A Public "Accept-the-Offer" Work Platform: Does It Exist, and Is It Crazy?

The idea isn't insane. Pieces of it already run at scale around the world, from India's rural job guarantee to a fully evaluated universal job guarantee in Austria. But no Western country has built the thing Chris describes: an app-based, on-demand public work marketplace that anyone can open on any day, pick a paid task, and get paid at a guaranteed wage floor. The closest real-world versions are last-resort job guarantees (MGNREGA, Argentina's Jefes, Austria's Marienthal pilot, France's "zero long-term unemployment" territories) and city day-work programs (Albuquerque's "There's a Better Way"). The gig-app interface part is the new piece. The mandatory prison-labour add-on is the weakest part of the proposal, legally and on evidence. A voluntary, fairly paid version, focused on skills and release, is defensible.

1Key Findings

1. Real-world analogues of a "last-resort employment" system

Job guarantee / employer of last resort (MMT and others). The Levy Economics Institute describes the job guarantee as "a public option for jobs," explicitly "an employment safety net." That is almost exactly Chris's framing. Levy's US model offers $15/hour plus benefits worth 20% of wages. It projects 11–16 million enrollees and costs of $378–$415 billion a year. A rival design by Mark Paul, William Darity Jr. and Darrick Hamilton, in a March 2018 paper for the Center on Budget and Policy Priorities, estimates "a total annual program cost of $543 billion, or just under 3 percent of GDP," employing 10.7 million workers (9.7 million full-time equivalents). These are model projections, not observed results.

Austria – Marienthal (Gramatneusiedl), 2020–2024. This was the world's first universal job-guarantee field experiment. It offered voluntary, tailored jobs to every long-term unemployed resident: carpentry, renovation, gardening, elder care, office work. The evaluation by Maximilian Kasy and Lukas Lehner (Oxford) combined randomized waves, a pre-registered synthetic control and comparisons with other municipalities. It found "strong positive impacts" on employment, income, security, social recognition, time structure and collective purpose. Everyone offered a job after the two-month preparatory phase accepted it. Sources differ on eligibility: 9 months of unemployment per the study page versus 12 months in INET's summary. INET Oxford reports that the Austrian government has since committed €50 million to extend the model. The pilot was tiny, though (the first wave was 31 people), so it proves feasibility, not national scalability.

France – Territoires zéro chômeur de longue durée (TZCLD). Since 2016, "employment-purpose enterprises" have hired long-term unemployed people on permanent contracts at the minimum wage (SMIC); by end-2024, according to a June 2025 Cour des comptes report, the experiment covered 83 territories, 86 such enterprises and 3,290 people. They do useful work that is meant not to compete with existing businesses. The 2025 official evaluation by DARES and France's Haut-Commissariat à la Stratégie et au Plan found a +74 percentage-point causal effect on employment over 24 months. It also found a net public cost of €11,300–€13,700 per full-time job per year, "far from the budget neutrality originally promised." An October 2025 Assemblée nationale finance amendment (CF343), drafted with the TZCLD association, states that "6,000 people have left long-term unemployment thanks to the project in 83 territories." This is the best Western evidence on both halves of the trade-off: it works, and it isn't free.

India – MGNREGA. This is the world's largest workfare program. It is a legal right to 100 days of unskilled paid work per rural household, and it reached 51 million households in 2016. Imbert and Papp found that it crowded out private employment roughly one-for-one and raised rural wages about 5%. Those effects fell mostly in the dry season, and indirect wage gains were worth up to half of total benefits for the rural poor. A large experiment by Muralidharan, Niehaus and Sukhtankar found that when the program was run better, 86% of the resulting income gains came from non-program earnings, meaning higher private wages and employment. So a credible public job offer sets a wage floor that private employers must match. Supporters call that the point; critics call it the cost.

Argentina – Plan Jefes y Jefas de Hogar (2002–2005/09). Argentina launched this after the 2001 collapse. It paid unemployed heads of household in exchange for up to 20 hours a week of community work, training, or subsidized private work, and it was administered through municipalities. More than 70% of participants were women. Critics note it was never truly universal, and local and movement-based administration raised patronage concerns.

Ethiopia – Productive Safety Net Programme (PSNP). The user's brief pairs it with Argentina, but PSNP is Ethiopian. It is a large rural public-works-for-food/cash program for chronically food-insecure households.

South Africa – EPWP and Presidential Employment Stimulus (PES). PES launched in October 2020 and has created more than 1.5 million work and livelihood opportunities. Its participants are 83% youth and 66% women, and it had a budget of R42 billion to March 2024. Its biggest component placed young people in schools as education and general assistants. SALDRU at the University of Cape Town is studying local multiplier effects. It shows a government can scale public employment within months when there is political will.

US New Deal (WPA/CCC). These are the historical proof that a Western state can directly employ millions on infrastructure and conservation during mass unemployment.

Hungary – közmunka (the cautionary tale). Under the Fidesz government, public work became the main activation tool. It paid around HUF 50,000 a month against a minimum wage of HUF 73,815, and between 2010 and 2016 it cost HUF 1,300 billion (€4.15 billion). Only 13.8% of participants moved into the regular labour market in 2014. Researchers documented mayors using placements as patronage in poor areas, and the program inflated employment statistics. It shows what happens when a guarantee pays below the floor and is run as a work test.

Denmark/Nordic "flexicurity." Flexicurity combines easy hiring and firing, generous benefits and strong activation. It isn't a job guarantee, but it shows that making people engage in exchange for support is compatible with a high-trust welfare state.

Canada today. Canada has pieces but no guarantee:

• Job Bank is a listings board, not an offer engine.
• Employment Insurance is income replacement.
• Canada Summer Jobs and youth employment programs are wage subsidies for employers.
• Ontario Works is the province's last-resort assistance. It was created under the Harris government in 1996–97 as Canada's first work-for-welfare program. It had three tracks: job search, unpaid "community participation" placements, and paid employment. By fiscal 2001 it ran close to 99,000 placements, according to the Fraser Institute. Qualitative research (Vaillancourt; articles in Social Policy & Administration) found the placements often failed to lead to jobs and supplied "subsidized and cheap labour" to agencies. That is the Canadian brand-damage Chris would need to design around.

US city programs. Albuquerque's "There's a Better Way" is the closest thing to Chris's on-demand model. It launched under Republican mayor Richard Berry in 2015. A van picked up panhandlers for a five-hour shift of city beautification, paid $9/hour (about $45) in cash the same day, and connected them to services. The city reports 6,666 jobs provided since September 2015. Its successor, "Better Way Forward," has employed more than 50 people as "Duke City Ambassadors," and 5 graduates have moved into full-time outside jobs. Chicago, Seattle, Denver and Dallas copied the model. Local critics (Joy Junction) noted that visible panhandling persisted, and early figures were thin: 309 participants, 17 employed and 5 housed at one early count.

2. Private gig/temp platforms that already do "accept the offer"

The "tap to accept a shift" mechanism is mature in the private sector:

• Instawork operates in the US and Canada.
• Wonolo, Shiftsmart, Bluecrew, WorkWhile and Veryable operate in the US.
• Indeed Flex started as Syft in the UK.
• Coople operates in Europe.
• TaskRabbit and Handy cover errands and home tasks.
• PeopleReady (formerly Labor Ready) runs the traditional day-labour branch model.

The key design split is worker status. Indeed Flex, Bluecrew and WorkWhile employ workers as W-2 employees, with taxes withheld and some benefits. Wonolo and most delivery apps use 1099 contractors, and Wonolo charges a per-hour "trust and safety" fee. Much of the comparative information online is published by the platforms themselves (Indeed Flex and Instawork comparison pages), so treat it as marketing.

In Ontario, the Digital Platform Workers' Rights Act came into force on July 1, 2025. It guarantees minimum wage for ride-share, delivery and courier workers, but only for "engaged" time from accepting a task to completing it. Waiting time isn't covered, which the Workers Action Centre calls a major gap. A public platform could fix exactly this by paying for availability blocks rather than only per task.

No government runs a gig-style public-work app of this kind that I could find. Public employment services list jobs; they don't guarantee a paid task today.

3. Evidence on effectiveness, costs and design problems

• Placement into regular jobs is the weak spot. Card, Kluve and Weber's meta-analysis covered 199 estimates from 97 studies. It found that "public sector employment programmes are less effective" than job-search help or training. Their 2018 update of more than 200 studies found larger effects for the long-term unemployed and in recessions, which is exactly a guarantee's target group and timing.
• Displacement and wage pressure. MGNREGA's one-for-one crowd-out shows that public jobs pull people out of low-end private work. Niskanen Center critics argue that a $15-plus-benefits US guarantee would impose a "massive" cost shock on low-wage employers. Cato calls the idea a "jobs guaranteed economic disaster." Supporters reply that firms paying less than a decent floor should adapt.
• Stigma and patronage. The Hungary and Ontario Works experiences show that coercion, sub-minimum pay and local political control turn a safety net into a punishment.
• Cost. France's roughly €12,000 net cost per job per year is the most credible Western benchmark. Offsets come from reduced benefits, but they rarely reach full self-financing.
• Wellbeing. The strongest, most consistent finding is non-economic. Structure, recognition and social contact improve sharply, which matches the 1930s Marienthal sociography of unemployment's damage.

Gig-economy downsides a public version must avoid. These include pay only for engaged time, no benefits, deactivation without appeal (Instawork's own worker reviews complain of this), algorithmic dispatch, and one-strike cancellation rules.

Immigration and labour-supply shocks. Chris's intuition is partly borne out in Canadian data. Youth unemployment hit 14.2% in April 2026, per The Hub, and teen unemployment reached 20.2% in Q1 2026. Ottawa has since tightened the Temporary Foreign Worker Program:

• It refuses low-wage applications in metro areas with 6%+ unemployment. Officials say this cut eligible low-wage applications 70%.
• It cut caps to 10% of an employer's workforce.
• From July 2026, the low-wage threshold is 120% of median wage.

Ontario even declined a federal rural-cap increase, with Labour Minister David Piccini citing youth unemployment above 15%. But ESDC notes that TFWP workers are about 1% of the workforce. CFIB research argues that TFWs mostly fill jobs youth don't take, and economists also blame AI and minimum-wage increases. A job guarantee interacts with supply shocks in two ways. It absorbs slack automatically, but its cost rises with inflows. Nearly every design therefore limits eligibility to citizens and permanent residents, or to those with work authorization.

4. The mandatory prison/criminal work idea

Canada already has it, partially.

• CORCAN, a special operating agency of Correctional Service Canada, runs 103 workshops in 36 of 53 federal institutions. Its work covers furniture, textiles, construction, printing, laundry and military-vehicle refurbishment.
• Pay under Commissioner's Directive 730 is $6.90/$6.35/$5.80/$5.25 a day across levels A–D. There is a $2.50 allowance, and $1.00 a day for inmates who refuse all assignments. The 1981 top rate was $7.55.
• The 2013 changes under the Harper government added mandatory deductions for room and board (22%) and telephone-system administrative costs (8%), per the Office of the Correctional Investigator's 2022-23 Annual Report, and ended CORCAN incentive pay. The same OCI report puts take-home pay at "about 46 cents an hour" for the most common pay level after deductions, and notes the 1981 top rate of $7.55 a day equals $23.07 in today's dollars.
• Refusing work "without reasonable excuse" is a disciplinary offence under s.40(p) of the Corrections and Conditional Release Act. Sanctions range from reprimand and loss of privileges to fines and extra duties.
• In court, Guérin v. Canada, 2018 FC 94, rejected inmates' Charter s.7 and s.12 challenges to the pay cuts. It held that payments are incentives for program participation, "rather than as compensation for work," so inmates aren't employees under the Canada Labour Code. The 2019 appeal (2019 FCA 272) is understood to have been dismissed. I could not verify its text.
• Conflicting information: one legal-information site claims room-and-board deductions no longer apply as of 2026. I could not confirm this from CSC primary sources.

International law. ILO Forced Labour Convention No. 29 permits compulsory labour only as a result of a court conviction, and only under public supervision. Prisoners may not be "hired to or placed at the disposal of private individuals, companies or associations." Work for private firms must be voluntary, without threat of losing privileges. A "mandatory program to replace imported labour in private industry" would collide directly with this.

United States.

• The 13th Amendment exempts "punishment for crime" from its slavery ban.
• States that removed their own exceptions by ballot: Colorado (2018), Nebraska and Utah (2020), Alabama, Oregon, Tennessee and Vermont (2022), and Nevada (2024, about 60%).
• States that rejected removal: Louisiana (2022) and California's Prop 6 (2024, 46.7% yes). Both defeats partly came down to ballot wording and fears about mandatory work programs. The issue cuts across party lines.
• UNICOR (Federal Prison Industries) workers "typically earn between 23¢ to $1.15 per hour," according to the Federal Bureau of Prisons, which says only 8% of work-eligible inmates take part and about 25,000 are on the waiting list.

Does it reduce recidivism? Modestly, and it depends on the design.

• PREP, the Bureau of Prisons' 1990s study of more than 7,000 releases, found UNICOR workers 24% less likely to reoffend, plus better in-prison behaviour. It used matched comparisons, not randomization.
• CrimeSolutions (US National Institute of Justice) rates correctional industries only "Promising," based on just 4 studies.
• A female-inmate study found no UNICOR effect on recidivism.
• Norway gives the strongest causal evidence. Bhuller, Dahl, Løken and Mogstad used random judge assignment and found that rehabilitation-focused prison cut reoffending by 29 percentage points over five years. Among previously non-employed people, job-training participation rose 34 points and employment rose 40 points. The mechanism was training and reintegration, not forced output.
• After release, New York's Center for Employment Opportunities, a transitional work-crew program, cut recidivism 16–22% in an MDRC randomized trial. Its employment gains faded after a year, and a four-city Midwest replication (TJRD) found no recidivism effect.

2Where the idea is strong, and where it's weak

Strong:

• The core insight is right. Long-term unemployment is corrosive, and a guaranteed offer measurably improves lives (Marienthal, TZCLD).
• The "accept-the-offer" user experience is a genuine innovation. It lowers the barrier to near zero, which is what people at the bottom need (Albuquerque's same-day cash pay worked for this reason).
• It has cross-partisan appeal. The left frames it as a right to work and wage floor (Levy, Darity/Hamilton). The right frames it as work-first reciprocity over handouts (Harris's Ontario Works, Berry's Albuquerque, Hungary's Fidesz).

Weak:

• Public jobs rarely launch people into private careers on their own.
• There's a tension between "abundant work" and "useful work." Make-work erodes legitimacy.
• Wage floors impose costs on low-wage employers.
• Local administration invites patronage.
• The cost is real: roughly €12,000 net per job per year in France.
• Mandatory prison labour for production risks ILO violations, Charter litigation and displacing free workers, and its evidence base is thin.

3Recommendations: what a well-designed version could look like

The big swing: a "Public Work Exchange." This would be a municipal–provincial–federal app on which any resident with work authorization can, on any day, accept a paid task. Tasks would include park cleanup, snow clearing for seniors, grocery and errand runs for homebound people, library and school support, tree planting, home-care companionship, and digitization. It would pay at least the provincial minimum wage, for scheduled availability blocks, not only engaged minutes.

1. Voluntary, never a sanction. Marienthal worked partly because it was unconditional. Hungary and Ontario Works show what coercion does. Keep existing benefits intact; the Exchange is an add-on option.
2. Wage floor at or slightly above the minimum wage, with same-day pay. Albuquerque proved that immediacy drives uptake. The floor also disciplines exploitative employers, which MGNREGA's evidence shows.
3. An employee-status public employer. Tasks should be delivered through municipalities, non-profits and social enterprises, TZCLD-style. Include WSIB coverage and an appeal right for any "deactivation." This is the anti-Uber part.
4. An "additionality" rule. Tasks must not replace unionized public jobs or existing contracts. France's non-competition test and Hungary's rules are the models. A local board should vet the task catalogue.
5. A ladder built into the app. Tier 1 is day tasks. Tier 2 is multi-week crews with a supervisor, following the CEO crew model. Tier 3 is subsidized placement with private employers, plus training credits. The meta-analyses say placement comes from training and job-search help, not from the public job itself, so bundle them.
6. Automatic counter-cyclical budget. Funding should expand with regional unemployment, because active labour-market programs work better in recessions. It could be partly financed by redirecting Ontario Works employment-assistance spending and EI top-ups. Be honest that it won't be fully self-financing.
7. Eligibility tied to work authorization. The Exchange should run alongside TFWP rules like the current 6% unemployment refusal-to-process trigger. An Exchange vacancy count could even become the evidence employers must answer before importing low-wage labour.
8. Pilot first. Start in one CMA, such as St. Catharines–Niagara, with a pre-registered evaluation like Kasy and Lehner's.

For justice-involved people: reframe "mandatory" as "guaranteed."

• Offer every prisoner a voluntary, Norway-style training-plus-work track paid a real wage, minus transparent deductions for restitution and family support.
• Route people into the Public Work Exchange on release, since release is when CEO-style transitional jobs showed effects.
• Keep any compulsory work public-only and non-commercial to stay within ILO C29.
• This keeps the "contribute to production" goal without state-enforced cheap labour competing with free workers.

4Caveats

• Marienthal and TZCLD are small, local pilots, and scaling may change the results. The Levy and Paul–Darity–Hamilton figures are projections.
• MGNREGA and Jefes operate in developing economies with very different labour markets.
• The TFWP-to-youth-unemployment link is contested. Ottawa and CFIB dispute it, while some economists support it.
• Several gig-app comparisons are vendor-published. The Albuquerque results are largely city-reported.
• I could not verify the 2019 FCA Guérin text or the claim that CSC room-and-board deductions have ended. Some US ballot-measure percentages come from news reports rather than certified results.