How to check us
29 sources sit behind this site. Each one is named here with what it covers, when it was pulled, which committed file holds it, and the thing that would make a careless reading of it wrong.
Six rules
These are load-bearing. Each one costs something, and each one is why a figure on this site is worth more than the same figure without it.
Source, publisher, period, retrieval date, file name, and the caveat that makes a careless reading wrong. A figure that cannot carry one does not go on the page.
Point-in-time counts undercount by design. The Toronto CMA is not Toronto. Estimates are revised backwards. None of that lives in fine print here, because none of it is optional.
If a number here is wrong, that is a bug. The fix is a published correction with a date, not a quiet change to the page.
Participation and public data only. No candidate preference, no party data, no endorsement. The ballot section counts whether a race exists and whether anyone published anything; it never counts who they are.
Every figure is in the HTML before any script runs. Turn JavaScript off and the charts are still finished, the tables still complete, the receipts still open.
Where a file has no date, it says undated. Where a county has no rental geography of its own, the proxy is named. Where two neighbourhoods lack a comparable year, they sit outside the trend chart rather than inside it with an invented value.
The registry
Every card on this site names one of these. The caveat is the part worth reading twice.
Elections Canada
What would make a careless reading wrong: Advance zones, not polling divisions, are the stable grain. A by-election can renumber polling divisions; zones survive. Merged and void polls carry notes, and receiver polls read slightly high.
Elections Canada
What would make a careless reading wrong: 2021 used the 2013 representation order and 2025 uses the 2023 order. Neighbourhood is the comparable unit; polling-division numbers are not comparable across the two years.
uniteTOlove, derived working files
What would make a careless reading wrong: A project-derived planning proxy, not an Elections Canada measure or a housing census. Street addresses are rows labelled parcel classes L/LE in the working extract; the upstream address-point query and its retrieval metadata are not registered here. Implied unit doors divide electors by 1.9, and workload divides those implied doors by 25 doors per pair-hour. Both are explicit assumptions, not observed household size or canvassing speed. The ratio cannot identify building type, lobby access, resident motivation, or cause.
City of Toronto / Statistics Canada
What would make a careless reading wrong: Neighbourhood boundaries are the City’s 158-neighbourhood system and do not line up with federal electoral geography. Comparisons across the two are indicative, not exact.
City of Toronto and Ontario Service Managers
What would make a careless reading wrong: Point-in-time counts undercount by design. They find people in shelters, in known outdoor locations, and in some institutions on one night; they miss people couch-surfing, doubled up, or staying out of sight. Count dates differ by region, so the totals are not a single national snapshot.
Canada Mortgage and Housing Corporation
What would make a careless reading wrong: Purpose-built rental only: no condos, no basement apartments, no rooms. The Toronto CMA figure spans Peel, York, Halton and part of Durham, so it is not a Toronto number and not a suburb number either. Several county rows are proxied from the dominant urban centre and carry a quality flag.
Ministry of Municipal Affairs and Housing (Ontario)
What would make a careless reading wrong: Own-source revenue is property tax, user fees, fares, water billing, permits and investment income. The remainder is transfers from the province and Ottawa plus other revenue. Single-tier, lower-tier and upper-tier municipalities raise money under different rules, so a raw per-capita ranking mixes three kinds of government.
Statistics Canada
What would make a careless reading wrong: Annual estimates, not census counts. Estimates are revised backwards as new census data lands, so a figure for 2019 today is not the figure that was published in 2019.
uniteTOlove, from council minutes
What would make a careless reading wrong: Hand-assembled from published council resolutions. It is a sample of what councils asked senior governments for, not a census of every resolution; absence from this table is not evidence a council stayed quiet.
uniteTOlove, from municipal clerks
What would make a careless reading wrong: Provisional. Nominations close 21 August 2026, so registrant counts move until then and zero-candidate races may fill. Platform scoring reads what a candidate published on their own site; a candidate campaigning entirely at doors and on paper reads here as having published nothing, which is a statement about the public record, not about the campaign.
uniteTOlove
What would make a careless reading wrong: Built one verified row at a time against each organization’s own page. It is a working map, not a complete one, and the categories are ours rather than any official taxonomy.
Toronto Police Service and Ottawa Police Service
What would make a careless reading wrong: Reporting thresholds and categories differ by service and by year, and at least one year is published as zero where the real answer is "not released". Treat this table as a record of what was disclosed, not of what happened.
Toronto Public Health / Shelter, Support and Housing Administration
What would make a careless reading wrong: Includes residents of City-funded emergency shelters and hotel programs who died while active in the system or shortly after discharge. Excludes unhoused people dying outside the shelter system unless verified by TPH.
Toronto Public Health
What would make a careless reading wrong: Mean observed age among shelter-system decedents. It is not life expectancy: comparing this mean with Ontario life expectancy describes two different statistics and cannot estimate years of life lost or the causal effect of homelessness.
City of Toronto SSHA
What would make a careless reading wrong: Reflects base capacity and operational models across 25 municipal wards. Motel and temporary family flex rooms vary nightly based on emergency activations and funding limits.
uniteTOlove / Toronto Auditor General & Budget Committee
What would make a careless reading wrong: Every figure cross-referenced with Toronto Auditor General reports (2022, 2025), City of Toronto budget notes, and Street Needs Assessments. Unit economics (e.g. $359/night warming centre costs) reflect audited actuals.
Toronto Public Health / SSHA
What would make a careless reading wrong: Quarterly counts of suspected non-fatal overdose events responded to by shelter staff or EMS at named facilities. Any count below five is suppressed to protect the people behind it and arrives as the string “< 5”, and a further set of site-quarters carry no value at all; 56.5% of the file is therefore not a number. Every total drawn from it is a floor, not a count, and the site publishes the interval rather than a midpoint. Sites also open, close and rename between quarters, so a site absent from a year may have closed rather than reported nothing.
uniteTOlove candidate monitoring
What would make a careless reading wrong: Platform word counts and stances extracted across 14 municipal issue domains. A count of zero indicates no published policy text on the candidate website, reflecting the public digital record rather than in-person campaigning.
uniteTOlove Digital Commons Census
What would make a careless reading wrong: Subscriber counts and activity signals verified directly from public platform registries. Private or restricted communities are flagged with accessible status notes.
uniteTOlove
What would make a careless reading wrong: Every link is checked against the provider’s own page rather than an aggregator. Hours and eligibility still change without notice; the guides name the three families that are drafted but not yet published rather than hiding them.
Daily Bread Food Bank / North York Harvest / PROOF U of T
What would make a careless reading wrong: Audited food bank visits (4,124,313 in 2025), employed household ratio (49%), and PROOF national food insecurity prevalence (25.5%). Reflects income inadequacy rather than food supply.
TCHC / City of Toronto Executive Committee
What would make a careless reading wrong: Capital repair backlog ($2.6B) across 58,000 units and average turnaround timelines for vacant social housing units.
uniteTOlove research library
What would make a careless reading wrong: Every document anybody on this project logged, including the ones that turned out to be useless: 418 of the 1,302 entries support no claim at all. The `category` column was written by hand across several ingestion passes and has drifted into 285 near-duplicate values, so category counts are indicative and the domain rollup is the reliable cut. `extraction_quality` is unrecorded for 1,285 rows and should not be read as a quality judgement.
uniteTOlove research library
What would make a careless reading wrong: One row per sentence pinned to a primary source, carrying the verification path that stood it up. Thirteen rows are marked UNVERIFIED-REMOVED: they were written, checked, and could not be confirmed. They stay in the file because a corpus that deletes its failures cannot be audited. Every claim here is about homelessness policy specifically and does not generalise to other domains.
uniteTOlove research library
What would make a careless reading wrong: The 174 documents the pinned claims actually cite. Roughly a third are published by the governments whose performance the claims describe, which is not a flaw — primary records are the point — but it is the reason the jurisdiction and publisher splits are printed rather than summarised.
uniteTOlove research programmes
What would make a careless reading wrong: Open questions, not findings: nothing in this file has been answered, and a question appearing here is evidence that somebody thought to ask it, not that it matters. Entries carry filing prefixes such as “[EXT]” which are stripped before the opening word is read. Programmes maintain their banks at different rates, so a large programme is an active one, not necessarily a more uncertain one.
uniteTOlove
What would make a careless reading wrong: File stems carry a place suffix — `housing-supply-affordability` holds the backgrounder, card and day-one sheet while `housing-supply-affordability_toronto` holds the brief — so the 296 stems are merged on the part before the first underscore into 149 subjects. Counting the stems instead would report that no subject is ever complete, which is false. Two snapshot stems are pre-merge backups and fold into their base for the same reason.
uniteTOlove research library
What would make a careless reading wrong: These seven CSVs index the project’s own research artifacts. A row proves that a brief, figure, precedent, claim, or uncertainty note was recorded; it does not independently validate the claim inside that artifact. Extraction is deterministic, but the source prose was written by multiple researchers and remains uneven.
City of Toronto / SSHA
What would make a careless reading wrong: The file carries its own warning on every row: historical baseline only, verify current status before public use. Sites have opened, closed and changed operator since, and the pandemic hotel programme is not in it at all. It is used here as a denominator of record for the period it covers, never as a description of the shelter system today.
What the build refuses to ship
A dataset that silently changes shape is the most expensive kind of bug on a site whose whole promise is that the figures are checkable. So the figures below are asserted in code: if one of them moves, the build fails instead of the page rendering a quiet lie.
| Assertion | Expected | Current | Holds |
|---|---|---|---|
| Beaches—East York advance zones | 15 | 15 | yes |
| Canvass coverage rows, both ridings | 28 | 28 | yes |
| Ontario FIR filers with usable revenue | > 350 | 402 | yes |
| Toronto total revenue, 2024 | $18,201,182,088 | $18,201,182,088 | yes |
| Ontario census divisions | 49 | 49 | yes |
| Years of population estimates | 25 | 25 | yes |
| Toronto average rent | $1,916 | $1,916 | yes |
| Toronto 2024 point-in-time count | 15,400 | 15,400 | yes |
| October 2025 count, three settings | 12,196 | 12,196 | yes |
| Shelter mortality data years | 20 | 20 | yes |
| Toronto shelter facilities | > 50 | 59 | yes |
| Weighted mean age among 2020–2024 shelter-system decedents | 49.9 yrs | 49.9 yrs | yes |
| Road safety reporting years | 4 | 4 | yes |
| Municipal policy issue domains | 14 | 14 | yes |
| Digital commons civic subreddits | > 20 | 271 | yes |
| 2026 races | 2,289 | 2,289 | yes |
| 2026 candidates | 4,751 | 4,751 | yes |
| Acclamation-track races | 956 | 956 | yes |
| Contested races | 917 | 917 | yes |
| Verified organizations | 2,863 | 2,863 | yes |
| Council resolutions | 81 | 81 | yes |
| Overdose file site-quarters | 1,963 | 1,963 | yes |
| Site-quarters carrying a number | 853 | 853 | yes |
| Site-quarters withheld as “< 5” | 716 | 716 | yes |
| Overdose total the file supports | 5,320 – 7,468 | 5,320 – 7,468 | yes |
| Documents in the resource registry | 1,302 | 1,302 | yes |
| Sources a claim actually cites | 174 | 174 | yes |
| Claims across both claim files | 968 | 968 | yes |
| Claims removed for failing verification | 17 | 17 | yes |
| Open questions in the harvest | 1,234 | 1,234 | yes |
| “How” to “why”, opening word | 81 : 1 | 81 : 1 | yes |
| Subjects on the publishing shelf | 149 | 149 | yes |
| Subjects carrying all four documents | 144 | 144 | yes |
| Races with nobody registered | 316 | 316 | yes |
| Races publishing no registrant count | 16 | 16 | yes |
| Press-kit figures that disagree with the code | 0 | 0 | yes |
| Rows differing between the races file and its export | 0 | 0 | yes |
All 37 hold as of this build. Two of them caught real errors while this site was being built: a population file with no province row, and a miscount of Ontario census divisions. Two more were caught by this column, which did not exist until the table grew past thirty rows: the mortality file had gained a year. A later source audit caught a more important interpretation error: mean age among observed shelter-system decedents had been subtracted from Ontario life expectancy and mislabeled as a life-expectancy gap. That comparison has been removed; the build now asserts the observed 2020–2024 weighted mean age directly.
If a number here is wrong
The correction gets published with a date and a note about what changed. The original figure stays visible in the record. Nothing is quietly edited.
Every figure on this site names the file it came from and the day that file was pulled. That is enough to check it against the publisher directly, which is the point. If your check disagrees with ours, ours is the one that needs to change.
Nothing on this site asks for money, an email address, or an account. There is no newsletter gate, no donate button, and no analytics that follows you off the page.
