As governments debate guaranteed income programs to combat poverty and inequality, they face a recurring question: if you give low-income families cash with no strings attached, will parents stop working, and how will it affect their home lives? To answer this, policymakers and economists still rely heavily on a series of large-scale government field experiments conducted in the United States and Canada during the late 1960s and 1970s.
A new study published in the Journal of Public Economics re-examines the data from those historic trials for two-parent families, incorporating newly digitized archival records and applying modern statistical standards. The researchers find that the long-standing consensus surrounding these experiments is largely wrong: the cash programs reduced work hours far more than previously believed, did not cause a spike in divorces, and noticeably improved marital satisfaction and happiness.
Revisiting the negative income tax
Economists Chris Riddell of the University of Waterloo and W. Craig Riddell of the University of British Columbia set out to audit and re-evaluate the five major North American income maintenance experiments, which took place in New Jersey (including a site in Pennsylvania), Gary, Seattle, Denver, and Winnipeg. These pioneering trials were designed to test a policy known as a negative income tax.
Under a negative income tax, a household receives a guaranteed minimum cash benefit from the government. As family members earn money from working, that government benefit is gradually reduced at a set percentage rate until their earnings reach a break-even threshold and the payments phase out completely. While universal basic income proposals often dominate public discussion today, the authors note that many economists consider income-tested programs like a negative income tax to be much more fiscally realistic, and cities such as Barcelona and Seoul have recently launched pilot programs using this exact design.
For decades, the prevailing view in economics textbooks and policy reviews was that the 1970s programs caused small, mostly statistically insignificant reductions in work effort, with the sole exception of the combined Seattle and Denver trial. At the same time, the Seattle-Denver experiment produced a widely publicized finding that guaranteed income dramatically increased marital breakups, a conclusion that chilled political support for the policy in Congressional hearings. The Riddells suspected that both of these takeaway lessons were distorted by outdated statistical methods, incomplete datasets, and flaws in how the original experiments were carried out.
Digging through the archives and testing the data
In the case of the Canadian experiment in Winnipeg, known as Mincome, high inflation and shifting political priorities caused the government to shut down the project in the late 1970s without funding any analysis or a final report. Only a fraction of the data was digitized in the 1980s, and the lone published study on how the Winnipeg program affected work hours relied on a dataset that could not be replicated and pooled married fathers together with childless single men.
To reconstruct the Canadian experiment, the researchers traveled to the Archives of Canada in Winnipeg and hand-collected the missing paper records. They digitized data for households that had been excluded from public files, such as couples who separated during the study, as well as a supplemental sample of roughly 260 families added after the initial rollout. This archival recovery expanded the Winnipeg sample by 65 percent to 821 couples and unlocked a previously unexamined survey module detailing how husbands and wives divided housework, how happy they were, and how they spent their free time.
For the American experiments, the researchers re-analyzed the public datasets using modern econometric protocols. In the original New Jersey analysis, for example, early investigators had discarded roughly half of the 1,357 enrolled families by excluding couples who separated or missed interviews, and they ignored the first and last six months of the three-year trial. The Riddells restored these dropped observations and adjusted their statistical models across all the experiments to properly account for the specific income, racial, and program-length brackets used to sort families before they were randomly assigned to treatment or control groups.
Next, the authors conducted balancing tests to verify whether the treatment groups (who were offered the cash program) and the control groups (who were not) had similar employment and demographic profiles before the experiments began. They found that random assignment failed in both Gary and Seattle. Both cities suffered severe, sudden economic downturns in their dominant local industries, steel in Gary and aerospace in Seattle, right as participants were being enrolled on uneven schedules. Because treatment and control families entered those two studies at different points during massive local layoffs, the groups had major pre-existing differences in earnings and welfare receipt, leading the researchers to set Gary and Seattle aside and focus on the valid experiments in Denver (1,503 families), New Jersey, and Winnipeg.
A consistent drop in working hours
When the researchers analyzed the complete, properly balanced datasets for Denver, New Jersey, and Winnipeg, the supposed inconsistencies across the experiments vanished. Instead of finding negligible employment effects in Canada and New Jersey, the authors found substantial and statistically significant reductions in work across all three locations.
Married women showed the largest percentage shifts, though they started from a much lower baseline of paid employment in the 1970s than men. Being offered the negative income tax reduced the likelihood that a married woman was employed by 17 percent in Winnipeg, 22 percent in Denver, and 32 percent in New Jersey. Total hours worked by married women fell by 26 percent in Winnipeg (about 144 hours per year), 21 percent in Denver (101 hours per year), and 36 percent in New Jersey (86 hours per year).
For married men, the work reductions were smaller in percentage terms but consistent across borders. In both Winnipeg and Denver, men offered the guaranteed income reduced their probability of working by 5 to 6 percent, while their total work hours dropped by 8 percent in Winnipeg (123 hours per year) and 10 percent in Denver (173 hours per year). Men at the Pennsylvania site of the New Jersey experiment showed a nearly identical 9 percent drop in hours and a 6 percent drop in employment, leaving the New Jersey sites as the only places where male work effort did not noticeably decline.
Rethinking family breakups and life at home
Having established that parents worked less in the labor market, the researchers examined what happened inside the home, starting with the famous claim that the Seattle-Denver experiment broke up marriages. Because earlier researchers had always pooled the Seattle and Denver data together, no one had checked whether the marital breakup findings held up in each city individually.
When the Riddells separated the two cities, they discovered that the spike in marital separations was driven entirely by the contaminated Seattle experiment. In Denver, across 12 different ways of defining and measuring divorce, the cash program had no statistically significant effect on marital breakups, with estimates hovering near zero. The exact same zero-effect result appeared in the newly reconstructed Winnipeg data.
In fact, the newly digitized survey responses from Winnipeg indicated that the income guarantee improved home life. Two years into the Canadian experiment, both husbands and wives in the treatment group reported a roughly 6 percent increase in overall happiness compared to the control group, while marital satisfaction rose by 8 percent for men and 17 percent for women.
The cash support also coincided with a shift in how couples managed domestic chores and community life. Married women in the Winnipeg treatment group reported a 27 percent increase in satisfaction with their husband’s help around the house and a 27 percent increase in agreement over who did the housework, alongside a 20 percent increase in social activities as measured by church attendance. Looking at baseline patterns in the data, the authors also found that couples who agreed on housework and felt satisfied in their marriages were much less likely to separate over the three-year period, suggesting that financial relief might actually support marital stability over the long run by easing household tensions.
What the 1970s can and cannot tell us today
While the study clarifies the historical record on both the labor costs and domestic benefits of a negative income tax, the authors caution against mechanically applying 1970s work reduction numbers to modern policy proposals. Family economics and work habits have changed substantially over the past half-century. In the early 1970s, men in these studies worked four to seven times as many paid hours as their wives, whereas married women today participate in the workforce at much higher rates and have become far less sensitive to changes in wages and household income when deciding how much to work.
However, the researchers point out that lower-income women working fewer hours remain more responsive to financial incentives than higher earners, meaning families eligible for basic income support today might still reduce their work activity. At the same time, the modern expansion of subsidized childcare has altered how parents with young children balance paid jobs and family duties, leading the authors to argue that new negative income tax trials are needed to test how these programs function in contemporary labor markets.
