Universal Basic Income Economics: Real Data from 12 Pilot Programs (2026)

TL;DR. Twelve major completed UBI pilots — Finland, Stockton, Kenya, Brazil, India, Namibia, Iran, Ontario, Spain, Germany, Compton, Hudson — produced more consistent results than the political debate suggests. Employment effects were small or zero. Mental health and financial-stress improvements were large and consistent. Cost-benefit picture depends on whether you're modeling rich or poor country contexts. None of the pilots is a clean test of "AI automation forces full UBI" — they are tests of partial cash transfers at modest scale. The economics are real but more nuanced than either side claims. Below: every pilot, what it actually measured, and what the data implies for AI-era policy.

The premise this article is testing

If generative AI displaces 20-40% of routine knowledge work over the next decade — the median forecast from the studies in our GDP impact piece — labor policy has to respond. The four serious policy responses on the table are:

1. Universal Basic Income (UBI) — flat cash payments to every adult, conditional on nothing
2. Negative Income Tax (NIT) — phased cash payments that taper as income rises
3. Job Guarantee — government-funded employment for anyone who wants a job
4. Sectoral wage subsidies / EITC expansion — targeted income support tied to work

This piece focuses on UBI because it's the policy with the most actual empirical evidence — twelve completed or substantially-completed pilots spanning rich and poor countries. The other three have far less real-world data.

The pilots tested here are not full national UBI implementations. They're partial cash transfers, typically 12-30% of median income, given to a defined population for a defined window. The question they answer is "what happens at this dose?" — not "what would happen if we paid every adult $24,000 a year forever?"

The 12 pilots compared

| Pilot | Years | Country | Sample | Amount/month | Key result | |---|---|---|---|---|---| | Finland Basic Income Trial | 2017-2018 | Finland | 2,000 unemployed | €560 | No employment effect; wellbeing up | | Stockton SEED | 2019-2021 | US (CA) | 125 | $500 | Full-time employment up; wellbeing up | | GiveDirectly Kenya | 2017-2029 | Kenya | 23,000 | ~$22 (UBI-level) | Investment up; health up; no work decline | | Madhya Pradesh UBI | 2011-2013 | India | 6,000 | ~$3 (children) / $6 (adults) | Nutrition up; school attendance up | | Namibia BIG | 2008-2009 | Namibia | 1,000 | N$100 | Child malnutrition halved; employment slight up | | Iran cash transfers | 2011-2016 | Iran (national) | 75M | ~$45 | Labor supply unchanged or slight up | | Ontario Basic Income | 2017-2018 (canceled) | Canada | 4,000 | C$1,400 | Incomplete; early data showed health improvements | | Spain IMV | 2020-ongoing | Spain (national) | 800k households | €462 avg | Targeted income floor; mixed implementation reviews | | Germany "Mein Grundeinkommen" | 2021-2024 | Germany | 122 | €1,200 | Mental health up; work hours unchanged | | Compton Pledge | 2021-2023 | US (CA) | 800 | $300-$900 | Financial stability up; employment unchanged | | Hudson HudsonUP | 2020-2025 | US (NY) | 25 | $500 | Wellbeing up; new business formation up | | Cherokee Nation EBCI | 1996-ongoing | US (NC) | All tribal members | $4,000-12,000/yr | Long-term: education up, mental health up, no work decline |

Notes: "no employment effect" or "employment up" generally means change vs control group, statistically significant unless noted. Amounts are nominal at time of pilot.

Pilot-by-pilot detail (the ones worth the depth)

Finland Basic Income Trial (2017-2018)

Design: 2,000 randomly selected unemployed people aged 25-58 received €560/month tax-free for two years, with no obligation to seek work or report on activities. Control group continued under standard unemployment benefits.

Headline finding: No statistically significant difference in employment days vs control group. Treatment group reported better mental wellbeing, lower stress, higher trust in government and institutions.

What the press reported: "Finland's basic income experiment failed."

What the research actually concluded: The official report (Kela, 2020) explicitly said the program improved wellbeing without harming employment. The "failed" framing was a political narrative — the incoming Finnish government had no appetite for expanding the program regardless of results.

What this tells us about AI-era UBI: Even at €560/month — a meaningful but not transformative amount — there's no evidence of work disincentive. The mental-health and wellbeing improvements were measurable and durable.

Stockton SEED (2019-2021)

Design: 125 randomly selected Stockton residents living in census tracts at or below the city's median household income received $500/month for 24 months. Randomized control group of 200.

Headline findings (from the published JAMA paper, 2021):

• Full-time employment increased from 28% to 40% among recipients vs from 32% to 37% in control (difference statistically significant)
• Income volatility decreased significantly
• Self-reported anxiety and depression dropped significantly (clinically meaningful effect size)
• Spending breakdown: 37% on food, 22% on home goods/clothes, 11% on utilities, 9% on auto repair/transport — predominantly basic-needs categories

What this tells us about AI-era UBI: Cash income at modest levels can boost labor force attachment, not reduce it, by enabling people to take risks (looking for better jobs, going part-time during job search) they couldn't otherwise afford. This is the strongest single counterargument to the "UBI will destroy work ethic" claim.

GiveDirectly Kenya UBI (2017-2029)

Design: ~23,000 people across 295 Kenyan villages, four treatment arms:

• Long-term UBI: $22.50/month for 12 years (a "true" UBI test)
• Short-term UBI: same amount for 2 years
• Lump-sum: equivalent amount up front
• Control villages

Published interim results through year 6:

• No measurable reduction in work effort across any treatment arm
• Long-term UBI villages: 25% higher productive investment (livestock, small business) than control
• Lump-sum villages: highest one-time investment but more variable downstream outcomes
• All treatment arms: better mental health, lower cortisol (biomarker for chronic stress)
• All treatment arms: lower intimate-partner violence
• Long-term UBI shows the most durable economic gains

What this tells us about AI-era UBI: In low-income contexts, UBI relieves binding capital and credit constraints that prevent productive investment. The long-term-vs-lump-sum comparison is the most important data point in the entire UBI evidence base — it suggests durability of payment matters as much as amount.

Madhya Pradesh UBI (2011-2013)

Design: Two villages in rural India had every adult and child receive a basic income for 12-18 months. Control villages received standard welfare.

Findings:

• Child nutrition improved measurably (weight-for-age up significantly)
• School attendance increased
• Productive work hours increased (people had cash to repair tools, buy seeds, etc.)
• Women's economic independence increased (recipients controlled their own funds)
• Indebtedness decreased

What this tells us about AI-era UBI: In low-income agricultural contexts, UBI compounds with existing economic activity rather than substituting for it. The economic effect is large per dollar but the dollar amounts required are small. Direct translation to AI-displacement-era rich-country contexts is weak.

Iran Universal Cash Transfer (2011-2016)

Design: Iran replaced fuel and bread subsidies with cash transfers of ~$45/month to almost every Iranian (~75 million people). This is the only example of a truly national near-UBI in a large country.

Findings:

• Labor supply changed by less than 1% (no measurable disincentive effect)
• Income inequality fell significantly
• Recipient households used the cash predominantly for food, education, and debt repayment
• Inflation effects were the largest negative — the cash transfers contributed to broader inflationary pressures
• Program was gradually phased back as oil revenues fell post-2014

What this tells us about AI-era UBI: Even at national scale across a large country, no significant work disincentive emerged. The implementation challenge was fiscal (funding the program) and inflationary (large cash transfers without offsetting tightening) rather than behavioral.

Stockton-pattern US pilots: Compton, Hudson, Magnolia Mother's Trust, plus 20+ others

A wave of US guaranteed-income pilots followed Stockton, all in the $300-1,000/month range over 12-36 months. Aggregate findings across them (Mayors for a Guaranteed Income, 2025):

• No employment decline in any pilot
• Modest employment increases in 6 of 23 pilots that reported full labor outcomes
• Consistent improvements in self-reported mental health (16 of 17 pilots that measured it)
• Consistent improvements in financial stability (debt reduction, savings up, income volatility down)
• Spending patterns: ~80% on basic needs categories (food, housing, transport, utilities)

What this tells us about AI-era UBI: The Stockton result was not a fluke. The pattern is consistent across ~25 US pilots in different cities, demographics, and pilot designs.

Cherokee Nation Eastern Band Casino Dividend (1996-ongoing)

Design: Not a pilot but a long-running natural experiment. Eastern Band Cherokee tribal members have received per-capita casino dividends (currently $4,000-12,000/year) since 1996.

Findings (Costello et al., Duke; long-term follow-up):

• Children whose families crossed out of poverty due to dividend showed better educational outcomes, lower behavioral problems, higher rates of high school completion
• Adult recipients showed lower rates of mental health and substance abuse problems compared to demographically similar non-recipients
• No measurable reduction in work hours

What this tells us about AI-era UBI: This is the closest thing to a long-term, durable, generationally-tested UBI program in any high-income context. The findings are robustly positive. The amounts are similar to what most national UBI proposals target.

What the pilots collectively tell us

Stripping away ideological framing, four findings hold across the dataset:

Finding 1: Modest work disincentive at best, none at most pilots

Across 12 pilots, employment effects range from -0.5 percentage points (Finland, statistically insignificant) to +12 percentage points (Stockton, statistically significant). The bulk of pilots cluster near zero employment effect. This is the most robust finding in the literature.

The strong-pro-UBI claim (that UBI will let people pursue meaningful work) is partially supported by the labor-mobility data from Stockton — recipients took the risk of leaving bad jobs to find better ones. The strong-anti-UBI claim (that UBI will destroy work ethic) is not supported in any pilot at any payment level.

The honest caveat: pilots typically test payments of 12-25% of median income. We have no rigorous evidence on what happens at payments of 75-100% of median income, which is what a "full" UBI implementation would require.

Finding 2: Large and consistent mental health benefits

Every pilot that measured mental health outcomes found significant improvements. The effect sizes are clinically meaningful, not just statistically detectable.

This finding is so consistent across rich and poor countries, different program designs, and different amounts that it's reasonable to treat it as a robust empirical fact. The mental-health benefit is the strongest single argument for UBI as policy.

The mechanism is straightforward: chronic financial stress is one of the largest known predictors of anxiety, depression, and stress-related physical illness. Relieving that stress at the margin produces measurable health benefits.

Finding 3: Financial stability dominates the spending profile

The "UBI will be wasted on alcohol and lottery tickets" claim has been tested in every pilot. The spending data uniformly shows 75-85% of UBI cash going to food, housing, utilities, transportation, education, and debt repayment. Discretionary and "vice" spending is a small minority everywhere.

Finding 4: The economics scale differently in rich vs poor countries

In low-income agricultural contexts (Kenya, India, Namibia), UBI compounds with existing economic activity and produces large productivity gains per dollar. In high-income contexts (Finland, US), the economic gains per dollar are smaller but the mental-health, financial-stability, and time-use benefits are larger.

This is important for AI-era policy: the case for UBI in rich countries is primarily a mental-health and financial-stability case, not a productivity case. The case in poor countries is primarily an economic-development case.

What the pilots do not tell us

Four important questions remain unresolved by the existing evidence:

1. Macro-scale labor effects at full UBI levels

Every pilot tested partial cash transfers. None tested a payment that fully replaced labor income. The behavioral question — "if we paid every adult $24k/year, would labor supply collapse?" — remains open. Plausible answers range from "no, the Iran data suggests behavior changes are tiny even at scale" to "yes, the lab-experiment data on income shocks suggests significant work-hour reductions at high payment levels."

2. Inflation effects

The Iran experiment is the only data point on inflation at scale. It showed meaningful inflationary pressure, though confounded by other macro factors. The honest answer is: we don't know what a $3 trillion/year US UBI would do to consumer prices.

3. Fiscal sustainability

A $1,000/month US UBI to all adults is roughly $3.3-4.1 trillion per year — about 12-15% of GDP. Net of program consolidations and tax clawbacks, ~$1.0-1.8 trillion per year, or 4-7% of GDP. That's a substantial new tax base requirement that no major economy has demonstrated it can sustainably build.

4. Long-term work-ethic / culture effects

The Cherokee Nation data is the only long-term high-income evidence and it's positive, but it's one population in one specific context. Whether multi-generational national UBI would produce different cultural effects than time-limited pilots is genuinely unknown.

How does this connect to AI displacement?

The economic case for UBI as an AI policy response rests on the assumption that AI will create large-scale unemployment. The current evidence (see our GDP impact piece) suggests the bigger near-term effect is not unemployment but wage stagnation in displaced sectors. UBI addresses the unemployment scenario directly but addresses the wage-stagnation scenario only indirectly.

This is causing serious labor economists to increasingly favor hybrid approaches: a modest UBI (say, $500-700/month) layered with wage subsidies (expanded EITC, sectoral subsidies for AI-complement skills training) and a complement-rather-than-substitute industrial policy.

The pilots discussed here are the strongest empirical case for the UBI component of such a hybrid. They are not the case for UBI as a stand-alone full-replacement policy — there's just not enough evidence there yet.

The 12-line implications matrix

For the impatient, the empirical case in 12 lines:

1. UBI does not destroy work ethic at the dosages tested. This is a robust finding.
2. UBI produces large and consistent mental-health benefits. This is the strongest finding.
3. UBI dollars predominantly fund basic needs, not vice. This is well-documented.
4. UBI in low-income contexts has high economic returns per dollar.
5. UBI in high-income contexts has smaller economic returns but larger wellbeing returns.
6. UBI compounds with existing economic activity; it doesn't generally substitute.
7. UBI funded by consolidating other programs (food stamps, housing vouchers) is approximately revenue-neutral at low dosages.
8. UBI funded as an addition to existing programs requires substantial new tax base.
9. UBI at full-replacement levels has never been tested.
10. UBI as a sole policy response to AI displacement is probably under-targeted.
11. UBI as one component of a hybrid policy (with wage subsidies, training, sectoral policy) has the strongest empirical case.
12. The "UBI obviously works" and "UBI obviously won't work" framings both lose against the actual evidence.

What to track over the next 24 months

If you follow this space, four data feeds are worth subscribing to:

1. GiveDirectly Kenya year-9 and year-12 results — the only true long-term UBI experiment, due 2027 and 2029 respectively
2. Spain IMV expansion data — Spain's program has gone from pilot to national permanent; first credible cost-effectiveness study expected 2026
3. Cherokee Nation longitudinal follow-up — 30-year cohort data due in 2026-2027
4. US Mayors for a Guaranteed Income aggregate — 60+ US pilots have completed by 2026; the meta-analysis paper expected 2027 will be the largest UBI evidence synthesis to date

Calculators related to this

We don't yet have a UBI-specific calculator, but the following help frame the economics:

• AI Headcount Equivalent — Convert hours saved to FTE equivalents (useful for sizing AI displacement)
• Job Automation Risk Score — Estimate task-level automation risk
• Job Automation Risk Analyzer — Task-level analysis (more detailed)
• AI vs Human Cost Analyzer — Total cost: AI pipeline vs human team
• AI Salary Premium — Estimate AI-skill salary lifts (where the wage-distribution problem is concentrated)

Bottom line

The UBI debate would benefit enormously from being based on the 12 actual pilots rather than rhetorical framings. The evidence shows: small or zero work disincentive, large mental-health benefits, basic-needs spending patterns, and high economic returns in low-income contexts. The evidence does not show: what happens at full-replacement payment levels, whether large national programs cause inflation, or whether multi-generational UBI has different cultural effects than time-limited pilots.

For AI-era policy, UBI's role is most defensible as part of a hybrid: modest UBI to address wage stagnation and provide a financial floor, combined with wage subsidies, training, and industrial policy to address the complement-vs-substitute distributional problem the IMF identified. UBI alone is a 70%-good answer to a 100%-complex question.

The honest framing for the next 5 years: "What does $500-700/month of unconditional cash do to a US workforce undergoing AI-driven labor disruption?" That's the question the existing pilots have already substantially answered. The answer is: it helps. Less than the strongest UBI advocates claim, more than the strongest critics admit.

---

Pilot data from primary source publications: Kela (Finland), JAMA (Stockton), GiveDirectly published reports (Kenya), Davala et al. (Madhya Pradesh), Haarmann et al. (Namibia), Salehi-Isfahani & Mostafavi-Dehzooei (Iran), Mayors for a Guaranteed Income (US pilot aggregate), Costello et al. (Cherokee Nation). Part of AI Economy Hub's labor economics series; for related coverage see AI's Real GDP Impact 2026 and Which Industries Hire AI Engineers 2026.