The Town with No Poverty: Forget on Manitoba's Mincome Experiment (2011)

Summary

Evelyn L. Forget (University of Manitoba) reopened the archival record of Mincome, Canada's only guaranteed-annual-income (GAI) field experiment, run jointly by the federal and Manitoba governments from 1974 to 1979. Mincome combined a negative-income-tax design in Winnipeg with a "saturation site" — the small town of Dauphin, Manitoba — where every resident was eligible for a guaranteed floor income (a 50% benefit-reduction rate against a roughly $3,800/year guarantee for a family of four). Because the whole town was covered rather than a scattered sample, Dauphin lets researchers study community-level effects individual-level NIT sites cannot. Forget's contribution, from the mid-2000s, was matching a rediscovered roster of Mincome participants against Manitoba's population-health administrative database (the "Repository") and Department of Education enrollment data, decades later, to measure effects the original 1970s-80s analysis never completed.

Source-version note (important): the file fetched and read in full is Forget's 12 May 2008 draft, "The Town with No Poverty: A History of the North American Guaranteed Annual Income Social Experiments" (revenudebase.info), not the final peer-reviewed article. The canonical published version is Evelyn L. Forget (2011), "The Town with No Poverty: The Health Effects of a Canadian Guaranteed Annual Income Field Experiment," Canadian Public Policy 37(3): 283-305 — a narrower health focus and, per secondary citation below, effect sizes not present in the 2008 draft. This page cites the 2008 draft directly for design/mechanism/qualitative findings; the publisher's own PDF (utppublishing.com, Cloudflare-gated) still returns a 403/bot challenge on direct fetch (checked again 2026-07-17). Partially resolved (2026-07-18, T2 verification): citations to the 2011 article's specific content are now independently corroborated via two channels, though still not from the publisher's PDF itself. First, the published abstract, mirrored on IDEAS/RePEc (a standard bibliographic database, distinct from the BC panel), states the study found "an 8.5 percent reduction in the hospitalization rate for participants relative to controls, particularly for accidents and injuries and mental health." Second, David A. Green's reanalysis — now traced to its published form, "A Reanalysis of 'The Town with No Poverty: The Health Effects of a Canadian Guaranteed Annual Income Field Experiment,'" Canadian Public Policy 48(4): 539–548 (December 2022) — quotes Forget (2011) directly with page citations: "hospitalization rates among Dauphin subjects fell by 8.5 percent relative to the comparison group," that about a third of Dauphin families qualified for MINCOME stipends and "many of those stipends would have been quite small" (p. 291), that the health data cover "almost every physician and hospital contact in the province" (p. 292), the new-hospital confound discussion — "[n]one of the health or census variables that we examined could explain the persistent gap between subjects and controls before 1974, but we note that there was a fairly new hospital, which may have led to some supply-induced demand" (p. 295) — and Forget's own caveat that "one must be careful in generalizing potential health system savings, particularly because we use hospitals differently today than we did in 1978" (p. 300), matching the wording and pagination already used on this page via the BC Expert Panel's citation. Because Green's paper is a critical reanalysis working from Forget's own dataset and text (not a casual secondary mention), its exact quotations are strong independent confirmation that the BC Panel's quoted wording and page numbers are accurate. These two independent channels corroborate each other and the figures already on this page; the marker is resolved on that basis, though a first-hand read of the publisher PDF is still preferable if the paywall/bot-gate ever clears.

The Core Argument / Findings

The 2008 draft, read in full, reports these quasi-experimental findings for Dauphin residents ("experimental subjects," matched 1:3 against propensity-score-matched controls in comparable small towns and rural municipalities in south/central Manitoba) relative to before-Mincome baselines and to controls:

• Fertility. Before Mincome, women under 25 in Dauphin were more likely than controls to have given birth; by the end of Mincome this had reversed to significantly less likely — attributed to young women staying in school rather than marrying early.
• Hospitalization (overall). Dauphin subjects were more hospitalized, and for longer stays, than controls before Mincome; "by the end of the period, the results were reversed."
• Mental-health and accident/injury hospitalizations. "The same patterns held" for both subcategories — these specifically drove the overall reversal, consistent with reduced financial stress lowering accident and mental-health-crisis rates.
• Fade-out. "By 1985, seven years after the money stopped flowing, there were no significant differences between Dauphin and the controls on any measure."
• Education. Dauphin's Grade 11-to-12 enrollment continuation rose relative to Winnipeg/non-Winnipeg comparators during the 1975-78 Mincome years, "falling back to prior levels after the experiment ended."
• No effect on physician visits, birthweight, or family formation/dissolution — not every claimed GAI benefit shows up in the data.

The 2011 published article is described by third-party citations (notably the British Columbia Expert Panel's 2020 report — see below) as quantifying the hospitalization effect precisely: hospitalizations fell about 8.5% in Dauphin relative to controls, driven by drops in accidents/injuries and mental-health diagnoses (quoted via the BC panel's citation of Forget 2011, p. 295) — consistent with, and a numerical sharpening of, the qualitative "reversal" in the 2008 draft.

Relation to the Georgist Case

Mincome is general-revenue-funded — a federal-provincial negative-income-tax pilot, not a rent-funded dividend — so it bears on the Georgist case only through the citizen's-dividend lane, not as direct land-rent evidence. A Georgist citizen's dividend is one specific funding mechanism (captured land/resource rent) for a broader policy class — the unconditional or near-unconditional cash guarantee — and Mincome is among the oldest, most-cited empirical tests of that class's social effects, independent of funding source. Where the wiki's direct rent-funded-dividend evidence comes from Alaska's oil-royalty PFD (Jones & Marinescu 2022) and Maricá, Brazil (Balakrishnan et al. 2024), Mincome supplies a health-outcomes channel neither of those cases tests. It is general basic-income evidence a rent-funded dividend could plausibly replicate on health and schooling — not evidence about rent capture itself.

Nuances and Limits

• Version uncertainty: this entry is built primarily on a 2008 pre-print; the widely quoted 8.5% figure is now corroborated via a second, independent primary-quoting source (Green's published reanalysis, below) but the publisher's own version-of-record PDF has still not been read directly.
• A serious methodological challenge exists, and is now formally published. The BC Expert Panel's 2020 report (see companion entry) cited then-unpublished work by economist David Green disputing the causal read; that work is now published as David A. Green (2022), "A Reanalysis of 'The Town with No Poverty': The Health Effects of a Canadian Guaranteed Annual Income Field Experiment," Canadian Public Policy 48(4): 539–548 — see Source 5 — with a published reply, Evelyn L. Forget (2022), "A Reanalysis of 'The Town with No Poverty': A Reply," Canadian Public Policy 48(4): 549-556 (checked 2026-07-17). Plotting the Dauphin-minus-controls hospitalization gap year by year, Green finds "the convergence ... happens almost entirely after the experiment has ended," and "there was already a trend toward convergence underway before Mincome" — a pre-existing downward trend before MINCOME began that continues for seven years after it ended — more consistent with a pre-existing trend (possibly a newly built Dauphin hospital clearing an admissions backlog, a confound Forget herself flags at p. 295) than a treatment effect timed to the payment years. Green argues Forget's regression specification — which lets MINCOME shift the trend but not the level of hospital use, and treats every year from 1974 onward (not just the 1975-78 payment years) as "treatment" — drives the negative result. Once Green restricts "treatment" to the actual payment years and allows a level effect, the sign flips: hospital use in Dauphin rises by roughly 5-12% relative to controls during MINCOME, the opposite of Forget's own conclusion, which he attributes to a plausible alternative channel (an income floor letting people take time off work to address health issues, rather than reduced financial stress) — a direct, published academic dispute the wiki should treat as live, not settled in Forget's favor. He is explicit that Forget's own reported non-effect on newborn birthweight — a health outcome not subject to discretionary timing — is left standing either way. Forget's published reply maintains her original reading of the data; this page has not independently verified the reply's specific rebuttals against its primary text (checked only via the Green paper and secondary summaries), so her side of the exchange should be read at the source before being cited further.
• Non-random saturation design. Dauphin was chosen, not randomly assigned; "control" towns are matched, not randomized — a quasi-experiment, weaker than an RCT.
• Labour-supply offset. Riddell & Riddell (2020), cited in the BC panel report, find married women with children in Mincome households cut mean paid work hours by roughly 40% (participation unchanged), shifting time toward home/child care and reporting higher life satisfaction — a real labour-supply response, though not the large work withdrawal critics predict.
• Small, homogeneous population and era. External validity from 1970s rural Manitoba to a large, diverse, urban national program is untested by this design.

Bears On

• Concept: Citizen's Dividend — the oldest North American evidence for unconditional/near-unconditional cash guarantees, a broader category the citizen's dividend specializes to rent funding.
• Benefit (indirect support, general basic-income evidence, not rent-specific): Rent dividends reduce poverty and inequality — not a rent-funded case and not in that page's direct evidence base, but it strengthens the underlying mechanism (income guarantees improve poverty-linked outcomes) a rent-funded dividend also relies on.
• Objection (response side): Universal transfers are an inefficient way to reduce poverty — Forget's on-record rebuttal to the BC panel (oversight costs, the "deserving/undeserving" distinction) is part of that page's response.

See Also

• Covering All the Basics: the BC Basic Income Panel's Final Report — cites and critiques this same Forget hospitalization finding
• A Distributional Analysis of a National Guaranteed Basic Income (PBO) — a costing companion on a different basic-income design
• Citizen's Dividend
• Rent dividends reduce poverty and inequality
• Maricá, Brazil: an oil-royalty-funded permanent dividend
• Canada

Sources

1. Evelyn L. Forget (draft, 12 May 2008), "The Town with No Poverty: A History of the North American Guaranteed Annual Income Social Experiments." PDF (revenudebase.info mirror) — fetched and read in full — used for program design, methodology, and the qualitative fertility, hospitalization, mental-health, accident/injury, fade-out, and enrollment findings.
2. Evelyn L. Forget (2011), "The Town with No Poverty: The Health Effects of a Canadian Guaranteed Annual Income Field Experiment," Canadian Public Policy, 37(3): 283-305. Publisher PDF — the canonical published version; this URL returned a 403/Cloudflare bot-gate on repeated attempts (last checked 2026-07-17) and was not directly fetched. Cited here for the 8.5% hospitalization figure and pagination (pp. 291, 292, 295, 300), confirmed via the published abstract (Source 4) and page-cited quotation in Source 5, not read first-hand.
3. David A. Green, Jonathan Rhys Kesselman & Lindsay M. Tedds (2020), Covering All the Basics: Reforms for a More Just Society, Final Report of the British Columbia Expert Panel on Basic Income, Part 5 §5.5. PDF — fetched and read; used for the verified 8.5% quote from Forget (2011, p. 295), the new-hospital confound, the Riddell & Riddell (2020) labour-supply finding, and the panel's original (pre-publication) citation of Green's critique. See also the companion research page.
4. Evelyn L. Forget (2011), abstract as indexed by IDEAS/RePEc, record — fetched 2026-07-17; used as an independent (non-BC-panel) confirmation of the published abstract's 8.5% hospitalization-reduction figure and the paper's headline claims on physician contacts, schooling, and fertility/family-dissolution non-effects.
5. David A. Green (2022), "A Reanalysis of 'The Town with No Poverty: The Health Effects of a Canadian Guaranteed Annual Income Field Experiment,'" Canadian Public Policy 48(4): 539–548, DOI 10.3138/cpp.2021-025 — the now-published version of the reanalysis the BC Panel cited as unpublished; a working-paper text (dated 5 March 2021, matching the published abstract) was read in full this session via UBC Vancouver School of Economics PDF (the publisher's final version at utppublishing.com returned a Cloudflare bot-gate and was not directly fetched) — used to independently confirm Forget (2011)'s pp. 291, 292, 295, and 300 wording, the pre-trend/timing re-analysis, and for the direct quotation of Green's contrary conclusion (hospital use in Dauphin rising, not falling, roughly 5–12% relative to controls, once treatment years and a level effect are specified as Green argues they should be).
6. Evelyn L. Forget (2022), "A Reanalysis of 'The Town with No Poverty': A Reply," Canadian Public Policy 48(4): 549-556, DOI 10.3138/cpp.2022-017 — existence, venue, and headline positions (Forget maintains her original conclusions) confirmed via publisher/IDEAS-RePEc listings and secondary summaries; the reply's full primary text was not fetched (utppublishing.com Cloudflare-gated, Project MUSE login-gated) and its specific rebuttal arguments are not independently verified here.