Money for Nothing? Universal Basic Income as Health Policy

To make a point, the Marxist sociologist Erik Olin Wright (1997) once borrowed a character from the 1960s comic strip Lil’ Abner: a big blobby creature called the Schmoo that could magically provide any basic need: a burger, a baked potato, a salad.¹ Not caviar and filet mignon, but the basics. And the Schmoo could reproduce itself endlessly and costlessly, so you could share it with anyone. Wright asked readers to consider how workers and owners would distribute Schmoos if they could. Workers, he argued, would want even their bosses to have one. After all, would you rather labor for a desperate employer squeezing every last penny, or for one who knows he’ll eat whether or not the quarter turns a profit? Owners, meanwhile, would rather destroy all the Schmoos than let workers get one, since workers who don’t need wages to survive can demand to be treated decently.

The Schmoo, of course, is a metaphor for a Universal Basic Income (UBI)—a government payment to all citizens large enough to meet their basic needs. The debate about UBI has traditionally been conducted in the currency of economics: employment disincentives, inflationary pressure, fiscal cost. But for readers of this journal, there is a prior question that tends to get buried under those macroeconomic concerns: Would a UBI actually improve population health? The income-health gradient—the robust positive correlation between wealth and wellbeing—is among the most replicated findings in all of social epidemiology. On average, men in the top income quintile live nearly 15 years longer than men at the bottom; the gap for women is about 10 years.² This gradient is apparent early in life, widens with age, and is observed across countries and historical periods.³ If we could close even a fraction of that gap through implementation of an income floor, the public health implications would be enormous. But the evidence for whether we can is, to put it charitably, complicated.

The most suggestive historical evidence comes from Canada. Between 1974 and 1979, the governments of Canada and Manitoba jointly ran an experiment called MINCOME, a guaranteed annual income offered to residents of the small town of Dauphin, Manitoba—the only saturation site among the five North American Negative Income Tax (NIT) experiments of that era.⁴ Because Canada had recently introduced universal health insurance, economist Evelyn Forget was later able to link MINCOME participation to provincial health administration data. What she found was striking: hospitalization rates among Dauphin residents fell by roughly 8.5% relative to a carefully matched comparison group, with the largest reductions in admissions for accidents and injuries and for mental health diagnoses. Physician contacts declined as well, especially for mental health. More adolescents completed grade 12.⁵ Here, it seemed, was evidence that a relatively modest income guarantee—set at 60% of the Statistics Canada low-income threshold—could produce real, measurable improvements in population health and reduce the burden on the health care system.

The NIT experiments running simultaneously in American cities during the 1970s were less encouraging,⁶ and the MINCOME findings sat largely unanalyzed for decades—a casualty of political change and bureaucratic neglect. When Forget’s work was finally published in 2011, it generated a wave of enthusiasm among those who saw a guaranteed income as an upstream intervention for population health. The logic was intuitive: poverty is stressful, chronic stress degrades physiological systems, and relieving financial pressure should—over time—translate into better health.⁷ Remove the Sword of Damocles that hangs over low-income households, the argument went, and you remove a major source of allostatic load.

More recent, and more rigorously designed, experiments have complicated that picture considerably. The OpenResearch Unconditional Income Study (ORUS)—the largest randomized controlled trial of a guaranteed income conducted in the United States to date—assigned 1,000 low-income adults aged 21 to 40 in Illinois and Texas to receive $1,000 per month unconditionally for three years, while 2,000 controls received $50 monthly over the same period. The results, published in 2024 by Miller, Rhodes, Bartik, Broockman, Krause, and Vivalt, tell a nuanced but ultimately sobering story about income and health.

On the hopeful side: transfer recipients reported meaningful improvements in stress and food security, at least early on. They used more health care—including dental care, a notoriously neglected dimension of health among low-income populations—and their medical spending increased by about $20 per month relative to controls.⁸ On the less hopeful side: the mental health improvements faded by the second year, and researchers could rule out even very small positive effects on physical health across a battery of validated survey measures and blood-based biomarkers. The health effects that the cross-sectional income-health gradient would predict—were the correlation causal—lay well outside the confidence intervals. In other words, giving low-income adults $1,000 a month for three years did not move the needle on their physical health in any detectable way.⁸ A companion analysis of the Baby’s First Years Study (BFY)—a randomized trial that provided low-income mothers of newborns with $333 per month for the first 40 months of a child’s life—similarly found no effects on maternal mental health, maternal or child BMI, or overall child health as measured over four years.⁹

Here is the puzzle that should occupy anyone who cares about income policy as health policy: the income-health gradient is enormous in observational data, yet well-designed income experiments—some of them now quite large—fail to reproduce it. Why? Several explanations merit consideration.

The first is timing. Health, unlike financial stress, is slow-moving. The physiological consequences of sustained poverty—accumulated allostatic load, toxic stress responses, chronic inflammation—build over years and decades.^7,10 Three years of income support, however generous, may simply not be sufficient to undo biological damage already done, particularly in a sample of adults who have been living in poverty for most of their lives. The MINCOME effect on hospitalization, by contrast, may have been more visible precisely because the entire community received support simultaneously, and because the experiment ran long enough relative to the severity of the population’s prior need.⁵

Second, there is the issue of what economists Best, Lobel, and Pinho Neto (2026) call “productive inclusion”—the idea that income transfers can only improve health when recipients are operating below a subsistence threshold where basic needs, including medication and adequate nutrition, are genuinely unmet. In a difference-in-differences analysis of the 2012 expansion of Brazil’s Bolsa Família program—the world’s largest cash-transfer scheme—they find that a guaranteed income top-up for the extreme poor reduced hospitalization by 8% and mortality by 14%, saving roughly 1,000 lives. Medication expenditure rose about 50%, and hospital admissions for undernutrition fell by 38%. The magnitude of these health effects closely mirrors the MINCOME results, and they emerged because beneficiaries were previously too poor to afford essential medications and adequate food—the transfer relaxed constraints that were literally killing people. Critically, these effects accumulated progressively over eight years rather than materializing quickly, reinforcing the timing argument above. When the fiscal savings from reduced hospitalizations are incorporated, the program’s marginal value of public funds becomes effectively infinite: the health benefits alone more than offset the direct cost of the transfers.¹¹

This “productive inclusion” framework offers a parsimonious answer to the core puzzle. ORUS recipients in Illinois and Texas were low-income by American standards, but they were not subsistence-constrained in the way Brazil’s extreme poor were in 2012. They already had access, however imperfect, to food assistance, emergency Medicaid, and other safety-net programs. Additional cash improved their financial cushion and their stress levels, and increased their use of health care—but the physiological pathways through which extreme deprivation damages health were not the binding constraints on their wellbeing. The income-health gradient in observational data, on this reading, reflects the cumulative health consequences of prolonged deprivation rather than a relationship between current income and current health that cash transfers can quickly undo. A higher income might increase health care utilization (as ORUS found) without yet translating into better health outcomes if those constraints are not binding—though the Alaskan quasi-experiment, which found a 4 percentage-point greater increase in primary care-seeking following an exogenous dividend increase, suggests that income boosts do shift health behaviors even when the effects on health itself take longer to appear.¹²

Taken together, the experimental record also sheds light on a longstanding theoretical debate about what the income-health gradient is actually measuring. A large body of evidence, associated most prominently with epidemiologist Michael Marmot, suggests that health gradients are shaped not just by absolute deprivation but by one’s relative position in a social hierarchy—what Marmot calls the “status syndrome.”³ But note that targeted experiments like ORUS are actually well-positioned to detect a relative-status effect: the treated group does improve its standing vis-à-vis the control group, so if rank-based psychosocial stress were a primary driver of health, we should see it. The null results cast doubt on relative status as the dominant mechanism—at least over three-year time horizons. Paradoxically, a truly universal income guarantee would fail to shift anyone’s relative rank, since everyone’s floor rises together. The productive inclusion results from Brazil point in the same direction: what seems to matter most is whether absolute subsistence needs are met, not where one stands in a hierarchy. The experimental evidence, in other words, is more consistent with absolute deprivation and cumulative biological wear-and-tear as the key pathways than with relative social standing per se.

What should health policy advocates take from all of this? The evidence neither definitively supports nor definitively refutes the view of UBI as a health intervention. The MINCOME and Bolsa Família results together constitute the strongest experimental evidence that income guarantees can reduce hospitalizations, ease mental health burdens, and even save lives—but both effects emerged in contexts where recipients were severely deprived of the material prerequisites for health.^5,11 The more recent RCT evidence from the United States is discouraging for those who hoped that three-year cash transfers would close measurable fractions of the health-income gradient in a high-income country context,^8,9 though the increased health care utilization found in ORUS at least raises the prospect of longer-run gains that studies with finite endpoints cannot capture.

There is also the question of what we are comparing UBI against. The case for guaranteed income as a population health tool is not just about whether it improves health in the abstract—it is also about whether it does so more effectively and equitably than the current patchwork of means-tested programs, each with its own bureaucratic friction, stigma, and gaps in coverage.⁴ Conditional welfare programs may also generate stress: the surveillance, the compliance requirements, and the periodic threats of disenrollment.¹³ A guaranteed, unconditional income floor might not produce dramatic health improvements visible in a three-year study window, but it might reduce the cumulative psychosocial toll of navigating an adversarial benefits system. That effect would be difficult to capture in any single RCT.

It is also worth remembering that the landmark income experiments of the 1960s and 1970s—the Seattle-Denver Income Maintenance Experiment, the New Jersey NIT, and MINCOME—were conducted in a radically different economic landscape. Manufacturing employment was still robust, unionization rates were far higher, and the labor market could plausibly absorb workers whose skills and education were modest. The modest work disincentive effects observed in those experiments⁶ played out in an economy where the main question was whether a guaranteed income would make people slightly less likely to take available jobs. Today, the question may be structurally different—and the Bolsa Família results add a further wrinkle: in a setting where recipients face binding subsistence constraints, a guaranteed income can actually increase employment and earnings by raising productivity.¹¹ The canonical disincentive story appears to be a high-income-country phenomenon, not a universal law.

Almost a century ago, the economist John Maynard Keynes wrote an essay called “Economic Possibilities for My Grandchildren,” speculating that rising productivity from machines and automation would soon mean we’d need to work only a few hours a week.¹⁴ That future did not arrive on schedule. But today, with generative artificial intelligence (AI) beginning to automate cognitive tasks that once required years of training, we may be approaching a moment when the premise of those earlier experiments—that paid work is reliably available for those who seek it—can no longer be taken for granted. Andrew Yang was not wrong to put UBI on the 2020 Democratic primary agenda partly as a response to automation, even if his numbers were contested. Some economists estimate that AI and robotics could expose the majority of current occupational tasks to automation within a generation.^15,16

This changes the health policy calculus in ways we have barely begun to think through. If large-scale structural unemployment becomes a feature rather than a bug of the coming economy, the relevant comparison for UBI is no longer “cash transfer versus working poverty” but “cash transfer versus mass involuntary idleness.” The health consequences of the latter—social isolation, lost purpose and identity, despair—are well-documented and severe.¹⁷ The Case-Deaton literature on deaths of despair already shows what happens to working-class communities when good jobs disappear without replacement.^18,19 And, if AI displacement pushes larger fractions of the American workforce toward genuine subsistence constraints—toward a world where people cannot afford medications, adequate nutrition, or stable housing—then the Bolsa Família results, not the ORUS results, may be the more prophetic guide to what a guaranteed income could accomplish for population health. A guaranteed income in that context is not a supplement to employment; it may be the only institution standing between large populations and the worst health outcomes associated with economic abandonment. None of the existing experiments were designed to test that scenario. Designing them—and thinking seriously about what health infrastructure would need to accompany an income floor in an AI-disrupted economy—is the next frontier.

Wright’s Schmoo thought experiment was ultimately about power—about who benefits from others’ precarity and who suffers from it. The biomedical version of that question is equally pointed: Who bears the health burden of an economy that leaves low-income households perpetually one emergency away from catastrophe? The evidence now accumulating from guaranteed income experiments suggests that the answer is more complicated than the income-health gradient would lead one to hope. Cash transfers may open health care doors, relieve immediate stress, and increase dignity—but translating those proximal changes into durable improvements in population health appears to require more time, and possibly more structural support, than any individual experiment has yet been able to provide. That is not an argument against a guaranteed income. It is an argument for designing the next generation of experiments—and the policies they evaluate—with health outcomes and health care systems integrated from the start, rather than treated as afterthoughts to economic debates about inflation and labor supply.

References

About the Author