Would Anyone Still Work?

If cash is given to people with no strings attached, how much of the labour supply actually disappears — and who does the disappearing?

For three years, the largest guaranteed-income experiment in American history quietly ran in Texas and Illinois. One thousand low-income adults received $1,000 a month, unconditionally, while a control group received $50. The results, published by NBER in 2024, are the closest thing the debate over automation and income has to hard data. They show a labour market that bends under a guaranteed income, but does not break — and bends unevenly, concentrated almost entirely in two groups: young adults choosing education over early employment, and single parents choosing their children over a second shift. Two-parent households and non-parents over thirty, the largest share of the sample, showed no statistically significant change in employment at all. This paper walks through what the data actually says, where the honest disagreements are, and why a result that looks reassuring at the individual level raises a harder question once you scale it to a national policy.

• Participants receiving $1,000 a month for three years reduced work by 1.3–1.4 hours per week on average, and were about 2 percentage points less likely to be employed than the control group.
• The effect was not evenly distributed. Two-parent households and non-parents over 30 — most of the sample — showed no statistically significant change in employment or hours worked.
• Single parents worked roughly 2.8 fewer hours a week and were about 4 percentage points less likely to be employed, driven largely by increased caregiving.
• Adults under 30 were about 4% less likely to be employed and worked 1.8 fewer hours a week, with qualitative interviews pointing to increased pursuit of education.
• Total individual income (excluding the transfer itself) fell by roughly $1,500 a year relative to the control group — the labour-supply reduction was not offset by new income-generating activity elsewhere.
• There was no measurable effect on business formation, despite participants reporting greater interest in starting one, and no measurable change in the quality of jobs held.

Guaranteed-income pilots have run for decades — the negative income tax trials of the 1970s, Alaska's oil dividend, casino payments to the Eastern Cherokee. What made the 2024 OpenResearch/NBER study different was scale, duration, and design: a genuine randomized controlled trial, three full years, unconditional cash, with a research team led by Eva Vivalt, Elizabeth Rhodes, Alexander Bartik, David Broockman, and Sarah Miller tracking outcomes through detailed surveys, administrative records, and a mobile app with response rates above 95%.

The headline number — a moderate reduction in labour supply that was not offset by other productive activity — has been read two ways since publication, and both readings are defensible from the same data.

The first reading, favoured by guaranteed-income advocates, is that the aggregate effect is small and the underlying story is humane: the labour that disappeared was disappearing into childcare and education, not idleness. A wider synthesis published by the Economic Security Project in 2024, drawing on nine separate RCTs and over 10,000 families, reinforced this pattern — the reductions clustered almost entirely among single parents and young adults, while the broadest demographic categories showed close to nothing.

The second reading, favoured by skeptics, takes the same numbers and asks what happens at scale rather than in a 1,000-person pilot. As discussed in a WBUR/On Point segment on the findings, an hour or ninety minutes less work per person sounds trivial until it is multiplied across a labour force of 100,000 or more — at which point it becomes a measurable macroeconomic effect, not a rounding error. The NBER authors themselves are careful on this point: their estimate is a partial-equilibrium result from a small guaranteed-income pilot, not a general-equilibrium forecast of what happens if the policy were rolled out nationally, where wage effects, price effects, and behavioural responses at scale could all differ from what a 1,000-person trial can capture.

Neither side is wrong. The honest position is that the data answers a narrower question than "would America stop working" — it answers "did a specific group of low-income Americans work less when given $1,000 a month with no conditions," and the answer to that narrower question is: modestly, unevenly, and mostly in ways that look like people making reasonable trade-offs rather than opting out of the labour force.

Every conversation about a future with less human labour eventually runs into the same practical objection: if you give people income untethered from work, won't they simply stop working? The guaranteed-income literature is the best empirical proxy available for that question, even though none of these pilots were designed to test an automation-driven transition specifically — they test income replacement, not the underlying cause of that replacement, and the two may produce different behavioural responses.

The Labour Machine model on this site treats labour-force participation as a variable, not a constant, in part because of results like these. A "high automation" scenario that assumes near-total workforce exit at high dividend levels is not supported by the current evidence. A scenario that assumes zero behavioural response is equally unsupported. The honest range, based on what has actually been measured, sits closer to a single-digit percentage point shift in participation for a transfer at this scale — concentrated in caregivers and students — than to either extreme.

This paper synthesizes published, peer-reviewed and working-paper research rather than presenting original data collection. The primary source is Vivalt, Rhodes, Bartik, Broockman, and Miller, "The Employment Effects of a Guaranteed Income: Experimental Evidence from Two U.S. States" (NBER Working Paper 32719, 2024), a randomized controlled trial of $1,000/month unconditional transfers to 1,000 low-income individuals over three years, against a $50/month control group of 2,000. Findings are cross-referenced against the Economic Security Project's 2024 synthesis of nine guaranteed-income RCTs covering more than 10,000 families.

• Behavioural responses observed in these pilots (three-year, means-tested-eligible populations) are treated as directionally informative for — but not a precise forecast of — a permanent, universal guaranteed income at national scale.
• Labour-supply responses to a guaranteed income are assumed to be a reasonable, though imperfect, proxy for responses to an automation dividend, since both represent unconditional non-labour income.

• No existing pilot has run long enough, or at national scale, to capture general-equilibrium effects — how wages, prices, or employer behaviour might shift if guaranteed income were universal rather than targeted.
• All cited pilots targeted low-income populations; effects for middle- and higher-income households are not directly evidenced by this data and may differ substantially.
• The NBER study population was concentrated in two states (Texas and Illinois); regional cost-of-living and labour-market conditions may not generalize nationally.

• Vivalt, E., Rhodes, E., Bartik, A., Broockman, D., & Miller, S. (2024). The Employment Effects of a Guaranteed Income: Experimental Evidence from Two U.S. States. NBER Working Paper No. 32719.
• Economic Security Project (2024). Addressing Misconceptions in 2024 Guaranteed Income Research.
• OpenResearch Lab, NBER Working Paper: Employment — Findings.
• On Point, WBUR (2024). The results are in on America's largest universal income experiment.