Publications, opinions, and speeches


Critical lessons from Ontario about how to set up a Basic Income experiment

Published on 25/08/2021

This presentation was recorded for the Basic Income Earth Network World Congress 2021 in Glasgow. In his presentation, Michael Mendelson summarizes his 2019 paper, Lessons from Ontario’s Basic Income Pilot.

In 2017 the Ontario Basic Income Pilot began making its first benefit payments. The Ontario Basic Income Pilot was the first full-scale Basic Income experiment in a wealthy nation since the 1970s. It featured several thousand participants, oversight by research committees, and a cost of hundreds of millions of dollars. The pilot was meant to “test whether a basic income would provide a more efficient way of delivering income support, strengthen the attachment to the labour force, and achieve savings in other areas, such as health care and housing supports.” Ontario’s Basic Income experiment was not just symbolic.

But having barely begun, the experiment came to a crashing halt. In 2018, Ontario elected a Conservative government, and the project was cancelled.

Ontario’s Basic Income experiment is now over. But there are still critical lessons to be learned about how to set up a Basic Income experiment. Any jurisdiction considering a Basic Income experiment can benefit from understanding limitations in Ontario’s Basic Income experiment’s design.

Watch the presentation

But before discussing its limitations, I need to tell you a little about the experiment’s parameters. Although the Ontario project was called a “pilot” this is incorrect. If a pilot is primarily a test of the administration of a new program, and an experiment is primarily a test of the outcomes of a new program, the Ontario project was not a pilot; it was a social experiment. Although Ontario called its program a “Basic Income,” more precisely, it was a Negative Income Tax, because it made payments to households inversely related to household income, rather than making payments to all individuals regardless of income.

The amount of income guarantee was relatively generous: 37.5 percent of median income. A family of four would have about $40,000 per year, untaxed. This was close to the official poverty line and far more generous than any other income security benefit in Canada.

In any Negative Income Tax, the amount of benefit paid is reduced by a percent of reportable income. The experiment’s reduction rate was 50 percent, so for example, with a benefit of $17,000 a recipient would need $34,000 income before the benefit was reduced to zero.

Household income reported through the income tax system was used to calculate the amount of benefit. Canada’s income tax system is retrospective; for example, taxable income for 2020 is reported in 2021. So, using the tax system meant that the amount paid depended upon last year’s income, not current year’s income. It also implied that the benefits would ordinarily be fixed for a whole year and be unresponsive to change during the year.

The Ontario government selected three communities to participate in the experiment: two small cities and one town, called Lindsay, with a population of about 20,000. In the two cities participants and controls were selected randomly: a Randomized Control Trial or RCT. In Lindsay the whole town was eligible in the enrollment period. In all three sites, including Lindsay, eligibility was limited to those who at the time of enrollment were entitled to at least some cash benefit.

Of the four North American negative income tax experiments in the 1970s, only the Canadian experiment – the Manitoba Basic Annual Income Experiment known as Mincome – had a “saturation site.” In a saturation site everyone is eligible for the program. This is very different than selecting geographically scattered participants in a classic RCT. A saturation site allows us to look at the effects of a Basic Income on everyone in a community – rich, poor, and in-between – as well as the community as a whole. A saturation site therefore more realistically simulates an actual Basic Income program.

For a limited time, Mincome functioned in the saturation site like a fully implemented Basic Income, albeit as a temporary program and without the taxes needed to pay for the program. It was therefore possible to assess effects of a Basic Income on a community, as opposed only to isolated households as in the RCTs.

In a seminal 2011 paper, Professor Evelyn Forget undertook a new analysis of Mincome’s impact. Her findings pointed to important effects on the population as a whole; mainly a significant reduction in hospitalization, especially for admissions related to mental health and to accidents and injuries. As well, a greater proportion of high school students continued on to grade 12.

Forget’s research sparked renewed Canadian interest in Basic Income and the possibility that prior RCTs had missed important, positive implications for communities. Her findings were one of the inspirations for the Ontario Basic Income Pilot and its decision to make the whole town of Lindsay eligible. You would therefore think Ontario would have been careful to make Lindsay a true saturation site.

And this brings me to the first of three critical problems in the design of the Ontario experiment.

To have been a true saturation site, every resident of Lindsay regardless of income would have had to be eligible for the basic income guarantee throughout the planned three years of the experiment. Instead, eligibility was limited to only those households which at the time of enrollment had income low enough to be entitled to a payment. This is not a saturation site. Saturation would have meant, for example, that a young adult could decide to quit work and go back to school to get a diploma or a degree. Or a Lindsay resident who had decided not to enroll in the first year could change her mind and sign up in year two. Or, perhaps, the guarantee of a minimum income could have created a greater sense of security and reduce economic anxiety throughout the whole community, as everyone would know that they had a real safety net available.

There are potential effects of a Basic Income implemented across a whole community which are simply not evident in an RCT. Some of these effects may be positive – such as on health services and crime – while others may be seen as negative – such as reduced labour force participation; but whether positive or negative, the implication is that critically important consequences of a Basic Income cannot be understood through an experiment that does not have a saturation site.

A second critical issue has to do with randomization.

You might think it is easy to give away free money, but it is not. The government’s initial enrollment process was to mail out invitations, but this failed miserably. According to Greg Mason, an academic assisting in the early design of the project: “Ontario started mailing invitations in June 2017 and by September, after mailing 37,000 invitations, it had managed to enrol barely 150 participants…”

Due to the continuing difficulty of recruiting participants, social agencies set up group meetings of households likely to be eligible. Those wishing to enroll could complete an application on the spot. This recruitment method meant that the Ontario experiment was no longer a randomized study.

A third critical issue has to do with the way income was tested.

As noted, the income tax system was used as an annual “income test” to calculate the amount of benefits a recipient would get. But this meant that the income used to assess the benefit was not current monthly income. Most income security systems providing for basic needs, such as food and shelter, are calculated according to as current income as possible and are designed to be responsive to changes in income. The received wisdom has been that low-income households need highly responsive programs because they cannot “average” their income over months or years; they need to pay the rent and buy groceries this month.

However, it is possible that the received wisdom may be incorrect or at least overstated. The Ontario experiment could have been a valuable test as to whether annual tax data could indeed be used as the basis for income testing.

The use of income tax is not a minor technicality. On the one hand, if it is possible to use retrospective tax data, the administration of a Basic Income could be greatly simplified, administrative expenses substantially reduced, and potential stigma could be all but eliminated. On the other hand, using retrospective tax data implies that benefit payments in any given month may be wildly out of sync with household needs in that month.

Finding out whether it would be possible to use the income tax system to “test” income could have been an important result from the OBIP. Instead, this unique feature was ignored in the research program while ad hoc work-arounds were used to adjust for out-of-date income.

There was another missed opportunity: Lindsay could have become a true population-wide saturation site by administering the basic income as an automatic tax-transfer program. Everyone in town could have been enrolled using tax data with an opt-out rather than an opt-in provision, so that take-up would have approached 100 percent. The result would have been a ground-breaking and truly unique experiment to understand potential community-wide effects of a Basic Income.

On the one hand, the three issues highlighted here – saturation site, enrollment and use of the tax system – compromised the integrity of the experiment; on the other hand, taken together these issues represented an opportunity to obtain evidence based on a more realistic simulation of a fully implemented Basic Income.

Given world-wide interest, almost any project with the label “Basic Income” can gather a wave of publicity far beyond the project’s potential significance. But if an experiment is to contribute meaningfully to evidence about the impact of Basic Income, the experiment needs to be designed to simulate as closely as possible the operation of a full-fledged Basic Income as it would be implemented for a whole jurisdiction. One of the implications of simulating a full-fledged Basic Income is that the experiment must include a true saturation site. Without a saturation site, we can never obtain information about the effects of an actual program on a community, rather than just the effects on individuals.

With a real saturation site in which the whole population is automatically enrolled, and a well-thought-out protocol for dealing with in-year income variation, the Ontario project could have been the world’s first population-based Basic Income experiment in a whole community. It was my intent in writing this paper on the now defunct Ontario experiment to encourage others setting up new Basic Income experiments to avoid Ontario’s errors and to take advantage of Ontario’s missed opportunities.


Evidence-based policy, Income security


Ontario’s Basic Income experiment is now over. But there are still critical lessons to be learned about how to set up a Basic Income experiment.