50 Years of UBI Experiments Already Answered the Big Question

Fifty years of experiments across five continents, over 50,000 participants in a dozen pilots. And the single most common objection to universal basic income, "but won't people just stop working?" has been tested so many times that at this point, repeating it without checking the data is a choice.

The answer, across nearly every pilot ever run, is no. People don't stop working. What they do with the money is more interesting than any op-ed prediction, and the one experiment that showed a real work reduction tells a story the critics never finish.

This is Part 1 of a series on UBI. Strictly the record. What actually happened when governments and researchers gave people unconditional cash and watched.

43 Years Running in Alaska

The longest-running UBI-equivalent in the world is the Alaska Permanent Fund Dividend, paid annually to every Alaska resident since 1982, and it's not a pilot. In 2024, that check was $1,702. Over 730,000 Alaskans received it. The fund itself, built from oil royalties, holds roughly $78 billion.

A 2019 NBER study by Damon Jones and Ioana Marinescu (later published in the American Economic Journal: Economic Policy) found the dividend had no significant effect on employment. Part-time work actually increased by 1.8 percentage points. The University of Alaska's Institute of Social and Economic Research found the PFD keeps 15,000 to 25,000 Alaskans out of poverty every year and reduced the poverty rate among rural Indigenous Alaskans from 28% to under 22%.

Alaska's inflation rate has trended below the national average since the program began. No price spikes attributable to the dividend have been documented. Forty-three years of data, and the standard predictions about dependency and inflation haven't shown up.

Alaska isn't representative of the broader U.S. economy, obviously. Small population, vast oil wealth, unique funding model. But when someone says "UBI has never been tested long-term," the PFD is sitting right there with four decades of tax records and labor data saying otherwise.

Finland Tried It. Recipients Worked More.

In 2017, Finland's Social Insurance Institution (Kela) randomly selected 2,000 unemployed people and gave them 560 euros per month with zero conditions. No work requirements, no means test. A control group of similarly unemployed people received standard benefits.

The final results from Finland's Ministry of Social Affairs and Health: recipients worked an average of 78 days during the reference period, versus 72 for the control group. Six more days of employment among the people getting free money. Recipients also reported less depression, less loneliness, and a stronger sense of autonomy.

The pilot had real limitations. Everyone in the sample was already unemployed. The amount was below Finland's poverty line. Two years is too short to capture long-term behavior. Researchers originally wanted a broader test but were constrained by political compromise. Still, the headline finding ran directly against the laziness assumption, and that's the finding that stuck.

Stockton's $500 Checks

The Stockton Economic Empowerment Demonstration (SEED) gave 125 randomly selected residents $500 per month from February 2019 through February 2021. A 200-person control group received nothing.

Full-time employment among recipients jumped from 28% to 40% in the first year, according to the UBRI interim evaluation. The control group rose from 32% to 37%. Recipients who could cover an unexpected expense with cash went from 25% to 52%. Less than 1% of the money went to alcohol or tobacco. The bulk went to food, utilities, clothing, and car payments.

That spending breakdown is one of the most consistent findings across every UBI pilot on record. The fear that people will blow cash on vices has been tested in the U.S., Kenya, India, and Namibia, and it keeps not happening.

SEED had 125 recipients. That's tiny, and the pandemic overlapped with the program, which complicates isolating what the cash itself did versus what COVID-era circumstances produced. The funder (Economic Security Project) has an advocacy mission, which is worth knowing. But the directional findings match pilots run by completely different organizations on different continents.

The Box That Sat for 30 Years

Between 1974 and 1979, roughly 10,000 residents of Dauphin, Manitoba received a guaranteed annual income through Canada's MINCOME experiment. When the federal government changed hands in 1979, the program was killed. The data was boxed up and warehoused.

In 2011, economist Evelyn Forget at the University of Manitoba pulled those boxes out and matched the original participants against Manitoba Health records. What she found: an 8.5% reduction in hospitalization rates, with the biggest drops in accidents, injuries, and mental health admissions. More teenagers finished high school. The only groups that reduced their work hours were new mothers (who took longer parental leaves) and teenagers (who stayed in school). Both outcomes were classified as positive by the research team.

That 30-year gap between data collection and analysis is genuinely unusual. A 2021 reanalysis by UBC researchers raised questions about confounders, and Forget herself acknowledged that the introduction of universal health insurance in Manitoba during the same period complicates the health findings. The education and labor results have held up under scrutiny anyway, and the basic pattern (cash didn't cause a work collapse) matches what every subsequent pilot found.

The Largest Study Ever Run

GiveDirectly's ongoing experiment in rural Western Kenya is the biggest UBI study in history. Around 23,000 individuals across 195 villages are receiving monthly payments, with 100 control villages. The study tests multiple designs: 12-year long-term payments, 2-year short-term payments, and lump sums.

The standout number from Egger et al. (2019): every dollar transferred generated $2.60 in local economic activity. A 2.6x multiplier. Households didn't withdraw from work. They shifted from wage labor to self-employment and small business formation. During COVID, a J-PAL analysis found UBI recipients experienced less food insecurity and illness.

Here's the finding I think matters most from the Kenya data: the 12-year long-term arm outperformed both the short-term and lump-sum arms, even when the total dollar amounts were identical. People who knew the money would keep coming invested it differently. Started businesses, planned further ahead. Certainty of income changed behavior more than the amount.

The Counter-Evidence Nobody Finishes Reading

The U.S. ran four Negative Income Tax experiments between 1968 and 1982, and these produced the strongest evidence for a work reduction effect. The largest, SIME/DIME in Seattle and Denver, found work hours dropped by 5% to 25% depending on the group, roughly equivalent to 1 to 5 weeks of full-time work per year.

That's the number critics cite. But the ASPE report shows who reduced work and why. Husbands barely changed. The biggest reductions came from youth, who traded work hours for school: an 11% increase in school attendance. Secondary earners (mostly women) also reduced hours, partly because the guarantee reduced pressure to hold a second household job.

Nixon proposed a version of this as the Family Assistance Plan in 1969. It passed the House. In the Senate, it died, killed by a coalition that couldn't agree on whether it was too generous or not generous enough. The SIME/DIME results arrived too late and were too mixed to save it. The political takeaway hardened into "people stopped working" — which is accurate for the aggregate hours number and a distortion of what actually changed.

What the Pilots Agree On

None of these experiments ran in the same country, the same decade, or under the same conditions. They mostly agree anyway.

Work effort didn't collapse in any of them. Alaska showed no employment drop over 43 years. Finland and Stockton both showed increases. Kenya showed a shift from wage labor to self-employment. The NIT experiments showed the largest reductions, but those were concentrated in teenagers trading work for school enrollment and mothers extending parental leave, not prime-age workers exiting the labor force.

Mental health improvements showed up in Finland, Stockton, Dauphin, Ontario (before the pilot was cancelled), and the OpenResearch study. Consistently, across wildly different populations.

Spending on alcohol and tobacco has been tracked across pilots in the U.S., Kenya, India, and Namibia. It stays negligible everywhere. The money goes to food, housing, health costs, and car payments. Probably the most durable result in the entire literature: it has never failed to replicate.

The multiplier effect from GiveDirectly's data (every dollar transferred generating $2.60 in local activity) has a parallel in domestic modeling: the Roosevelt Institute projected a $1,000/month UBI funded by new revenue could expand GDP by $2.5 trillion over eight years.

And the certainty finding from Kenya gets the least attention: 12-year recipients outperformed short-term and lump-sum recipients at identical dollar amounts. Knowing the money keeps coming changed behavior more than the amount did.

Who Benefits from the Data Staying Buried

Means-tested welfare programs employ enormous bureaucracies. Politicians on both sides use the "will they stop working" framing to sidestep the actual evidence, because the evidence doesn't fit clean ideological boxes. The NIT experiments showed effects that can be read as either positive (kids stayed in school) or negative (people worked less), depending on which sentence you stop reading at.

Which lets opponents kill programs without engaging with the data. Doug Ford cancelled Ontario's pilot 18 months in, before any results were collected, while claiming it "did not help people become independent contributors to the economy." The talking point about dependency was never actually tested because the pilot wasn't allowed to run long enough to generate data.

The incentive structure isn't complicated. Complexity in the safety net protects administrative budgets, political positioning, casework employment, and the assumption that poverty requires surveillance rather than cash.

What 50 Years of Pilots Can't Tell Us

Short pilots can't predict what people do when payments are permanent. Small samples in specific cities don't generalize to national populations. No wealthy country has ever run a full-scale, permanent UBI. The macro effects on inflation, housing prices, and aggregate labor supply at scale remain theoretical: modeled, but never observed at that scale.

The Hawthorne effect is real: people behave differently when they know they're in a study. Sam Altman's OpenResearch pilot found recipients worked about 1.3 fewer hours per week, roughly 12 minutes a day. But 87% of participants self-selected into the study, the sample was limited to ages 21 to 40, and Guy Standing's critique in BIEN called it "less a test of basic income than many previous experiments." The methodology matters, and so does who funded it.

The Question the Debate Keeps Asking

The funding question is real. So is displacement: which existing programs get cut, who loses coverage in the transition. And who controls the payment infrastructure matters enormously; a conditional payment with strings is not UBI. Those debates haven't been settled. The "will people stop working?" one has been answered, the same way, across 50,000 participants on five continents, and it keeps getting asked anyway.

Next in this series: who's funding all this UBI research, what they're building alongside it, and what they get out of the answer.