The universal basic income idea keeps returning, and the evidence on work

Every generation or so, an idea arrives with a friendly face and a one-sentence pitch. Give everybody money. No forms, no caseworkers, no conditions. The idea is called a universal basic income, and it has an unusual political coalition: technologists who think robots are coming for the jobs, progressives who want a cleaner safety net, and a handful of libertarians who think it is better than the welfare bureaucracy. It sounds like a bargain.

It is worth treating seriously, and it is worth treating critically. My thesis is this: a national guaranteed income would cost far more than its fans admit, it would reduce work among the people it reached, and it would confuse a transfer of money with a restoration of purpose. I will state the best case for it first, because it has one.

Terms

A universal basic income is a regular cash payment from the government to every citizen or resident, at a level meant to cover basic needs, without a work requirement and, in the pure version, without a means test. A guaranteed income is a looser term: it can mean payments to everyone below a threshold. A negative income tax is a design in which a household with income below a set level receives a payment that shrinks as its earnings rise, so that the government pays a fraction of the shortfall. All three are different ways to give cash, and they differ mostly in who is eligible and how the payment phases out.

The case for it, at its strongest

Here is the argument as its best advocates would make it.

Simplicity first. America runs dozens of overlapping programs for the poor, each with its own rules, offices, applications and cliffs. A single cash payment would replace that tangle with one transfer, cut administrative cost and treat recipients as adults who know what they need. Second, automation. If machines take over work faster than new jobs appear, a society needs a way to share the gains, and a basic income would give people a floor. Third, the conservative lineage. Milton Friedman proposed a negative income tax in Capitalism and Freedom in 1962, arguing that it would give the poor cash, preserve market incentives and replace a patchwork of programs with a single efficient one. Fourth, human flexibility. With a floor under them, people could take a risk, start a business, care for a parent or retrain without fear.

That is a respectable case, and part of it is right. The existing system is a mess, and Friedman was right that cash is a more respectful form of help than a maze of in-kind programs. Where I part company is on the word "universal", on the evidence about work and on a premise that sits under all of it.

What the experiments found

America has actually run the experiment, at a scale larger than most people realize. In the 1970s the federal government sponsored a set of negative-income-tax experiments. The two largest were in Seattle and Denver, run together as the Seattle and Denver Income Maintenance Experiments, known as SIME and DIME. Families were randomly assigned to receive guaranteed incomes at different levels for a limited period, with a control group that received nothing.

The question was whether guaranteed payments would change how much people worked. They did. Participants who received payments worked fewer hours than those in the control group. The reductions were smallest for husbands and larger for wives and for single mothers. The researchers who ran them reported the reductions plainly, and they became part of the political debate about guaranteed income, a debate that had already seen President Nixon's Family Assistance Plan pass the House in 1970 and then fail to win approval in the Senate.

I would ask the reader to be fair to the evidence. The experiments were temporary, and a family that knows a payment will end behaves differently from one that expects it to last a lifetime. Some of the reduced hours went to schooling, to job search and to child care, which are not frivolous. But the direction was consistent across the sites: more money with no strings meant less work. That should surprise no one who has taken a first-year economics course. Pay people not to work, even gently, and some will accept.

Finland

The most cited recent experiment is Finland's. From January 2017 to December 2018 the Finnish government paid a monthly sum of 560 euros to a randomly selected group of two thousand unemployed people, without conditions, and compared them to a control group on ordinary unemployment benefits. The results, reported in 2019 and 2020, were mixed in a way that deserves a straight account. Recipients reported better well-being, less stress and more trust. On employment, the preliminary results reported in 2019 found essentially no change, and anyone who wants the final employment figures should read the Finnish agency's own report.

I want to be fair here, because supporters point to Finland and say that the fear of reduced work was refuted. It was not, for two reasons. First, the participants were unemployed people, who already were not working, and what they did with the extra money tells us little about what would happen if it went to everyone, including those who are working. Second, the Finnish trial replaced an unemployment benefit that penalized people for taking a job with a payment that did not, which is a fix to a specific problem and not a test of universality. Finland shows that removing a penalty on work helps. It does not show that paying everyone a living wage leaves work untouched.

The arithmetic

Now the part that the advocates tend to skip, and the one that any honest argument demands: do the sum.

Take a payment of one thousand dollars a month to every adult. That is twelve thousand dollars a year, which is not lavish. The adult population of the United States is on the order of two hundred sixty million, so I will use that round figure and label it as approximate.

260,000,000 adults × $12,000 = $3,120,000,000,000.

That is about three trillion dollars a year, every year. For scale, the total public debt stood at $40.56 trillion on November 5, 2026, according to the Treasury. A new annual program costing $3.12 trillion is equal to about 7.7 percent of the entire national debt added each year: 3.12 divided by 40.56 is roughly 0.077. And the national debt, as do the arithmetic on $40 trillion and the interest bill it carries explains, already carries an interest bill that grows with every dollar added to it. You can follow the running total on the national debt clock.

Advocates answer that the program would replace other programs. Partly it could, but the poor are not the only ones receiving the payment: a universal program sends the same check to the middle class and the wealthy. If you want to keep the cost down you must either shrink the payment, which defeats the purpose of a living income, or add a means test, which means you no longer have a universal program but a negative income tax with a phase-out, and the benefit cliffs and marginal rates that come with it. This is the same trap that benefit cliffs can make a raise cost a single mother real money describes in the programs we already run. And to pay for it, taxes must rise, which reduces work from the other side. You cannot tax people to fund a transfer without affecting what they earn.

A negative income tax shows the same trade-off in miniature. Suppose, purely as a hypothetical, that the government guarantees a family of three twenty thousand dollars and reduces the payment by fifty cents for every dollar the family earns. A family with no earnings receives the full twenty thousand. A family that earns ten thousand receives fifteen thousand, for a total of twenty-five thousand. A family that earns forty thousand receives nothing, since the payment has fallen to zero at that point. Notice what the fifty percent phase-out does: every extra dollar earned between zero and forty thousand is worth only fifty cents to the family, which is a marginal tax rate of fifty percent before any ordinary taxes. Lower the phase-out rate and the program reaches further up the income scale and costs more. Raise it and the work penalty grows. There is no setting of the dial that makes the trade-off go away, and that is the lesson Friedman's own proposal quietly taught.

The rebuttal, in numbered points

1. A payment that does not depend on work reduces work. The 1970s experiments found fewer hours worked among the families who received payments. This is the opposite of what the sales pitch promises.

1. The cost is not a rounding error. At twelve thousand dollars an adult, the program runs about three trillion dollars a year on these round figures. A smaller payment fails as a living income, and a means test brings back the complexity the program was supposed to eliminate.

1. Cash is not a cure for a loss of purpose. A person without a job suffers more than a loss of income. The research on long-term unemployment is clear that it brings lower well-being, weaker family ties and weaker community involvement. A check addresses the first and leaves the rest. Replacing the paycheck with a transfer does not replace the work.

1. The automation argument has been wrong every time. The fear that machines will make human labor obsolete is two centuries old. Farm employment fell from most of the workforce to a sliver, and the nation did not run out of work. It found new work, as Creative destruction, why the buggy whip makers lost and America won shows. A policy that treats the next wave as different, and builds a permanent dependency on that bet, risks building the wrong thing.

1. Welfare reform already tested the opposite approach. The 1996 welfare law added work requirements and a time limit to cash assistance, and employment among single mothers rose afterward. The evidence from that era points the other way from the guaranteed-income promise, toward the view that expectations of work help people.

The dignity of earned success

Now the part that cannot be put in a spreadsheet, and the one I believe matters most.

A job is not only an income. It is a way of being useful to other people, and being seen to be useful. It is where most people learn discipline, form friendships, find a spouse and earn the respect that does not come from a government. The oldest moral traditions of the West treat work as part of what it means to be human. The book of Genesis speaks of work as part of the human condition, and the Apostle Paul told the Thessalonians that if anyone would not work, neither should he eat, which is a hard saying, but one that reflects the belief that contribution is part of community.

I would not push that too far. A society should care for those who cannot work: the disabled, the very old, the very young, the sick. Conservatives have always said so, and they should say so again. The distinction is between caring for those who cannot and paying everyone whether they can or not. A universal basic income erases that distinction, and with it the idea that a person earns his place by what he gives.

Many conservative thinkers have argued that a person's life has meaning when he has created value for others. You do not have to share every view of every thinker to see the point. Money handed over without a task attached is not worth less in dollars. It is worth less as a source of purpose.

What a conservative would do instead

None of this means shrugging at poverty. Fixing a cruel safety net is a worthwhile project. A conservative alternative would:

• Consolidate overlapping programs, and return more authority to states, which can experiment.
• Smooth the benefit cliffs so that a raise never leaves a family poorer.
• Keep work requirements and make work pay, through a well-designed earned income credit, a program that began as a conservative idea, as described in The EITC began as a conservative idea and now pays out billions in error.
• Cut the costs that make life expensive, such as housing and energy, so that wages go further.
• Leave room for private charity and the family, which can see what a Treasury check cannot.

Alaska's Permanent Fund Dividend, paid from the earnings of a fund built mainly on oil and mining revenue since 1982, is often cited as a working example of a universal payment. It is, within limits. But it is small, it is funded by a natural-resource fund and not by taxes on work, and it was never meant to be a living income. It is a dividend, not a wage.

The case in brief

1. A universal basic income pays everyone regardless of work, and the 1970s Seattle and Denver experiments found that guaranteed payments reduced hours worked.
2. Finland's 2017 to 2018 trial found better well-being and no clear employment gain in its preliminary results, but it tested unemployed people and removed a penalty on work, and so does not settle the question for a universal payment.
3. At $12,000 per adult, a national program costs about $3.12 trillion a year on round figures, which is more than the country can add to a debt that stood at $40.56 trillion on November 5, 2026.
4. Work provides purpose, connection and dignity that a check cannot, and care for those who truly cannot work is different from paying everyone.
5. Better answers exist: consolidate programs, remove benefit cliffs, keep work requirements, and lower the costs that eat the paycheck.

What you can do: find out which programs your own state runs for working families, and ask your state representative whether the benefit cliffs in them have ever been measured.