophone with many small organizations” [2] [8]. Scientific American, reporting the paper, put the remaining doubt plainly: employees who wanted to keep their extra day off might have exaggerated the benefits [9]. The paper establishes that something real happens to wellbeing. It does not establish how large the effect would be in a firm that did not volunteer.

A second problem is definitional. “Four-day week” is used for at least three different arrangements, and the evidence for one does not transfer to the others. The trials run by 4 Day Week Global use what organisers call the 100-80-100 model; national schemes in Belgium, Tokyo and Dubai are something else entirely [10] [11] [12]. Iceland’s celebrated experiment was neither [13]. The distinctions that matter are these:

• Reduced hours at full pay (100-80-100): staff keep 100% of pay for roughly 80% of the time in exchange for maintaining output. This is the model tested in the UK, North American, German and Portuguese pilots, although actual hours in the UK fell from 38 to 34, not to 32 [1] [14] [15]
• Compressed hours: the same 38 or 40 hours squeezed into four longer days. This is what Belgium’s 2022 law offers, what Utah tried in 2008 and what Dubai’s summer scheme uses [16] [17] [12]
• Modest weekly reductions: Iceland’s 2015–2019 trials cut a 40-hour week to 35 or 36 hours, usually without a day off [18] [13]
• Short holiday experiments: Valencia’s 2023 “four-day week” was four consecutive Mondays off created by moving public holidays, covering 360,000 workers for one month [19]

The distinction is not pedantry. When Gallup surveyed 12,313 full-time American employees in 2022, the 8% already working four days reported higher burnout than those working five, and no significant gain in thriving ✓ Established [20]. That finding is regularly cited as evidence against the four-day week, and the Nature Human Behaviour paper is cited as evidence for it, but they are mostly measuring different things: Gallup’s four-day group mixes compressed ten-hour days with genuine reductions, while the 4 Day Week Global trials cut average hours by about five a week [20] [8]. Gallup’s chief workplace scientist Jim Harter drew the cautious conclusion that the arrangement “might work for some people and some jobs, but could cause issues in other situations” [21]. The clearest single lesson from the full body of evidence is that working fewer hours appears to help wellbeing, while working the same hours in fewer days may not ◈ Strong Evidence [20] [3].

This report audits what the trials found in the order a sceptical reader would want: the mechanism that might explain the results, the strength of the evidence, the human outcomes, the international record, the policy response and the genuine points of disagreement. The argument it builds is narrower than either camp prefers. The four-day week, in its reduced-hours form, has a solid and repeated record of improving how workers feel and a credible record of not damaging the small, mostly office-based firms that choose it [3] [14] [15]. It has no rigorous record yet of raising productivity, no randomised trial, and almost no evidence from the hospitals, care homes, factories and shops where most hours are worked [22] [23]. Advocates have built a strong case from weak designs; critics have often dismissed real results because the designs were weak. Both habits obscure what is actually known.

The intuition behind the four-day week is that a large share of the working week is not productive work at all, and that time pressure forces firms to remove it. The trial evidence supports the first half of that claim in a specific way: the firms that kept output stable did so by changing processes, not by asking staff to work faster. In Portugal’s 2023 government-backed pilot, 92% of companies that made two or more process changes kept the four-day week, against 62% of firms that made none ◈ Strong Evidence [24]. In Germany, where the University of Münster monitored 45 organisations in 2024, more than half changed their meeting culture and a quarter introduced new digital tools [14]. The organising consultant Carsten Meier summarised the German experience as a discovery that “the potential of reduced working hours seems to be hidden under complicated processes, meetings and a lack of digitalisation” [14]. The 4 Day Week Global programmes build this into their design: firms receive around eight weeks of preparation to redesign workflows before the trial begins [9]. What is being tested, in other words, is not a shorter week alone but a shorter week plus a management overhaul.

The most famous example of that overhaul is also one of the least rigorous. In August 2019 Microsoft Japan gave about 2,300 employees five Fridays off with full pay as part of a “Work-Life Choice Challenge”, capped meetings at 30 minutes and encouraged remote collaboration [25] [26]. The company reported that sales per employee rose 39.9% compared with August 2018, that electricity use fell 23% and that 59% fewer pages were printed; 92% of employees said they liked the scheme [25] [26]. The “40% productivity” headline circled the globe. Yet the measure was sales per employee over a single month, compared with the same month a year earlier, with no control group and no adjustment for anything else that changed in the business. The experiment was not made permanent; Microsoft Japan’s planned follow-up in the winter did not include extra paid leave [25]. It demonstrates that meeting discipline can be imposed quickly. It does not demonstrate a 40% productivity gain ⚖ Contested.

There is older and harder evidence that very long hours are unproductive. The Stanford economist John Pencavel, studying British munitions workers during the First World War, found that “below an hours threshold, output is proportional to hours; above a threshold, output rises at a decreasing rate as hours increase” ✓ Established [27]. His estimates put that threshold at around 49 hours a week, with output at 70 hours barely higher than at 56. This is the scholarly basis for the claim that overwork wastes effort, and it is sound. But it cuts both ways. In Pencavel’s data, below the threshold each hour was as productive as the last, which implies that cutting from 40 hours to 32 would reduce output roughly in proportion unless something else changes [27]. The case for the four-day week therefore cannot rest on fatigue alone. It rests on the claim that modern office work contains enough slack — in meetings, interruptions and administrative drag — to absorb an eight-hour cut. That is plausible for some jobs and implausible for others, and the trial samples are heavily weighted towards the first kind [1] [23].

The dose-response finding is the most important piece of evidence that the wellbeing effects are caused by shorter hours rather than by enthusiasm for an experiment. If gains were purely a morale effect — staff happy to be chosen, or keen to keep a perk — there would be little reason for workers whose hours fell further to improve more. Instead, the Nature Human Behaviour team found a stepped relationship, with workers cutting eight or more hours gaining almost three times as much on the burnout measure as those cutting one to four ◈ Strong Evidence [2] [8]. The researchers identified three channels: better self-assessed work ability, fewer sleep problems and less fatigue [3]. The German pilot, which went further than self-report by collecting hair cortisol samples and smartwatch data from a subset of staff, recorded 38 more minutes of sleep a week among four-day workers than in a comparison group, and about 13 fewer minutes of stress a day [14] [28]. Those physiological measures were limited in scale, but they point in the same direction as the surveys.

The mechanism also has a cost side that advocates mention less often. In the UK pilot, 62% of employees said the pace of their work increased, and the authors recorded a small rise in work intensity, from 3.64 to 3.69 on their scale, that was “weakly statistically significant” ✓ Established [1]. Some 15% of staff reported working more hours by the end, not fewer, and only 71% reported a reduction [1]. At South Cambridgeshire District Council, the first British council to trial the model, 37% of staff worked beyond their contracted hours, typically two to three extra hours a week [29]. Wen Fan, lead author of the 2025 paper, framed the core risk precisely: “When workers want to deliver the same productivity, they might work very rapidly to get the job done, and their well-being might actually worsen. But that’s not what we found” [9]. On the average, she is right. The distribution hides a minority for whom compression, not reduction, is the lived reality.

A further mechanism is economic rather than psychological: the four-day week as a recruitment tool. South Cambridgeshire began its trial because it could fill only about 80% of the posts it advertised; over 27 months, job applications rose by more than 120% and staff departures fell by more than 40%, and the council calculated annual savings of £399,263, largely from reduced agency spending ◈ Strong Evidence [30]. Half the senior managers in the German pilot said recruitment had improved [28]. In the UK one-year follow-up, 32% of organisations reported better recruitment and 50% lower turnover [31]. These are real gains for the firms concerned, and for small employers competing with larger rivals on pay they may be decisive. But Julia Backmann, who led the Münster evaluation, identified their limit: “this advantage would be lost if every company were to offer the four-day week now” [28]. A recruitment edge is a relative benefit. It cannot be summed across an economy.

The mechanisms, taken together, suggest a coherent picture. Shorter hours reduce fatigue and improve sleep, with effects that scale with the size of the cut [2] [14]. Firms that redesign processes — fewer and shorter meetings, better tools, fewer handovers — can absorb part of the lost time, and those that do not are more likely to abandon the experiment [24]. Some workers intensify their effort and some quietly work extra hours [1] [29]. Small employers gain a recruitment advantage that would erode with wider adoption [28]. Nowhere in this chain is there a mechanism that guarantees output per worker will hold in a hospital ward or on a production line, where the work is the time. Fraser of Allander Institute economists David Eiser and Mark Mitchell put the point directly: in health, social care, education, personal services and the arts, “the opportunity for productivity improvements is likely to be limited” ◈ Strong Evidence [23].

The single most important fact about four-day-week evidence is who is in the samples. Firms are not drawn at random from the economy; they apply, often pay to join, and receive coaching from organisers who campaign for the policy. In the UK pilot, 66% of participating firms had 25 or fewer employees and only 22% had 50 or more; marketing and advertising was the largest single sector, followed by professional services and charities ✓ Established [1]. Veal counted only three manufacturing firms and described participants as “self-selected and unrepresentative”, “mostly office-based workers” [7]. The South African pilot run by 4 Day Week Global in 2023 was even more concentrated: 57% of its 28 firms had ten or fewer staff, and only 22 returned all surveys [32]. The 2025 Nature Human Behaviour sample was 65% women and 73% college-educated [2]. These are not flaws in any single study. They are the predictable result of a design in which the firms most confident that the model will work are the ones that test it.

Critics have catalogued the consequences with some consistency, and the trial authors concede most of the points themselves [2] [6]. Daniel Hamermesh, the labour economist at the University of Texas at Austin whose work on time use is standard in the field, said of the UK project in 2022 that he “would give the finding very little, if any, weight” because it “tells us nothing” about whether benefits would extend beyond the volunteers [33]. Michael Sanders, professor of public policy at King’s College London, made the same point in 2025: results from motivated participants “doesn’t tell us much about what would happen if someone else tried it” [34]. Antti Kauhanen of the Finnish research institute ETLA argued that such pilots “need to have control groups, ideally randomised ones” and that “there are not well-defined measurements of productivity and wellbeing” [22]. The main methodological weaknesses are:

• Self-selection: firms opt in, so the sample over-represents employers with cultures, products and managers suited to the change ✓ Established [2] [7]
• No randomisation: even the best comparison group, 285 staff at 12 US firms, was not assigned by chance and had a lower survey response rate (58% vs 80%) [2]
• Self-reported outcomes: hours, burnout, health and in most pilots productivity are reported by staff or managers with an interest in the result [2] [9]
• Attrition: in Portugal 96 firms entered, 55 dropped out during preparation and 21 completed the six-month trial; only those that finished appear in the headline results [24]
• Short duration and selective follow-up: the UK one-year follow-up drew detailed research participation from 28 of 61 organisations [31]

Attrition deserves particular attention because it shapes the most quoted number of all, the share of firms that continue. Portugal’s government reported that 95% of participating companies assessed the experiment positively [15]. That figure describes firms that reached the end. The funnel behind it, documented by Xataka from the final report, ran from 96 companies that signed up, to 41 that entered the main phase, to 21 that completed six months, of which 17 kept the four-day structure and 4 returned to five days ✓ Established [24]. Germany’s pilot had a similar shape on a smaller scale: 45 organisations started, two large ones dropped out for economic reasons, and of those that finished, 20% went back to a five-day, 40-hour week while 73% kept some form of reduction [28] [14]. The UK pilot’s 92% continuation rate is real, but it is a rate among volunteers who completed preparation and stayed the course [1]. It is closer to a measure of customer satisfaction than of general feasibility ⚖ Contested.

The defenders of the trials have answers to most of these criticisms, and some are persuasive. Autonomy argues that the Portuguese, German and North American pilots “all find no change in the control companies compared to well-being improvements in treatment organizations”, and that “the vast majority of the data points clearly in one direction” [22]. Randomising an organisation-wide change across private firms is genuinely hard; no employer can be compelled to adopt a four-day week for the sake of a cleaner experiment [2]. The German evaluation, led by an independent university team, used comparison groups inside participating firms and physiological measures, and still found wellbeing gains [14]. The one-year UK follow-up found that improvements in health, burnout and life satisfaction had been maintained, which Juliet Schor interpreted as evidence that “the strong findings at six months are not due to novelty or short term impacts” ◈ Strong Evidence [31]. The novelty, or Hawthorne, argument advanced by the Adam Smith Institute — that morale boosts “are just temporary and eventually subside” — is weakened, though not eliminated, by gains that persist for a year [35].

Where the defence is weakest is on productivity. Across the pilots, productivity has been measured by self-assessment, by managers’ ratings or by revenue — never by output per hour on a consistent basis ✓ Established [2] [7]. The first North American and Irish pilot reported revenue up 38%, but Fortune’s account did not state how many of the 33 firms supplied revenue data, and only 27 answered the final survey [36]. Firms in that pilot rated productivity 7.7 out of 10 [36]. In Germany, where the Münster team examined financial data, “although there were slight increases in key financial performance indicators such as turnover and profit, these were not significantly different from the previous year” [14]. In Portugal, 80% of managers rated the arrangement financially neutral [15]. The most defensible summary of the productivity evidence is therefore not that output rose but that, among firms that chose the policy and redesigned their work, no collapse in reported performance was detected. That is a meaningful finding. It is a much smaller one than the headlines.

Dean Baker, the co-founder of the Center for Economic and Policy Research and no opponent of shorter hours, captured the scepticism neatly in 2022: “A policy that could lead to an increase in productivity of 25 percent, overnight, is impossibly great” [33]. A 20% cut in hours with no fall in output implies a 25% rise in output per hour, achieved in a single step and sustained indefinitely — which is the arithmetic behind Baker’s scepticism. The trial evidence is consistent with a smaller story: some slack is removed, some intensity is added, some extra hours go unrecorded, and output per hour rises by enough to offset part of the cut in firms where work is flexible [1] [29]. Elsewhere, the cut must be paid for. Until a randomised or large quasi-experimental study measures output directly, the claim that four-day weeks raise productivity should be treated as ⚖ Contested — and the claim that they do no damage, in suitable firms, as well supported.

If the productivity evidence is thin, the wellbeing evidence is thick. In the UK pilot, mean scores on the burnout scale fell from 2.8 to 2.34; 71% of staff reported lower burnout against 22% reporting higher; mental health scores rose and anxiety fell from 2.26 to 1.96 on a four-point scale ✓ Established [1]. Fatigue fell for 46% of staff and sleep difficulties fell for 40% [1]. Some 54% said it had become easier to balance work with household jobs, and 15% said no amount of money would persuade them to return to a five-day schedule [4]. Portugal’s government-backed pilot, evaluated by researchers under the Birkbeck economist Pedro Gomes, found that the share of workers who struggled to balance work and family fell from 46% to 8%, while anxiety fell 21%, fatigue 23% and insomnia 19% [15]. The second UK pilot, run in 2024 and 2025 with 17 firms and almost 1,000 workers, reported 62% lower burnout and 41% better mental health [37]. The pattern is not confined to one country, one organiser or one year.

Independent public-sector evaluations confirm the direction. In Scotland, two government bodies — the Accountant in Bankruptcy and South of Scotland Enterprise — cut weekly hours to 32 for a year for 259 staff ✓ Established [38]. At the Accountant in Bankruptcy, mental health scores rose 18.4% and work stress fell 18.4%; at South of Scotland Enterprise, the share of staff “very satisfied” with their work-life balance rose from 4% to 84%, and psychological sick days fell from 357 to 265 [38] [39]. At South Cambridgeshire District Council, evaluated by the universities of Cambridge and Salford, employee health scores rose 18 points and staff turnover fell 39% during the first 15 months [29]. Daiga Kamerāde of the University of Salford concluded that “a four-day work week maintains the quality of public services” while improving staff wellbeing [29]. These results matter because the public bodies were not recruited by an advocacy group and their service metrics are published. They point to the same conclusion as the private pilots: fewer hours, on full pay, make people feel better.

The gains are not evenly distributed, and the averages conceal a minority who experienced the four-day week as compression. In the UK pilot 15% of workers reported working more hours by the end and 22% reported higher burnout [1]. Sleep difficulties fell for 40% but did not change for 45% and rose for 15% [1]. At South Cambridgeshire, more than a third of staff routinely worked two or three hours beyond their contract [29]. In the German pilot, the desire for more family time fell from 64% to 50% — a large change, but one that leaves half the participants still wanting more [28]. The Brazilian pilot coordinated by the Getulio Vargas Foundation recorded more energy and creativity but also difficulties with deadlines and with reconciling internal and external demands ◈ Strong Evidence [40]. For workers whose clients, suppliers or patients still operate on five days, a four-day week can mean five days of availability compressed into four days of pay — a pattern the trials’ aggregate numbers are not designed to detect.

Public preferences reveal the same tension. In August 2021 an Ipsos poll found that 66% of Americans and 70% of employed Americans supported a four-day week ✓ Established [42]. In 2023 Bankrate found 81% support among full-time workers, but 54% of supporters said they would work longer days to get it [43]. A YouGov survey in Britain in October 2022 found that 54% of adults believed a four-day week should mean working at least 8.25 hours a day, and 64% were unwilling to accept a pay cut for one [44]. In other words, much of the public enthusiasm for the four-day week is enthusiasm for a compressed week — the version with the weakest wellbeing record. Gallup’s broader finding is a corrective to both camps: the quality of the work experience, including management and engagement, had two-and-a-half to three times the impact on wellbeing of the number of days or hours worked [20]. A badly managed four-day week will not rescue a badly managed workplace.

The household effects are where the four-day week is most clearly distinct from other workplace reforms. In the UK, 60% of employees said it had become easier to combine work with care responsibilities, and 62% said it had improved their social lives [4]. In Portugal, 65% of workers reported spending more time with their families [15]. The gendered dimension has not been lost on governments: Tokyo’s governor Yuriko Koike announced the city’s option for staff in December 2024 explicitly so that “women do not have to sacrifice their careers due to life events such as childbirth or child-rearing” [45]. Lithuania’s 2022 law, in force from 2023, gives public-sector parents of children under three a 32-hour week at full pay ✓ Established [46]. Japan’s demographic context explains the urgency — a fertility rate of 1.2 in 2023 and 0.99 in Tokyo [45] — but the trial evidence linking shorter weeks to birth rates does not yet exist. The care benefits are self-reported and real; the demographic hopes are speculative.

The international overwork data explain why wellbeing is the axis on which this debate turns. Japan’s 2024 white paper on karoshi, or death from overwork, recorded 883 recognised cases of work-related mental disorders in the 2023 fiscal year, a record, and found that 8.4% of employees worked 60 or more hours a week ✓ Established [47]. South Korea, which cut its legal maximum week from 68 to 52 hours in 2018, still worked 1,874 hours per worker in 2023, 132 more than the OECD average [48] [49]. For workers in those labour markets, the relevant comparison is not between 40 hours and 32 but between chronic overwork and something closer to the European norm — a range in which Pencavel’s threshold suggests the productivity cost of cutting hours is lowest [27]. The wellbeing findings of the trials are strongest precisely where the economic case is strongest too. That overlap is the four-day week’s most compelling argument.

Iceland is the case most often cited as proof that a four-day week works at national scale, and the one that least supports the claim. Between 2015 and 2019, Reykjavík City Council and the national government ran trials covering about 2,500 workers, more than 1% of the working population, who moved from 40 hours to 35 or 36 hours a week without loss of pay ✓ Established [18]. A 2021 report by Autonomy and the Icelandic think tank Alda called the trials “an overwhelming success” and stated that 86% of Iceland’s workforce had since moved to shorter hours or gained the right to them [18]. Veal’s analysis of the underlying reports found that in 61 of the 66 workplaces the cut was between one and three hours a week, with a maximum of four. “A four-day week trial would have involved reducing the working week by seven to eight hours,” he wrote; “instead the maximum reduction in these trials was just four hours” [13]. Subsequent collective agreements cut 35 minutes a week in parts of the private sector and 65 minutes in the public sector [13]. Iceland shows that modest reductions can be negotiated without disruption. It is evidence for shorter weeks, not four-day ones.

Belgium illustrates the opposite confusion. Its February 2022 labour deal, in force from November 2022, gave full-time employees the right to request working their full weekly hours in four days — 38 hours as four days of 9.5 hours, for example — with employers able to refuse in writing [16] [10]. Prime Minister Alexander De Croo framed it as giving “people and companies more freedom to arrange their work time” [16]. Uptake has been minimal: the HR services firm Securex measured 0.45% of full-time staff in January 2023 and 0.52% in November 2023, while Acerta estimated about 0.8%, roughly one in 130 ✓ Established [41] [10]. Workers told Securex they feared longer days would damage their well-being, productivity and family life [41]. The employment ministry later said the measure “was not intended to apply to all employees” [41]. Belgium is regularly listed alongside the UK and Iceland as a four-day-week country. On the evidence, Belgian workers were offered compression and overwhelmingly declined it.

Japan and the Gulf have adopted the four-day label mostly for compressed or optional arrangements. The Japanese government’s 2021 Basic Policy on Economic and Fiscal Management encouraged employers to offer an optional four-day week for workers with caring duties, retraining or side jobs, while maintaining the five-day standard [50]. From April 2025, the Tokyo Metropolitan Government allowed its staff to take one weekday off per week — previously one per four-week period — on a schedule built around monthly hours rather than a cut in them ◈ Strong Evidence [11]. Only about 8% of Japanese companies allow three or more days off a week [11]. Dubai’s “Our Flexible Summer” scheme, run from 1 July to 12 September 2025 across 21 government entities, gave one group eight-hour days from Monday to Thursday with Fridays off and another group seven-hour days with a half-day Friday; total hours were unchanged [12]. These policies may have merit as flexibility measures. They are not tests of the reduced-hours model, and their results, where published, should not be read as such.

The genuine reduced-hours evidence comes from a small number of countries, almost all through voluntary pilots. Portugal’s 2023 trial, co-sponsored by the government, cut company-reported hours by 13.7%, from 39.3 to 34 a week [15] [51]. Germany’s 2024 pilot, organised by the consultancy Intraprenör with 4 Day Week Global and evaluated by Münster, used the 100-80-100 model [14]. Lithuania’s parental 32-hour week is the only statutory reduction of this kind found in Europe, and it is limited to public-sector parents of young children ✓ Established [46]. Poland launched the most ambitious state-funded test in 2025, selecting 90 employers with more than 5,000 workers from 1,994 applications, with a PLN 50 million budget and conclusions due by 15 May 2027 [52]. Labour minister Agnieszka Dziemianowicz-Bąk called it “another civilizational step—just as we once secured the eight-hour day and free Saturdays” [52]. The history of working time suggests that comparison is not hyperbole:

Set against that history, average working hours already vary enormously across countries without any four-day policy. OECD data compiled from national sources show annual hours per worker of about 1,833 in South Korea, 1,800 in the United States, 1,598 in Japan, 1,533 in the United Kingdom, 1,498 in France, 1,421 in Iceland and 1,332 in Germany, against an OECD average of 1,736 ◈ Strong Evidence [60]. A German worker, on these figures, already works about 27% fewer hours a year than a Korean one, through shorter weeks, part-time work and longer holidays rather than a four-day mandate [60]. Korea’s own figures show how much statutory caps can achieve: the share of workers on 50 hours or more fell from 47.9% in 2002 to 12% in 2022, and annual hours from 2,228 in 2008 to 1,901 in 2022 [61]. Korea’s employers’ federation still estimates hourly productivity at 77.4% of the OECD average [49]. The cross-country pattern is that richer, more productive economies tend to work fewer hours — but as Eiser and Mitchell warn, “correlation does not prove causation” [23].

The international record therefore sorts into three groups. A handful of countries — Portugal, Germany, the UK and, prospectively, Poland — have generated real reduced-hours evidence through pilots, almost all with volunteer firms [15] [14] [1] [52]. A second group — Belgium, Dubai and, largely, Tokyo — has adopted the four-day label for compressed or optional schedules that change when work happens rather than how much of it there is ✓ Established [10] [12] [11]. A third — France with its 35-hour law and South Korea with its 52-hour cap — has reduced hours by statute without a four-day framing, and generated the only large-scale quasi-experimental evidence on what happens to employment when hours are cut by law [57] [48]. That third group, examined in section 7, offers the hardest test of the four-day case. Its results are mixed.

The policy response to four-day-week trials follows a recognisable pattern: governments commission or tolerate pilots, welcome the wellbeing results and then decline to extend them. Scotland is the clearest example. Its two-agency pilot found stable productivity, an 18% improvement in mental health at one body and a rise from 4% to 84% in staff very satisfied with their work-life balance at the other [38]. In August 2025, finance minister Ivan McKee acknowledged that the pilot “has identified several examples of more efficient and innovative working practices” — and said the government “will not be moving to a 32-hour week” ✓ Established [39]. The union PCS responded that “the question is no longer whether it can work, but whether the government is prepared to back what works” [39]. The Scottish data also contained a warning that ministers noted: at South of Scotland Enterprise, client response times slipped from 94% to 92% within target [39]. Small as that change was, in public services the political cost of any visible decline weighs more heavily than an improvement in staff morale.

England’s experience was more confrontational. When South Cambridgeshire District Council began its trial in January 2023, the Conservative government opposed it. In July 2023 local government minister Lee Rowley wrote to ask the council to “end your experiment immediately”, arguing that “removing up to 20 per cent of capacity is not something which should be acceptable for a council seeking to demonstrate value for money” [62]. The TaxPayers’ Alliance welcomed the intervention, calling the arrangement a “council clock-off” [62]. The council continued, and the final evaluation by the University of Salford found that 21 of 24 services had improved or stayed the same over 27 months, with planning applications decided on time rising from about 75% to about 95% ✓ Established [30]. A resident survey found no statistically significant difference in nine of twelve services [30]. The episode illustrates the political asymmetry: the costs of a four-day week are visible and easily framed, while the benefits accrue to staff and appear only in evaluations months later. The main policy developments elsewhere are:

• Spain: a bill cutting the statutory week from 40 to 37.5 hours, covering more than 12 million workers, was defeated in Congress on 10 September 2025 by 178 votes to 170 [59]
• South Korea: the government plans a support law for a 4.5-day week, with KRW 32.5 billion for 2026 pilots and a roadmap to wider rollout in 2027 [49]
• United States: Senator Bernie Sanders’ 2024 Thirty-Two Hour Workweek Act would lower the overtime threshold to 32 hours over four years; no state has mandated a four-day week [58] [63]
• Australia: a Senate committee recommended in March 2023 that the government trial a 100-80-100 model; no national trial has followed [64]
• Poland: the only state-funded multi-sector pilot in Europe, with 90 employers testing from 2026 [52]

Spain’s defeat is instructive because the proposal was modest by four-day standards: a cut of 2.5 hours a week, not eight [59]. Second Deputy Prime Minister Yolanda Díaz, who championed the bill, conceded on the day of the vote that it would fail but insisted that “the reduction of the working week and the wage rise” had already been won “in the street” [65]. The opposition came from the Catalan party Junts as well as the conservative PP and far-right Vox, with sectors such as agriculture, retail and restaurants most exposed [59]. In South Korea, where President Lee Jae-myung’s government is pursuing a 4.5-day week, industry objections are similar; one auto-parts executive told the Korea Times that “if the working week is shortened without easing rigid labor practices, efficiency will fall and costs will rise” [49]. These are precisely the sectors absent from the pilot samples. The risks of scaling a policy from volunteer offices to the whole economy can be assessed as follows:

Public finance estimates sharpen the concern for state employers. In 2019 the Centre for Policy Studies, a centre-right think tank, estimated that a 32-hour week across the UK public sector would cost £45 billion a year at existing productivity, or £17 billion under generous productivity assumptions ⚖ Contested [67]. Its analyst Jethro Elsden argued that “we need to raise productivity before we can afford to shorten the working week, not vice versa” [67]. That estimate assumed a mandate, not the voluntary office-based adoption the pilots tested, and it predates the public-sector evaluations. But the Svartedalen nursing home trial in Gothenburg, which moved care workers to six-hour days between 2015 and 2017, showed the arithmetic in practice: the city hired 17 additional staff at a cost of about 12 million kronor, and returned to eight-hour shifts when the budgeted two years ended [66]. Daniel Bernmar, the Left Party politician who led the trial, conceded that “more money must be spent” to cut hours within a reasonable timeframe [66]. In round-the-clock services, there is no meeting culture to compress.

Employer surveys show interest without commitment. The CIPD found that 34% of UK organisations expected a four-day week to be reality for most workers within ten years, but 66% said it would depend on efficiency gains and 68% on technology adoption [69]. A 2019 Henley Business School survey estimated that UK firms already offering four-day weeks saved £92 billion a year, with 64% reporting higher productivity — figures widely cited by campaigners, but based on employers’ own estimates extrapolated to the economy ⚖ Contested [70]. In the United States, the Gallup data show 8% of full-time employees already on four days, a share that includes many compressed schedules [21]. Accredited UK employers grew from more than 200 in January 2025 to 253 later that year [71] [68]. The trend is upward. It is also, so far, a niche of small knowledge-work firms, which is exactly the population the pilots sampled.

The policy lesson from the past four years is that pilots have succeeded in shifting the burden of proof without settling the question. Few serious analysts now argue that a reduced-hours week is unworkable for an office-based small firm; the German, Portuguese and South Cambridgeshire evaluations make that position hard to sustain [14] [15] [30]. But no government has yet been persuaded that the evidence justifies a mandate, and the reasons are not purely ideological. A mandate would apply to the care homes, building sites and shops where the trials have almost no data; it would remove the recruitment advantage that drives part of the gain; and it would have to be paid for, either through productivity gains the trials have not demonstrated or through higher costs [23] [28]. The Poland and Korea programmes, if they publish independent and comparable data, could fill part of that gap ◈ Strong Evidence [52] [49].

The hardest evidence on cutting working hours does not come from four-day-week pilots at all. It comes from national laws, which economists can study with methods closer to natural experiments. France’s 35-hour week offers the best-known case. Matthieu Chemin and Etienne Wasmer exploited the fact that the law was applied less stringently in Alsace-Moselle, which has its own local legal regime, and found “no significant difference in employment with the rest of France, which casts doubt on the effectiveness of this regulation” as a job-creation measure ✓ Established [72]. An earlier study by Bruno Crépon and Francis Kramarz of France’s 1982 cut from 40 to 39 hours found that it increased job losses among directly affected workers by between 2% and 4%, with minimum-wage workers hit hardest because their monthly pay could not adjust [75]. Portugal’s 1996 cut from 44 to 40 hours produced a different result: Pedro Raposo and Jan van Ours found that hourly wages rose, monthly earnings held roughly constant and job destruction fell for affected workers [73]. None of these laws tested 32 hours at full pay.

What the national evidence does show is that the employment-sharing argument for shorter hours — that cutting each worker’s week creates jobs for others — has weak support ◈ Strong Evidence [72] [75]. That matters because it was the original rationale for France’s 35 hours and remains part of the case in Spain and Poland [59] [52]. The modern four-day campaign mostly makes a different argument, about wellbeing and productivity, which national laws were not designed to test. The two sets of evidence therefore talk past each other. Pilot advocates can fairly say that national mandates imposed uniform cuts without the process redesign that makes their model work [24]. Critics can fairly say that a mandate is the only way the policy would reach the majority of workers, and that the pilot evidence tells us nothing about how a mandate would perform [33] [23]. Eiser and Mitchell concluded in 2021 that “without meaningful results from the pilots, mandating a transition to a shorter working week (with no loss of pay) would be a very risky policy” [23].

Reversals provide a second kind of counter-evidence, anecdotal but telling. Utah moved most of its state employees onto a compressed four-day, ten-hour schedule in 2008, hoping to save $3 million a year in utilities; a legislative audit found the actual saving was about $502,000, and audit supervisor James Behunin reported that “some agencies” were “less productive because of the new schedule” because it did not align with client needs ✓ Established [17]. The state reverted in 2011 after complaints about Friday closures [17]. In the private sector, the payments company Bolt adopted a four-day week in 2022 and later returned staff to five days, with its chief executive saying staff perks would be “on hold” until revenue grew [74]. Treehouse, an online education company profiled for its 32-hour week, abandoned it in 2016 [76] [77]. In Germany, 20% of pilot firms returned to five days [28]; in Portugal, 4 of 21 that completed the trial did [24]. Reversals are a normal feature of management experiments. They are also a reminder that the 92% continuation rate is a ceiling, not an average.

The disagreement can be summarised as a contest between two readings of the same body of evidence. Both rely on real data; they differ on what that data can be generalised to. The optimist reading treats consistency across countries and organisers as evidence of robustness [22] [3]. The sceptical reading treats consistency as the expected result of a consistent selection process [33] [34]:

Some of the strongest criticism comes from free-market think tanks, and some of it is overstated. The Institute of Economic Affairs economist Julian Jessop argued in 2022 that a four-day week “is not a realistic option” in some services, asking how doctors could “see 25% more patients in a day” and how “a bar person can pull 25% more pints in a day” [78]. That is a fair point about time-bound work, but it describes a compressed-output model the pilots did not test; their firms reduced hours by redesigning work, not by asking staff to process 25% more per hour [24]. The Adam Smith Institute’s claim that morale gains are “just temporary” is contradicted by the one-year follow-ups [35] [31]. On the other side, advocates sometimes present the UK pilot’s revenue figure and the Microsoft Japan anecdote as productivity evidence they are not ⚖ Contested [1] [25]. The Cambridge sociologist Brendan Burchell called the 2025 paper “a game-changer”; on wellbeing, he may be right [79].

Hamermesh’s judgement was made about the UK pilot in 2022, before the peer-reviewed paper, the German and Portuguese evaluations and the British public-sector trials. Some of his scepticism has been answered: there are now comparison groups, independent evaluators, physiological data and one-year follow-ups [3] [14] [31]. The core of it has not. There is still no evidence on how widespread the benefits could be, because every reduced-hours trial so far has studied organisations that chose to take part ✓ Established [2]. The genuine disagreement, stripped of rhetoric, is narrow. Both sides largely accept that shorter hours improve wellbeing for those who receive them, and that office-based firms can absorb a substantial cut through process redesign. They disagree on whether the evidence can be extended to the economy as a whole — and on that question, the sceptics are correct that the data do not yet exist [22] [23].

Separated from the advocacy, the four-day-week trials support a set of conclusions that are less dramatic than the headlines and more durable than the critics allow. The wellbeing finding is the strongest: across six countries, two peer-reviewed or university-led evaluations and several public-sector trials, cutting roughly five hours a week at full pay reduced burnout and fatigue and improved sleep and mental health, with larger cuts producing larger gains ✓ Established [3] [2] [14] [38]. The business finding is moderate: among firms that volunteered and redesigned their work, revenue and service metrics were broadly stable, and most chose to continue [1] [15] [30]. The productivity finding is weak: no trial has measured output per hour directly, and the largest study collected no productivity data at all [2]. The generalisation finding is absent: there is no randomised trial and almost no evidence from shift-based, frontline or manufacturing work [22] [23]. Ranked by strength, the evidence says:

• Strong: reduced hours on full pay lower burnout and fatigue, with a dose-response pattern at individual level [3] [2]
• Strong: compressed weeks of four long days do not share these benefits and may raise burnout [20]
• Moderate: volunteer office-based firms that redesign work show no detectable loss of revenue or service quality [14] [30]
• Weak: claims of large productivity gains, including the Microsoft Japan 40% figure, rest on revenue or sales data without controls [25] [1]
• Untested: effects in frontline sectors, under a national mandate or in firms that did not choose the policy [2] [23]

The most underreported finding is the gap between what the label promises and what most policies deliver. The four-day week with the best evidence is a reduction of about five hours, usually with a day off, on full pay, in a firm that has rebuilt its meetings and workflows [8] [24]. The four-day week most governments have adopted — in Belgium, Tokyo and Dubai — is a compression of the same hours into fewer days, and the four-day week most workers say they would accept is, by a majority, the compressed version [10] [11] [12] [44]. Iceland, the best-known national case, was neither [13]. When the label travels without the reduction, it borrows credibility from trials it does not resemble ◈ Strong Evidence [20]. Policy debates would be clearer if “four-day week” were reserved for genuine reductions and compressed schedules were called what they are.

The second structural insight concerns who funds and frames the evidence. 4 Day Week Global and Autonomy have done more than anyone to generate data on reduced hours; without them, there would be almost nothing to evaluate [22] [1]. But the trials were designed by campaigners to demonstrate feasibility, and their results were released in a format optimised for headlines — 92% continuing, 71% less burnout — while the caveats sat in the body of the reports [4] [1]. The 2025 Nature Human Behaviour paper is a model of how to report this evidence honestly, stating both its effect sizes and its likely overestimation in the same document [2]. Independent evaluations by Münster, Salford and Cambridge have generally confirmed the wellbeing findings while tempering the financial ones [14] [30]. That pattern — consistent direction, smaller magnitude under independent scrutiny — is typical of policy research produced by advocates, and it is the right lens for reading future results.

For employers, the evidence supports a cautious but positive conclusion. A small or mid-sized office-based firm that wants to cut hours, invests in redesigning its processes and monitors output is likely to see healthier, more loyal staff and no collapse in performance ◈ Strong Evidence [24] [14] [31]. It will gain a recruitment advantage for as long as competitors do not follow [30] [28]. It should expect some staff to intensify their work or quietly work extra hours, and should measure that rather than assume it away [1] [29]. It should avoid confusing compressed schedules with reduced ones [20]. And it should be wary of the headline productivity numbers that circulate in the press, which the underlying reports do not support [1] [25]. For employers in care, hospitality or manufacturing, the evidence is close to silent, and the experience of Svartedalen suggests that the cost of reduced hours there is explicit: more staff [66].

For governments, the evidence supports funding better research rather than mandates. The questions that matter for public policy — whether reduced hours raise output per hour across sectors, whether they share out employment, what they cost in public services — are exactly the ones the voluntary pilots cannot answer [23] [72]. Poland’s state-funded pilot, with 90 employers from sectors including garment manufacturing, water utilities and a philharmonic orchestra, is the closest any government has come to testing the model outside its comfort zone ◈ Strong Evidence [52] [80]. South Korea’s pilots could add evidence from an economy where hours remain long and the potential gains from reducing them are largest [49] [48]. Scotland’s decision to stop at a pilot despite positive results suggests that even sympathetic governments see the evidence as insufficient for a public-sector mandate [39]. On the current record, that judgement is defensible.

A century ago, the five-day week was a radical experiment adopted by one carmaker; within two decades it was American law [54] [56]. In 1930, Keynes expected his grandchildren to work fifteen hours a week [55]. The four-day week may follow the path of the five-day week, or it may remain, like Keynes’s forecast, an aspiration that productivity growth made possible but that labour markets did not deliver. The trials have answered one question decisively: people who work fewer hours for the same pay feel substantially better, and many employers who try it do not regret it ✓ Established [3] [31]. They have not answered the question that would turn a workplace perk into an economic policy — whether the economy as a whole can produce as much in four days as in five. The honest summary of what the four-day-week trials actually found is that they found the first answer and assumed the second.