WORKING DRAFT — frozen research component (July 2026). This page is part of the Civil Society Research Library v2, imported whole as a frozen component: writing finished, human review never completed, evidence chain intact. 6 of 7 claim keys cited on this page are VERIFIED against primary sources in the library’s own ledger; 1 in other states (disputed/removed/unmatched) — shown, not hidden. Claim keys are shown as CSRL2-### and resolve in this library’s own claim ledger — not this site’s estate-wide claims register. About this collection.

Cost per bed night: a computed comparison for Homes First and Dixon Hall

Combining each operator's own audited financial data with the City's own primary shelter-occupancy Open Data yields a real, computed cost-per-bed-night trajectory for two of Toronto's largest shelter operators — a metric neither organization publishes directly, and one where an early calculation method required a real, disclosed precision correction before it could be trusted.

The finding

A note on method, stated plainly because it changed the numbers: an early pass at this calculation used calendar-year approximations for Dixon Hall's fiscal-year (April-to-March) spending against calendar-year occupancy data. A later, more precise pass computed true fiscal-year-aligned bed-night sums directly from daily occupancy records, and found the calendar-year approximation had produced meaningfully different — and, for FY2023 and FY2024 specifically, wrongly characterized as monotonically rising — figures. This page uses the corrected, fiscal-year-aligned numbers as authoritative.

For Homes First Society, cost per bed night is now computed for every year from 2017 through 2023, directly from primary Charity Intelligence retrieval: $104.53 in 2017, rising to $274.81 by 2023-24 — with the 2020-to-2021 transition marking a real, explained scope boundary between two internally consistent sub-series rather than one continuous line CSRL2-699. The single largest jump traces to the COVID-19 period specifically: occupancy fell from 97% to 82% while total spending rose 50%, mechanically explaining a 74% increase in the per-bed-night figure during the hotel-shelter expansion CSRL2-693.

For Dixon Hall, the true fiscal-year-aligned figure runs from $183.83 to a COVID-era peak of $313.23 at FY2021, across FY2018 through FY2024 CSRL2-705 — a different shape than Homes First's trajectory, peaking earlier and not simply rising monotonically once properly fiscal-year-aligned. An earlier, since-superseded calendar-year approximation had shown Dixon Hall's cost per bed night rising from $159.64 (2019) to $360.60 (2020), a 126% single-year increase mechanically explained by a steeper 34% bed-night decline against 49% program-spending growth — a real calculation, just since replaced by the more precise fiscal-year-aligned series above CSRL2-694.

A separate but related metric — bed nights by category, not cost — shows the two operators diverging in a genuinely different way. Homes First's overnight services moved substantially outside its base shelter category: by 2024, over two-thirds of its total overnight services occurred outside base shelter, following a real methodological correction (a naive continuation across a scope change in the underlying dataset would have falsely shown a near-tripling; isolating base shelter specifically produced a true comparable series: 152,180 to 271,717 base-shelter bed nights, 412,221 to 638,349 total) CSRL2-695. Dixon Hall shows the reverse pattern: base shelter bed nights rising (53,620 to 78,121) while its non-base share fell from 70.6% to 22.1% — a faster wind-down of its COVID-era hotel/respite footprint alongside core shelter growth, a genuinely different trajectory discovered only by computing both operators the same way CSRL2-696.

A forward-looking comparison point, from a fresh S1 discovery this pass: the City of Toronto's own 2026 announcement of its 10-year shelter strategy projects that new purpose-built shelters (approximately 80 people each) will achieve savings of up to $127 per bed per night compared to shelter hotels, representing $74 million in projected operating cost reductions over 20 years. This is a City-projected, forward-looking figure, not a computed historical actual like the Homes First/Dixon Hall series above, and is tagged VERIFY rather than VERIFIED pending a full Lane V pass [CL-90655].

Why it matters

Cost per bed night is a metric neither operator publishes directly — it had to be computed by combining each organization's own audited spending with the City's own occupancy Open Data, and the Dixon Hall correction shows why that computation needs real care: a plausible-looking calendar-year approximation produced a materially wrong trend characterization until fiscal-year alignment was applied properly. The two operators' genuinely different trajectories (Homes First's per-bed-night cost still elevated years after the COVID peak; Dixon Hall's peaking earlier and partly reverting) is itself a finding a single-operator view would miss. The City's own $127/bed/night projected savings figure for purpose-built shelters gives a forward-looking benchmark to weigh against these two operators' actual historical costs, though it should not be read as directly comparable without accounting for differences in service model, population served, and building type.

What this page does not cover

This page does not extend the corrected fiscal-year-aligned method to Homes First's own series beyond what CL-699 already provides for 2017-2023. It does not verify the City's own $127/bed/night savings projection independently — it is a City-published projection, not yet cross-checked against an independent analysis. It does not carry a "case against" either operator's cost structure, since the sources reviewed are computed primary-data comparisons and a City strategy announcement, not a documented opposing position on these specific figures.

Sources

Related pages

Jurisdiction: Toronto · Topics: cost-per-bed-night, shelter-operators, financial-transparency, per-diem · status: working draft · review: pending (the library’s own tags, kept visible) · published 2026-08-17 · corrections welcome.