Whose benchmark? The disagreement over how much housing Canada needs
Two institutions size the housing shortfall differently, construction proceeds at a measured pace against both paths, and part of the record has simply stopped updating.
Executive summary
The published record shows residential construction proceeding at a measured, steady pace. Two institutions responsible for estimating how much new housing Canada actually needs do not agree with each other on the size of the shortfall, and one part of the record -- how many homes are actually being finished, not merely started -- has stopped updating in the government's own machine-readable data.
People added per housing start: all-time low (-0.7 in 2026 Q2)
The latest value of people added per housing start (-0.7, 2026 Q2) is the lowest in the series history since 2007 (76 quarters of data). A record marks where the series sits in its own history; it carries no causal claim.
Technical details
Population growth per housing start: -0.7 (was 2.5 a year earlier)
In the four quarters to 2026 Q2, Canada's population declined by 187,499 while housing starts averaged 262,070 units (annualized), so the ratio is negative (-0.7; it was 2.5 a year earlier). New demand pressure from population growth has reversed while supply continues to be added. This says nothing about pre-existing shortage levels, and the inference ignores household-formation shifts, completions lags, and within-country migration.
Technical details
CMHC's 2025-06 housing benchmark uses a different measure from starts
Canada started 255,061 homes in the 4 quarters to 2026 Q2 (actual, all areas). CMHC's 2025-06 modelled benchmark — restore housing affordability to 2019 (pre-pandemic) levels by 2035, on a rolling 10-year horizon — implies ~478,000 net completions per year over 2025-2035. The comparison crosses unit bases: the benchmark is completions-denominated, but official all-areas WDS completions end in 2022 (latest annual figure: 219,942 in 2022). Direct CMHC metropolitan completion counts are now held, but they are gross monthly units on narrower geography, not this national net-additions basis. Current starts are shown as context, not compliance. The benchmark is a model with stated assumptions, not an observation; the rival benchmark differs by 3.8× on definitions alone (see the housing baseline conflict).
Technical details
What the record shows
Construction activity is measured and current. Its composition -- the mix of what gets built, not simply how much -- is tracked every quarter; the record does not currently show a standing shift in that mix.
Two institutions, two paths
CMHC and the Parliamentary Budget Officer each publish an independent estimate of how much additional housing Canada needs. Neither restates the other, and Countability does not merge, average, or pick between them -- both paths are tracked against the same construction record, in their own units, on their own terms.
Canada's Housing Supply Shortages: Estimating what is needed to solve Canada's housing affordability crisis by 2030 (with 2023 update)
Restore housing affordability to early-2000s levels by 2030 annual rate This vintage was expressed as a cumulative additional-units figure, not an annual rate.; cumulative 3,500,000.
Canada's Housing Supply Shortages: Moving to a New Framework
Restore housing affordability to 2019 (pre-pandemic) levels by 2035, on a rolling 10-year horizon annual rate 478,000; cumulative 2,600,000.
Household Formation and the Housing Stock: Estimating the Housing Gap in 2035 (RP-2526-007-S)
Return the total vacancy rate to its 2000-2019 historical average (6.4%) by 2035, while absorbing suppressed household formation (714,000 households projected by 2035) annual rate 290,000; cumulative 690,000.
Actual construction against both paths
The same construction pace reads very differently depending on which institution's path it is measured against.
CMHC's 2025-06 housing benchmark uses a different measure from starts
Canada started 255,061 homes in the 4 quarters to 2026 Q2 (actual, all areas). CMHC's 2025-06 modelled benchmark — restore housing affordability to 2019 (pre-pandemic) levels by 2035, on a rolling 10-year horizon — implies ~478,000 net completions per year over 2025-2035. The comparison crosses unit bases: the benchmark is completions-denominated, but official all-areas WDS completions end in 2022 (latest annual figure: 219,942 in 2022). Direct CMHC metropolitan completion counts are now held, but they are gross monthly units on narrower geography, not this national net-additions basis. Current starts are shown as context, not compliance. The benchmark is a model with stated assumptions, not an observation; the rival benchmark differs by 3.8× on definitions alone (see the housing baseline conflict).
Technical details
PBO's 2025-08 housing benchmark uses a different measure from starts
Canada started 255,061 homes in the 4 quarters to 2026 Q2 (actual, all areas). PBO's 2025-08 modelled benchmark — return the total vacancy rate to its 2000-2019 historical average (6.4%) by 2035, while absorbing suppressed household formation (714,000 households projected by 2035) — implies ~290,000 net completions per year over 2025-2035. The comparison crosses unit bases: the benchmark is completions-denominated, but official all-areas WDS completions end in 2022 (latest annual figure: 219,942 in 2022). Direct CMHC metropolitan completion counts are now held, but they are gross monthly units on narrower geography, not this national net-additions basis. Current starts are shown as context, not compliance. The benchmark is a model with stated assumptions, not an observation; the rival benchmark differs by 3.8× on definitions alone (see the housing baseline conflict).
Technical details
Where CMHC and the PBO differ
The two institutions are not disputing a fact -- they are answering different questions about what "enough housing" means, and reaching different numbers as a result.
How many additional homes does Canada need by 2035?
Units needed to restore housing affordability to 2019 (pre-pandemic) levels by 2035; implies ~478,000 net completions per year (range 430,000-480,000) → source
Units needed to return the total vacancy rate to its 2000-2019 average (6.4%) by 2035 and absorb suppressed household formation; implies ~290,000 completions per year → source
Why they differ: Different target definitions, not different data: CMHC models the supply needed to restore a chosen affordability level; PBO models the supply needed to normalize vacancy rates and absorb suppressed households. PBO's report states that CMHC's target would imply a ~13% total vacancy rate, which PBO characterizes as abnormally high.
The demand side
Housing supply does not move in isolation. The record also shows how population pressure and rent have moved alongside construction over recent quarters -- co-movement, stated as observed, never as cause.
People added per housing start: all-time low (-0.7 in 2026 Q2)
The latest value of people added per housing start (-0.7, 2026 Q2) is the lowest in the series history since 2007 (76 quarters of data). A record marks where the series sits in its own history; it carries no causal claim.
Technical details
Population growth per housing start: -0.7 (was 2.5 a year earlier)
In the four quarters to 2026 Q2, Canada's population declined by 187,499 while housing starts averaged 262,070 units (annualized), so the ratio is negative (-0.7; it was 2.5 a year earlier). New demand pressure from population growth has reversed while supply continues to be added. This says nothing about pre-existing shortage levels, and the inference ignores household-formation shifts, completions lags, and within-country migration.
Technical details
People added per housing start and Rent CPI, % y/y moved together over the 12 quarters to 2026-04 (r=+0.89). Correlation is not causation.
Rent CPI, % y/y and Housing starts, actual, trailing 4 quarters moved in opposite directions over the 12 quarters to 2026-04 (r=-0.71). Correlation is not causation.
What the record cannot say
Three limits are named rather than smoothed over. Each is part of the finding, not an omission from it.
How many homes is Canada actually completing right now?
What is missing: Narrowed September 8, 2026: a documented CMHC Excel route now supplies reproducible June and July 2026 metropolitan starts, under-construction stocks and gross completions, with July intended-market detail. The three held Statistics Canada completion mirrors still end in 2022. The broader CMHC provincial totals include smaller-centre values despite conflicting survey-coverage notes, so this reading uses the explicitly labelled metropolitan tables. A current, verified all-areas completion series and national net-additions measure remain outside the comparable evidence held here.
Consequence: The machine-readable-route branch of the original closing criterion is met for metropolitan counts. This is a narrowing, not restoration of all-areas coverage or a national benchmark-compliance measure. Gross monthly completions on metropolitan geography cannot be divided by either institution’s national annual net-additions benchmark. Cancellation and occupancy outcomes remain separate gaps.
How many authorized or started homes are cancelled before completion?
What is missing: No continuously updated machine-readable cancellation count or rate is held on a verified comparable basis for this reading. The qualified CMHC construction tables report starts, stocks and completions; stock adjustments for cessation or unit-count changes do not isolate cancellations. The separately unfetched Market Absorption Survey concerns absorption after completion, not pre-completion cancellation; it is not activated by this source admission.
Consequence: The Housing Delivery module now observes separate construction stages. It still cannot measure a cancellation rate, a linked permit-to-completion conversion rate or occupancy. Missing cancellations are never imputed as zero; broader source discovery remains open under the registered criterion.
How fast are new households forming right now?
What is missing: No official sub-annual household-formation series exists for Canada. The official household count (StatsCan 17-10-0159) is annual (July 1 reference, published each November, based on the 2021 Census adjusted for undercoverage plus the T1 Family File and postcensal estimates).
Consequence: Quarterly or monthly comparisons of housing supply against 'household formation' necessarily use modelled formation (e.g., PBO's projections) or population-based proxies. Countability publishes population-per-completion (observed) quarterly and completions-per-formation (observed) annually, and labels any sub-annual formation figure as modelled.
Is Canada building family-sized homes?
What is missing: Bedroom-count detail for new supply exists only in CMHC's Housing Market Information Portal, which has no machine-readable public API. The StatsCan mirror (34-10-0148) provides dwelling type (single/semi/row/apartment) but not bedroom counts.
Consequence: Countability uses dwelling-type composition (ground-oriented vs apartment) as a stated proxy for family-sized supply.
Source trail & method
- Housing shortages in Canada (2022/2023 framework, superseded modelled benchmark) — CMHC
- Canada's Housing Supply Shortages: Moving to a New Framework (2025, modelled benchmark) — CMHC
- Household Formation and the Housing Stock: Estimating the Housing Gap in 2035 (modelled benchmark) — PBO
- Population estimates, quarterly — Statistics Canada
- Housing starts, under construction and completions, all areas, quarterly — Statistics Canada / CMHC
- Housing starts, all areas, SAAR — Statistics Canada / CMHC
Every claim above states its own tier, confidence, and source. Correlation is reported as observed co-movement, never as cause. Where a required linkage is not public, the gap is declared rather than estimated across. This Deep Dive's editorial framing (section headings and connective text) introduces no numeral — every figure belongs to a member object shown in full on this site.
Where this sits in the record
- Big Question: Can governments still solve housing shortages? →
- Monitor: Canadian Capacity Monitor →
This Deep Dive and its underlying question are published in full. See Work with Countability for a commissioned one.
Work with Countability