Arguments about basic income often turn on predictions about behavior. Will people stop working? Will they spend wisely? Since the late 1960s researchers have tried to answer such questions with experiments, and North America ran the first large ones.
The negative income tax experiments
Between 1968 and 1982 the US government funded four negative income tax experiments: in New Jersey and Pennsylvania, in rural Iowa and North Carolina, in Gary, Indiana, and in Seattle and Denver. The New Jersey experiment, begun in 1968, was one of the first large randomized social experiments anywhere. Families were randomly assigned to different guarantee levels and phase out rates, or to a control group, so researchers could measure effects directly.
The results were fairly consistent. Primary earners, usually husbands, cut their hours only slightly, by a few percent, while second earners and teenagers cut theirs more. Some of the reduction came from longer spells between jobs, which may have led to better matches, and some teenagers stayed in school longer. Critics read the numbers as proof that guarantees discourage work; supporters called the effects small and partly beneficial.
The Seattle and Denver experiment, the largest, also reported a rise in marriage breakups among families receiving payments. The finding troubled supporters such as Senator Daniel Patrick Moynihan and was used against reform. A 1990 reanalysis by Glen Cain and Douglas Wissoker found the divorce result was largely a statistical artifact, but by then the political moment had passed.
Mincome in Dauphin
Canada ran its own trial, and its story shows how politics can bury evidence. From 1974 to 1979 the Mincome project in Manitoba offered a guaranteed income in Winnipeg and in the small town of Dauphin. In Dauphin every family was eligible, making it a "saturation site" where a whole community was covered. When funding ended, the data were boxed and stored without a final report. Decades later the economist Evelyn Forget dug into health records. Forget found that hospitalizations in Dauphin fell by about 8.5 percent compared with similar communities, with fewer admissions for accidents, injuries and mental health problems. More teenagers stayed in school to finish grade 12.
In 2017 Ontario launched a basic income pilot for about 4,000 low-income people in and around Hamilton, Thunder Bay and Lindsay, paying up to about 17,000 Canadian dollars a year to a single person, reduced by half of any earnings. Ontario's pilot was cancelled in 2018 by a newly elected provincial government before it could produce its planned results, leaving researchers with only surveys of participants.
Recent American trials
Over the past decade dozens of US cities have run guaranteed income pilots, usually paying a few hundred dollars a month to selected low-income residents. The best known began in Stockton, California, in 2019. The Stockton Economic Empowerment Demonstration gave 125 residents 500 dollars a month for two years. In Stockton's first year, full-time employment among recipients rose from 28 to 40 percent, faster than in the control group, which rose from 32 to 37 percent. Recipients also reported less anxiety and depression, and most of the money went to food, utilities and other basics, with less than one percent tracked to alcohol and tobacco. The sample was small and the pandemic arrived midway, so the results are suggestive rather than decisive.
Dividends from shared wealth
Two programs pay income to everyone in a community without having been designed as experiments. Alaska's yearly dividend is paid from the earnings of a fund built from the state's oil revenue, created by a 1976 amendment to the state constitution; the work course discusses its effect on jobs. In North Carolina, the Eastern Band of Cherokee Indians began sharing casino profits with every tribal member in the late 1990s. A long-running health study led by Jane Costello happened to be following children in the area. Costello found that Cherokee children whose families were lifted out of poverty by the payments had fewer behavioral problems, and later work found benefits that lasted into adulthood.
Worked example: reading a headline
A headline reads: "Guaranteed income makes people quit their jobs." The article reports a pilot in which recipients worked 1.5 hours a week less than the control group. A careful reader asks four questions. Who reduced their work: parents of young children, students, people caring for relatives, or people who simply preferred more rest? How long did the trial run, and did people know it would end? Was it a whole community, like Dauphin, or a scattered sample? And what else changed, such as health, schooling or job quality?
The New Jersey and Seattle results showed that an average can hide very different stories. A mother working less to care for an infant and a teenager finishing school instead of taking a low-wage job both count as a fall in work, yet many would call both a gain. A headline that reports only the average has told the reader the least interesting part of the result.