Executive summary

The evidence supports developing a phased Australian reduced-hours proposal. It does not establish that every workplace can maintain pay and services at no additional cost. The central design requirement is measurable time returned to the person: a four-day label is insufficient if hours are compressed, unpaid work increases or colleagues inherit the workload.

The selected evidence includes an international peer-reviewed employee study, the original UK pilot and its follow-up, Australian interviews including a discontinued trial, Iceland's shorter-hours experience and a UK council's published account. Together they provide credible examples and implementation lessons. They do not constitute a representative Australian national experiment. The international and UK evidence may overlap; their participant counts must not be added as independent replications. [A01–A06]

Historical reforms now extend this evidence base: substantial changes were secured through staged legislation, arbitration and collective bargaining. They strengthen the case for an implementation programme, while leaving employment effects mixed and the value of actual time distinct from a legal hours standard. A2 now uses verified Australian instruments to establish different starting hours; A3 develops successive campaign and delivery commitments. The detailed historical background remains separate review evidence.

For this What If programme, the design premise is that some people will use genuinely released time to experiment with free AI. A1 investigates whether the time arrangement can be made credible. It does not require everyone to learn AI, treat leisure as a failed outcome, or claim that these trials measured avoided AI job losses.

Current judgement: DESIGN PLAUSIBLE; DELIVERY UNRESOLVED. Carry forward unchanged ordinary weekly pay, actual reduced hours, voluntary personal use of the released time and explicit service/cost checks. A2 should compare sector-specific ways to deliver those requirements. B1's personal AI access should be available independently, including to people without a workplace time dividend.

Reader orientation, premises and confidence

This is the first of three reports on time. It hands trial findings and design limits to A2's Australian arrangements; A3 will consider the costed public proposal and adoption route. The companion B1 report designs individual AI opportunity. The integrated question is how implementation could occur by 2035.

Premises retained for design on 9 September 2026: P1, shorter paid hours plus free AI enables some voluntary experimentation; P2, personal learning by doing is the programme's preferred foundation; P3, that capability contributes to resilience against employment and malicious-AI cohesion harms. All three remain ASSUMED FOR DESIGN. Neither a trial result nor this report proves them as a complete causal chain.

Confidence is medium that well-designed reduced-hours arrangements can improve wellbeing in participating settings; low to medium for transfer to other sectors; and unresolved for nationwide costs, employment protection and cohesion effects. These are claim-level judgements, not probabilities of success. Exact primary records can support high confidence in what a study reported while giving much less confidence in Australian national extrapolation.

Pass 0 — Define what is being implemented

A genuine four-day week reduces the person's actual working time while retaining ordinary weekly pay. A compressed week redistributes unchanged hours over fewer days. A four-day service closure is a different decision again: employees can work fewer days while a service uses staggered rosters to remain open.

The 100:80:100 shorthand proposes full pay, 80% of previous hours and maintained output. It is a target, not a measured result. For illustration, 80% of 38 hours is 30.4 hours; four eight-hour days are 32 hours, a reduction of approximately 15.8% from 38. A2 now specifies instrument-based examples at 35, 37.5 and 38 ordinary hours, rather than mixing these offers. The proposed end target remains a 20% reduction from the applicable existing full-time standard, with a four-day pattern where suitable and protected equivalent time for other rosters. This is a candidate design recommendation: no legal entitlement or uniform industry schedule has been adopted. These are arithmetic examples, not a recommended universal roster. The Fair Work Ombudsman describes the maximum-hours framework and reasonable additional hours; it does not establish an existing universal right to a paid four-day week. [A07]

The pay definition needs a later employment-instrument specification. Preserving salary while removing regular overtime can still reduce a worker's ordinary lived income. A trial should therefore report base pay, usual total remuneration, leave and retirement-contribution treatment, alongside the legal definition being used. “Same pay” must not conceal a cost shifted to the worker.

Pass 1 — What the selected actual trials show

International employee evidence

Fan and colleagues' 2025 study analysed 2,896 employees in 141 participating organisations, with 285 employees in 12 control organisations in the descriptive comparison. It was not randomised. The full paper's Table 2 reports mean individual hours falling from 38.98 to 34.26, a 4.72-hour reduction, with improvements in burnout, job satisfaction and reported mental/physical health. Companies undertook work redesign before the six-month trial. Selection, mostly smaller organisations and subjective measures constrain transfer. The paper also records increased perceived job demands. Its disclosures identify research grants and state that the authors had no financial/legal connection to the trial organiser. [A01]

The implication is to measure the actual time dose and the whole redesigned job. Benefits cannot be attributed solely to deleting a day, and improved wellbeing does not establish that every participant experienced less pressure. A1 uses the paper for hours and wellbeing, not as a national output or fiscal evaluation.

UK pilot and durability

The original 2022 UK pilot involved 61 companies and about 2,900 workers. At the reporting point, 56 companies intended to continue. Revenue rose 1.4% across the 23 companies supplying the relevant start/end data; a different 24-company comparison with an earlier period reported 34.5%. The denominators and periods differ. Neither statistic alone measures labour productivity or causal profit improvement. The organiser-linked report combines company data, employee surveys and interviews. [A02]

The 2024 follow-up recorded 54 of the original 61 still operating the policy, five discontinued and two not responding; at least 31 had made it permanent. Its detailed follow-up engagements involved 28 organisations, a different denominator from the original cohort. This supports persistence among many early adopters, while leaving selection and missing-response limitations. It also prevents describing continuation as universal. [A03]

Australian evidence, including a stopped trial

Hopkins, Bardoel and Djurkovic interviewed 15 decision-makers from 15 Australian firms, with 2–85 full-time employees. Fourteen still offered the arrangement; one small marketing business stopped shortly after its trial amid substantial business disruption. Recruitment used visible adopters and referrals. This is qualitative leadership evidence, not a randomised employee or payroll audit. The authors report implementation lessons and no specific research grant or competing interests. Their sample gives limited coverage of larger, staffing-intensive services. [A04]

This adverse case is neither an Australian failure rate nor something to discard because external conditions were difficult. A resilience proposal must work through difficult conditions. A2 needs a plan for downturns, demand surges and small teams with few substitutes.

Broader services and smaller reductions

The Alda/Autonomy synthesis of Iceland's trials describes more than 2,500 workers, many moving from 40 to 35 or 36 hours without reduced pay. It covers varied public-service settings and reports broadly maintained service alongside wellbeing benefits. It is an advocacy-linked synthesis of earlier evaluations, not a new national randomised trial. Its historical 86% figure includes people gaining a right to shorter hours; it is not evidence that 86% worked four days or a 32-hour week. [A05]

South Cambridgeshire's current council account reports 21 of 24 monitored service measures improving or remaining unchanged in its July 2025 analysis, alongside a claimed annual staffing-budget saving of £399,263. It describes approximately 85% of contracted hours at unchanged pay. This is an employer's published account of university analysis; the underlying full 2025 evaluation was not independently reproduced in this review. Local recruitment conditions, agency costs and the treatment of pandemic periods matter. Do not substitute the earlier 2024 “22 of 24” figure for this later analysis. [A06]

Historical reductions since 1945: how change was made

History adds evidence about implementation at scale. It does not turn earlier five-day or shorter-hours reforms into tests of a contemporary four-day week. The following selection retains the distinctions and source IDs of the separate historical background; percentage changes in a standard are not measured changes in actual hours.

Research evidence and design table 1
Case Process and implementation Outcomes and relevance to the Australian proposal
Australia, 44 → 40 from 1948 A union application led to the 1947 Standard Hours Inquiry; awards and associated wage provisions were adjusted, with industry-level scheduling and exceptions Strong evidence of institutional and ordinary-pay change; no adequate causal national output/employment estimate established here. [HH01–02]
West German metal/electrical bargaining A 48 → 40 sequence in 1956–67 was followed by the 1984 dispute and staged settlements reaching 35 in 1995; pay compensation accompanied reductions Flexibility was part of the negotiated exchange. Hunt's final study finds the later reductions may have lowered employment; no automatic work-sharing gain. [HH03–06]
Sweden, 48 → 40 in 1958–73 Successive inquiries and legislative decisions set intermediate standards and implementation dates A documented choice to take some prosperity as leisure; this review does not establish a causal national outcome or uniform compensation across every stage. [HH07–08]
France, 39 → 35, 1998–2002 legislation/transition Deadlines differed by firm size; subsidies and bargaining shaped adoption, pay and scheduling Positive employment estimates concern particular comparisons or the whole package. A regional study found no significant employment difference and has a published correction. [HH09–13]
Portugal, 44 → 40 in 1996–97 Two implementation steps, with averaging and collective-agreement provisions Studies report reduced hours and protected monthly earnings, but employment effects differ by worker group and research design; newer adverse firm evidence remains a working paper. [HH16–19]
Quebec, 44 → 40 in 1997–2000 Social dialogue and legislation lowered the overtime threshold one hour annually, without mandated wage compensation An original evaluation finds fewer hours beyond 40 but no employment increase. This is a contrast to the same-pay proposition. [HH21–23]
Japan, general 48 → 40 transition, 1987–97 Law, phased coverage, exceptions and supporting measures Actual-hour effects were heterogeneous; the final threshold-based study does not find job creation. The nominal eight-hour change was not eight hours returned to everyone. [HH24–25]
South Korea, 44 → 40 in 2004–11 Phasing by establishment size came with changes to overtime, leave and averaging Manufacturing productivity evidence is favourable within its sample; satisfaction with hours and overall wellbeing did not move identically. The later total-hours cap is a different intervention. [HH26–28]

The Australian order is particularly relevant: it adjusted wage provisions to the shorter award week but left scheduling to industrial implementation. The later 38-hour route began through agreed/unopposed claims in 1983 and broadened in 1986. Neither episode establishes that future implementation will be costless. FWC history, HH01; 1947 order, HH02

The evidence supports staging as an implementation option, not a causal finding that staging itself guarantees success. Germany illustrates negotiated steps and compensation; Sweden illustrates legislated intermediate dates. IG Metall chronology, HH03; Swedish parliamentary record, HH08

Employment needs separate treatment. A European sector-level comparison found hours reductions without an identified employment increase. France and Portugal have competing findings; the relevant counterfactual, subsidies, wages and hiring/retention response matter. A leisure benefit can still be valuable without job creation, but it cannot be advertised as proven protection from AI displacement. Batut, Garnero and Tondini, HH20; Hunt's final German study, HH06

Implementation inference: seek a defined destination, an initial funded reduction, a binding schedule for further negotiation and expansion, and measured hours/pay/service outcomes at each step. Do not promise that an election endorsement immediately changes every workplace, or quietly replace the destination with indefinite trials. A3 applies that distinction to the campaign.

Pass 2 — What these findings permit us to infer

The evidence is enough to reject the assumption that shorter hours necessarily destroy output or service in every setting. It is not enough to guarantee a uniform national transition. That distinction leaves useful implementation work to do: select and price mechanisms capable of delivering the desired individual experience.

Research evidence and design table 2
Question Foundation judgement Consequence for design
Can people receive less work time at unchanged pay? Demonstrated in selected arrangements Record each arrangement's actual hours and remuneration
Can wellbeing improve? Credible, with selection and measurement limits Include wellbeing but also workload, fatigue and adverse experiences
Is output/service always maintained? Not established Use service-specific quality, timeliness, coverage and cost measures
Are all arrangements permanent? No Include continuation, withdrawal and termination conditions
Will national AI displacement fall? Not established by this corpus Carry this as a later programme outcome, not a trial finding
Will social cohesion improve? Not established by these trials Measure inclusion, relationships and fairness separately
How many will learn AI in their own time? No defensible estimate here Use transparent uptake scenarios; do not invent a conversion rate

The sample sizes, organisations, respondents and outcome denominators describe different units. A study may report positive average wellbeing without having complete output data. A service indicator can remain stable while its delivery cost rises. An organisation can retain a popular benefit without proving that the benefit caused business growth.

Pass 3 — Economics and the individual's usable time

For unchanged output with a proportional hours reduction r, measured output per paid hour must rise by 1/(1-r) − 1, assuming the same headcount and output definition. At r = 0.20, the arithmetic is 25%. That need not mean asking people to work 25% harder: eliminating low-value work or waiting can produce the ratio. It does mean the required change cannot simply be wished away.

Where output depends on maintaining staffed hours and efficiency does not change, headcount must instead rise by 25% to preserve coverage after a 20% hours reduction. As an illustrative example, if labour accounts for 60% of costs, a 25% increase in that component raises total cost by 15%, before offsets and secondary effects. These are sensitivities, not estimates for hospitals, aged care or any Australian industry.

Recruitment, vacancies, overtime, sickness, rework and turnover may provide offsets, but each must be measured in the same setting and period. National adoption can reduce an early adopter's recruitment advantage. The proposal cannot count the same efficiency gain once to preserve output, again to pay for additional staff and again to finance the AI entitlement.

For the individual, usable time is more specific than rostered time. An illustrative accounting identity is:

net discretionary time = reduction in actual work + commuting time saved − additional obligatory care/other work − extra recovery burden.

This does not devalue care or rest. It explains why a nominal day cannot be entered as eight hours of AI learning in a national model. A person who chooses family time has not breached the programme's purpose or forfeited pay. The employer should not inspect a private AI account to decide whether the person earned the shorter week.

Consider a hypothetical nurse, designer and unemployed carer. The designer's employer may reorganise tasks; the nurse's service needs a funded coverage plan; the carer receives no employer-funded hours reduction. All three may benefit from personal AI access. Their differences make a combined programme more useful, but make a single workplace rule insufficient as its inclusion policy.

Pass 4 — A transferable Australian design envelope

The following are proposed design requirements, not enacted rules:

Research evidence and design table 3
Requirement Observable check Owner proposed for A2
Less actual work Paid and unpaid hours, all jobs, out-of-hours contact and nominal days off Participating workers and employer; independent evaluator validates
Pay retained Base and usual total earnings, employment conditions and any offsets Employer/payroll, worker representatives and qualified employment advice
Services maintained or explicitly funded Timeliness, coverage, quality, queues, errors and service-user experience Service operator and evaluator
Sustainable workload Intensity, fatigue, unsafe pace and displaced work Workers, operator and relevant safety governance
Transparent full cost Additional staffing, redesign, capital, agency use and recurrent funding Operator finance team and evaluator
Fair distribution Part-time/casual, shift, disability, caring and lower-income outcomes Joint design process with affected people
Private voluntary time No requirement to study, start a business or support a campaign Programme and workplace agreements

For part-time and casual workers, A2 must compare equivalent percentage reductions at protected earnings, increased hourly pay, predictable scheduling and other options. A1 does not select an instrument for people whose hours fluctuate. An average full-time benefit cannot substitute for an explicit offer to them.

The right to disconnect supplies a relevant existing boundary around unreasonable out-of-hours contact; it is not a complete enforcement mechanism for this proposal. A2 must examine awards, agreements, coverage and practical enforcement with appropriate expertise. [A08]

Compare genuine reduction against compressed hours, predictable rosters, additional leave and shorter days. Those alternatives may solve particular problems; identify their effects without renaming unchanged hours as the proposition's time dividend. No public authority, employer or union has committed to implement this report.

Four pathways — the time arrangements A1 carries forward

These are the shared programme pathways, applied to A1's time question. Trial findings in A01–A06 inform mechanisms and limits; they do not supply four observed national outcomes.

Research evidence and design table 4
Pathway Individual experience and time mechanism Cost, responsibility and barrier to resolve
WF1 — Business as usual Current arrangements evolve; some employers already reduce hours while others retain existing schedules. People find learning time themselves. Existing employer/public costs continue. Establish the baseline's actual hours, pay, flexibility and service performance in A2; do not assume no improvement.
WF2 — Incremental reform More voluntary trials and flexibility release time for selected workers; conventional learning support expands. Employers and existing programmes support selected changes. A2 tests who misses out and whether favourable-trial economics transfer to staffing-intensive services.
WF3 — Individual opportunity, uneven time The personal AI offer operates, but fewer paid hours remain concentrated in feasible settings. Workers and carers without additional discretionary time still qualify for AI access. Personal-access costs sit in B2; A2 prices partial time coverage and records barriers. More compute cannot resolve an unfunded nursing roster or remove a caring obligation.
WF4 — Integrated success by 2035 Funded sector arrangements deliver genuinely reduced hours at unchanged ordinary weekly pay, maintain services and make some additional time usable for voluntary learning. Distribution is addressed explicitly. Employers, worker representatives and authorised funders secure the required redesign, staffing and agreements. A2 must establish coverage, pay/service protection, funding and contingency plans; B3 must examine the combined resilience outcomes.

The same hours/pay/workload/service ledger applies to all four. Compare benefits as well as costs, including where BAU or incremental reform works well. All pathways face the same demand surge, staffing shortage and fiscal-pressure cases. Do not define WF4 by successful office trials while omitting difficult sectors or people outside employment.

To build WF4, turn the remaining barriers into actions. Missing coverage funding → A2 prices redesign and additional staffing → employers, worker representatives and an authorised funder negotiate an affordable arrangement → obtain an agreement and baseline before a demonstration → retain service/pay protections and defer expansion if these cannot be sustained. Unequal usable time → A2 examines casual/part-time treatment and caring constraints → responsible service and employment bodies specify practical options → B3 tracks actual reach and discretionary time → report partial achievement if those barriers persist. Roles are proposed, not appointed. The existing backcast supplies the checkpoint windows.

Pass 5 — Backcast and handoff to 2035

This is the WF4 time-delivery backcast, judged against the programme's common success criteria. WF2 and WF3 can provide useful partial implementation without being reported as the full proposition achieved. A1's current evidence judgement remains unchanged: a designed success endpoint is not a demonstrated national result.

Research evidence and design table 5
Checkpoint Required output Failure response
Late 2026 Accepted definition of hours/pay, sector options and outcome ledger Revise ambiguous or unfunded offers
2027–2028 Costed candidate arrangements and an honestly evidenced public proposal Narrow claims if costs or informed support are unresolved
2029–2030 Separately authorised voluntary demonstrations spanning more than favourable office settings Pause expansion where service, pay or workload requirements fail
2031–2032 Longer follow-up including downturns and employee transitions Redesign financing and rosters; report terminated arrangements
2033–2034 Sustainable negotiated expansion with subgroup evidence Retain successful local arrangements without asserting national completion
2035 Report actual coverage, hours, pay, service and distribution alongside the remaining gaps Distinguish achieved arrangements from the original ambition

These dates are planning windows. Evidence and authority control progression. The programme brief separately addresses the next-election window; A1 does not estimate voter support or design electoral targeting.

Proposed accountability follows the milestone: the human research owner reviews the research design in 2026; the research team supplies priced options and the One World Community reviews the public proposal in 2027–2028; participating employers and worker representatives own any later agreements and staffing commitments; independent evaluators report service, pay, workload and distribution through 2035. These are proposed roles, not appointed organisations. Readiness evidence is respectively an accepted definition, a funded option, signed agreements and baseline records, then longitudinal outcomes and sustainable accounts. The community cannot authorise a national employment rule.

If national adoption is delayed, continue designing voluntary sector arrangements and independent personal access; revise the coverage goal rather than calling local adoption national success. In an adverse funding or staffing case, defer expansion, preserve existing obligations and publish the unmet service requirements. Do not recover costs by silently withdrawing the promised pay protection.

A2 handoff: preserve the genuine-reduction definition; cost both efficiency-led and additional-staffing routes; include casual/part-time and non-employed people; retain stopped trials; measure actual discretionary time; and do not assume that A1 supplied an AI-learning effect. B1 handoff: personal entitlement need not wait for national working-time reform. A3 handoff: describe evidence limits, including failures and sector costs, in any future public proposal.

Objections and unresolved evidence

“The trials prove that a nationwide shift pays for itself.” They show workable arrangements in selected settings. Missing cost and output data cannot be replaced by continuation rates. A2 needs sector economics and an account of costs borne by service users and other workers.

“Mixed results mean we should abandon the proposition.” They narrow the route. A partial-hours step, a funded roster change or a different sector sequence can still advance genuine time release. An underfunded universal design can be rejected without rejecting every shorter-hours arrangement.

“AI productivity will supply the missing efficiency.” That is an additional unmeasured input in this corpus. Specify the technology, quality-adjusted gain, training and operating costs, and who receives the gain before including it in a financing case.

“The extra day must be spent learning to justify the policy.” That changes the voluntary personal-learning premise into a condition. Evaluate uptake in consenting research participants; do not make private behaviour an employer entitlement test.

This review does not estimate a national average treatment effect, provide a complete sector map or exhaust all international failures. The latest council analysis is appraised at the employer-summary level. Detailed withdrawal explanations, long-term macroeconomic effects and Australian sector transfer remain named A2 evidence tasks. These limits permit continued design; they do not justify claiming national rollout readiness.

What this design can justify now

Recommend proceeding to A2's sector and fairness design under the envelope above. The current evidence makes that work worthwhile. It does not yet justify a funded national commitment, a universal cost-neutrality claim or recruitment into a live demonstration. The design envelope remains subject to human review before detailed downstream choices are treated as settled.

References and source locations

All sources were accessed on 9 September 2026. Report findings map to the public source register; third-party source files are linked, not republished.

Glossary

Actual hours: paid and unpaid work actually performed. Compressed week: unchanged hours across fewer days. Output: quantity adjusted for relevant quality; revenue is not automatically output. Transferability: whether a finding can reasonably apply in another population or setting. Selection bias: differences associated with who joins or remains in a study. Backcast: a dependency-based route from an intended future to present action.

Authorship and method

Research by Rick Molony, developed with AI assistance. This is bounded critical research and implementation analysis, not an independent audit, field trial, procurement quote, approved budget or legal opinion.

Historical source trail

The HH references map to the public source register, which provides direct sources, access-depth qualifications and limitations for each case. A1 uses that completed review without claiming a new full-paper audit. Australian instrument evidence is documented in A2 and the public source register.

Suggested citation

Rick Molony (2026). “What four-day-week trials and historical hours reductions establish”. Report A1, What If Research, Our Resilient World. Research dated 9 September 2026. CC BY 4.0.

How this page was derived: the full substantive source body was rendered after the declared public-context transformations, removal of private delivery/workflow material and omission of the private change log.