Ending Homelessness in Toronto · Chapter 7

What It Costs

A chapter of Ending Homelessness in Toronto — this project’s receipt-backed book on ending homelessness for less than the crisis costs now. Where a figure could not be confirmed against a primary source, the text says so on its face. How the checking works: how we verify · corrections.

Every chapter before this one has argued that Toronto already knows what works. This chapter asks the question a reader is right to ask before believing any of it: what would it actually cost, and how sure can anyone be of the number?

What Toronto pays today to manage the crisis

Toronto's direct spending on shelter and homelessness services was $897.96 million gross in 2025 and a budgeted $786.1 million gross for 2026 (City of Toronto, 2025 and 2026 Budget Notes, Toronto Shelter and Support Services). That figure does not capture the full cost of leaving people unhoused. A 2024 study linking Toronto's homeless population to provincial health administrative data found people experiencing homelessness incur roughly $12,209 a year in public healthcare costs each, against $1,769 for housed comparison groups — about six times higher even after adjusting for underlying health conditions — producing an estimated $69.8 million to $99.7 million a year in excess Toronto healthcare spending attributable to homelessness alone, before counting justice-system, shelter, or emergency-response costs (Lucie Richard, "The Cost of Inaction: Healthcare Expenses Associated with Homelessness in Toronto," Homeless Hub, October 17, 2024). Nationally, the Homeless Hub estimates homelessness costs the Canadian economy roughly $10 billion a year across healthcare and emergency services. By any reasonable accounting, Toronto is spending in the range of $900 million to $1 billion or more a year, all sources combined, to manage a crisis that keeps growing rather than to end it — and, as earlier chapters document, still turns people away most nights.

The per-person math the City has already done

Toronto's own Auditor General has already quantified how badly this money is being spent, relative to the alternative: emergency shelter costs more than three times as much per person as supportive housing, and seven to ten times as much as a rental subsidy (Auditor General of Toronto, "Part 1 of the Audit of Emergency Shelters," May 2022). At the unit level: a shelter-hotel room costs the City roughly $253 a night; a base shelter of any type costs roughly $136 a night — the figure behind this book's $49,640-a-year headline shelter-bed cost — against roughly $126 a night specifically for the newer subset of purpose-built, City-owned shelters. Of these, the $136 base-shelter figure is the one that appears in an Auditor General report — the February 2025 winter-services audit, Figure 3, where the AG's own note records it as TSSS-supplied and unaudited; the hotel and purpose-built figures come from the City's own budget materials and press reporting rather than from an Auditor General report. The 2022 audit's contribution is the cost-multiple finding above, not the nightly figures. Canada's At Home/Chez Soi national trial found the Housing First intervention cost $22,257 a year per participant receiving the highest-intensity clinical support and $14,177 a year per participant receiving moderate-intensity case management, falling to a net annualized cost — after subtracting averted shelter, hospital, and justice costs — of $6,311 and $7,868 respectively. Toronto's own site, in its high-needs cohort specifically, cost $21,089 per person per year and produced $31,747 in two-year service-cost reductions (Gozdzik et al., At Home/Chez Soi Project: Toronto Site Final Report, Mental Health Commission of Canada, 2014, p. 11). The most expensive thing Toronto currently does is exactly the thing it does most: pay for temporary shelter, indefinitely, for people who would cost the public purse less, and live better, permanently housed.

What ending it would cost: three scenarios, built bottom-up

No City of Toronto, provincial, or federal body has published an official estimate of the full cost of permanently housing everyone currently homeless in Toronto. This chapter builds one bottom-up, from the City's own reported and audited unit costs and the delivery-tool mix Chapter 5 recommends, because that number is the single most load-bearing claim in this book and deserves to be shown, not just asserted.

Three delivery tools, three unit costs. Rent supplements carry no capital cost at all, because they subsidize an existing private-market unit rather than acquiring or building one; their ongoing cost can be derived from the City's own confirmed cut to the Canada-Ontario Housing Benefit, which fell from $19.75 million in 2025 to $7.95 million in 2026, "eliminating support for roughly 530 households" — implying a benefit level of roughly $22,264 per household per year. Acquisition, through the City's Multi-Unit Residential Acquisition Program, has an audited program-wide average of roughly $150,000 per unit (the program has distributed $165 million for over 1,000 homes since 2021) against a City-set program cap of $200,000 per unit. Modular and purpose-built supportive housing is the most contested figure: the City's original Council-approved budget assumed $190,000 per unit, but the Auditor General's own 2023 audit of the program found actual and forecast costs $33 million, or 63%, over the adjusted budget, with only 159 of a planned 275 units completed at three of five planned sites — implying an audited actual cost of roughly $310,000 per unit, and an audit team that stated plainly it was "unable to conclude whether modular housing construction results in faster delivery and lower costs than traditional construction" (Auditor General of Toronto, City of Toronto's Modular Housing Initiative: The Need to Balance Fast Delivery with Stronger Management of Contracts and Costs, June 21, 2023). Ongoing support costs use the same national At Home/Chez Soi benchmarks as above: $14,177 a year for the acquisition population and $22,257 a year for the modular-build population.

Applied to Toronto's October 2025 point-in-time population of 12,196 people, across three plausible splits of that population among the three tools:

Scenario A is closest to Chapter 5's own stated sequencing, that rent supplements "should absorb the largest single share of the population." Scenario B is this book's recommended point estimate, splitting the population roughly along the line between chronic and non-chronic homelessness in the City's own 2024 count. Scenario C is not a prediction; it deliberately routes a much larger share of the population through the most expensive tool than the chronic-homelessness data actually supports, to test how high the honest range can go before it strains credulity.

The arithmetic behind Scenario B, shown in full: 6,098 people through rent supplements at no capital cost; 3,659 people through acquisition at $150,000 to $200,000 a unit, or $549 million to $732 million; and 2,439 people through modular build at $190,000 to $310,000 a unit, or $463 million to $755 million — a capital total of $1.01 billion to $1.49 billion, and an ongoing cost of roughly $242 million a year. The other two scenarios follow the same method against their own population splits.

Overrun sensitivity, named plainly. Applying the Auditor General's measured 63% modular-cost overrun to the build component alone moves each scenario's capital ceiling up by 24% to 43% relative to using the originally claimed rate throughout — the single largest source of range width in this model, larger in every scenario than the acquisition program's own cap-versus-average spread. The effect is largest in the build-heavy scenario, where the modular component is the biggest share of the bill.

Does the estimate survive against what Toronto already spends? Measured against $786.1 million a year in the 2026 budget — down from $898.0 million in 2025 — "less than three years of status-quo spending" means staying under roughly $2.36 billion to $2.69 billion, depending which budget year anchors the comparison. Every scenario modeled here clears that bar — including the worst case. The supplement-heavy scenario costs 0.8 to 1.3 status-quo years. The balanced, recommended scenario costs 1.1 to 1.9 status-quo years. And even the deliberately pessimistic, build-heavy scenario — priced at the Auditor General's audited overrun rate and the acquisition program's cap rather than its average — costs 1.6 to 2.85 status-quo years. That worst case is a thin margin, not a comfortable one, but it holds under three status-quo years in every single scenario this book modeled, including the one built specifically to be as unfavorable as the plan's own logic allows.

This range is markedly lower than an earlier estimate this book carried, of $2.4 billion to $2.7 billion, which did not survive its own arithmetic: it implicitly assumed the entire homeless population would be routed through an acquired or newly built unit, with none served through the no-capital rent-supplement path that Chapter 5 itself says should absorb the largest single share of the population. That assumption was never stated outright in earlier drafts of this argument, and once corrected, the honest range is both lower and better supported than the figure it replaces.

Ongoing Housing First support costs land in a tight $240 million to $255 million a year band across all three scenarios — well below the $786 million to $898 million the City spends today just to manage the crisis, and consistent with the return already demonstrated in Toronto's own At Home/Chez Soi trial (Chapter 5).

Two honest notes on what that band rests on. First, the band's tightness should not be mistaken for precision in its dominant input. The rent-supplement line — the largest ongoing-cost component in every scenario — is priced at $22,264 per household per year, a figure derived from a single sentence attributed to the City's own materials — an $11.8 million Canada-Ontario Housing Benefit cut said to eliminate "support for roughly 530 households" — whose household count this book's re-verification could not re-locate in the primary record (the $11.8-million cut itself is confirmed); it stands, flagged, as the model's weakest input. A ±10% swing on that rounded "roughly 530" moves the ongoing-cost line by roughly ±$13.6 million a year — real movement, though not enough to change which side of the status-quo-spending comparison any scenario lands on; the direction of this chapter's conclusion survives. Second, this model prices people as units. The 12,196-person point-in-time count is applied above as though every person needs a separate capital unit, when the underlying population includes families who share a unit — an error that runs conservative on the capital side (it overstates unit demand and therefore the capital bill, so the "under three status-quo years" finding is safe), while the same household-vs-person seam cuts the opposite way on the ongoing-cost side, where the $22,264 figure is a per-household benefit level applied above as a per-person cost.

This is an author's calculation, built transparently from the City's own published and reported figures, not an official City, provincial, or federal costing exercise, and it should be replaced with one before capital is committed at this scale. But the direction of the finding does not depend on getting the exact multiplier right: every actual cost comparison Toronto has produced — the Auditor General's three-to-tenfold finding, the At Home/Chez Soi return, the gap between a hotel room and a shelter bed — points the same way. The status quo is not the affordable option; it has simply never been priced against its alternative.

This project's homelessness research corpus, which turns the argument above into specific, government-by-government action lists (see Further Reading), holds itself to the same discipline at a much more granular level: every individual recommendation in it is sorted into one of four honest categories — reforms that cost nothing and save money, reforms that are fiscally neutral, reforms that are genuinely not yet costed because a specific input is still missing, and a small number of large anti-poverty investments that are real and worthwhile but are not homelessness-budget offsets and should never be summed in as though they were. That research states its own rule plainly: do not add the columns together into a single return-on-investment number. This chapter's scenario model follows the same instinct at the whole-of-book scale — a range, not a point estimate; an honest ceiling, not a marketing figure.