What does Toronto build, and who ends up holding it?
My thesis studied the demand side of Toronto housing: what happens to rents when homes leak out of the long-term market. This study builds the supply side of the same market, from the permit desk to the intended market to the final owner, and its last chart lands back where the thesis began.
This page comes in two parts: the story in plain language first, then technical notes with the permit universe, dedupe rules, lag tables, and sources. Companion to study 05, the planning queue, which covers the stage before this one.
A quarter-million homes crossed the counter
Every building in Toronto legally begins as a building permit. The city publishes every permit back to 2000: nearly a million of them once revisions are collapsed, of which 48,214 create homes.
How to read the first chart: each bar is one year's worth of new dwelling units in issued permits, 234,529 homes across twenty-five years. The number that matters most is hiding in the fine print: seven in ten of those homes sit in just 560 permits, the ones for buildings of a hundred units or more. Toronto's housing supply is a small number of very large decisions.
A quarter-million homes crossed the permit desk in 25 years, most of them in a few hundred towers
dwelling units in issued City of Toronto building permits, by issue year – latest revision per permit – 2001 to 2025
City of Toronto building permits open data · issue dates show administrative batching; annual totals absorb it
And the permit clock runs slowest exactly there. The second chart counts the median wait from permit application to issuance, in ticks of thirty days. A house permit takes about three months. A mid-size building takes seven. The towers that carry most of the housing wait a median of 791 days, more than two years, at the permit desk alone, and that clock starts only after the rezonings and site plans of study 05 are done, and stops before roughly two and a half years of construction begin.
The bigger the building, the longer the permit
Median days from application to permit issuance, 2001–2025 – one tick = 30 days – the slowest lane (100+ units, 557 permits) carries 69.9% of all permitted units
City of Toronto building permits open data · latest revision per permit
A generation without rental
When a building starts construction, CMHC records what it is intended to be: homeowner housing, condominium, or purpose-built rental. Thirty-six years of that record tell Toronto's supply story in one chart.
How to read this chart: the dark line is purpose-built rental, the housing built specifically to be rented out long-term. Through the 2000s it nearly vanished: 3.4 cents of every construction dollar's worth of starts, while the condo era took over. The revival after 2017 is real, with more rental units started in 2020–2024 (31,502) than in the entire 1990s (24,291). But it arrives after a thirty-year gap that the rental market of my other studies has been living inside. At the end of 2025, 90,568 units of all kinds were under construction across the region.
Toronto spent a generation not building rental; the revival is recent
purpose-built rental · condo · homeowner – annual housing starts by intended market, Toronto CMA – co-op and other categories, near zero throughout, are omitted
CMHC via Statistics Canada table 34-10-0148 · intended market at start is not final use
The newer the condo, the more of it investors own
The funnel's dominant output for two decades has been the condominium. Federal property records show who holds them.
How to read this chart: each row is a construction vintage of Toronto's 328,165 condo apartments, and the bar is the share owned as an investment property, meaning a property that is not its owner's primary residence. The staircase is unmistakable: 19% of condos built in the 1970s, 48% of those built 2011–2015, and 58% of condos built since 2016. For contrast, 10.9% of single-detached houses are investment properties.
The newer the condo, the more of it investors own
Share of Toronto condo apartments that are investment properties, by period built – 2022 – dashed line = all condos (40.5%) – single-detached houses, for contrast: 10.9%
Statistics Canada, Canadian Housing Statistics Program, table 46-10-0070 · City of Toronto · vintages before 1971 (4,610 condos) omitted for small counts
This is where the study lands back on my thesis and study 04. Investor-held condos are the stock that supplies Toronto's secondary rental market, and the same stock the short-term-rental sector draws from. The funnel takes the better part of a decade to produce a tower, the tower is majority investor-owned when it arrives, and how those owners deploy the units, long-term lease, 28-night loophole, or short stay, is exactly the margin my thesis measured from the rent side.
The honest fine print
Three geographies live on this page and are never mixed on one chart: permits cover the City of Toronto, CMHC starts cover the wider metropolitan area, and the ownership records cover the City. Intended market at the start of construction is a plan rather than a guarantee of final use. "Investment property" is the federal classification for any dwelling that is not its owner's primary residence, which bundles long-term rentals, second homes, and short-term rentals together; this study cannot split them. And the ownership data's latest year is 2022, so the newest vintages will keep evolving.
Pre-flight catches on this dataset: permit issue dates arrive in administrative batches (198 permits share one date in 2007), so nothing on this page uses daily granularity; and a million raw rows collapse to 988,069 permits once revisions are deduplicated, with the loss counted in the technical notes.
Technical terms, translated
- Purpose-built rental
- A building constructed specifically to be rented out long-term, as opposed to condos that individual owners may rent out or not.
- Intended market
- What a building is planned to be when construction starts: homeowner, condo, or rental. Plans can change; the record captures the intent.
- Investment property
- A dwelling that is not its owner's primary residence. A downtown condo rented to a tenant, a second home, and a short-term rental all count.
- Permit revision
- Big projects amend their permits, sometimes repeatedly. Counting every revision would count the same building twice, so this study keeps only the latest version of each permit.
Technical notes
The permit universe, the dedupe arithmetic, the lag tables, and the sources.
Data and universe
City permits: three open-data files (cleared 2000–2016, cleared 2017+,
active), 1,085,803 rows collapsing to 988,069 permits at the latest revision
per permit number. Analysis universe: permits with
DWELLING_UNITS_CREATED > 0 and an issue date in
2001–2025 (n = 48,214; 234,529 units; 202 units recorded lost on the
same permits). Scale classes: 1–2 units (n = 46,137), 3–99
(1,517), 100+ (560); the waiting-time table below drops the four permits
without a usable application date, so its class counts run 46,136 / 1,517 /
557. (An earlier version of this page used the smaller counts for both
purposes; the SQL rebuild described below surfaced the distinction.)
CMHC: table 34-10-0148 (starts by dwelling type and
intended market, Toronto CMA, monthly from 1988, annualized here) and
34-10-0154 (starts, under construction, completions; December values used
for the under-construction stock). CHSP: table 46-10-0070, Toronto City,
reference years 2020–2022.
Estimates
| Scale | Median application-to-issue | Median issue-to-completion (issued ≤ 2018) | 75th pct |
|---|---|---|---|
| 1–2 units | 83 days | 2.16 yrs | 4.45 yrs |
| 3–99 units | 210 days | 2.34 yrs | 5.17 yrs |
| 100+ units | 791 days | 2.52 yrs | 4.61 yrs |
Completion lags condition on cleared permits from cohorts issued to 2018 to
limit right-censoring; completion here means the permit file is closed, which
trails occupancy. Intended-market decade shares (Toronto CMA starts): rental
11.3% in the 1990s, 3.4% in the 2000s, 14.9% in 2020–2024; condo 58.2%
of 2015–2024 starts. CHSP 2022, Toronto City: 40.5% of 328,165 condo
apartments are investment properties (4.7% business- or government-owned);
10.9% of 303,060 single-detached houses. Every number traces to
output/results.json and the committed CSVs.
Threats to validity
This is a descriptive pipeline study by design: no stage is modeled causally, and no cross-geography ratio is computed (City permits are never divided by CMA starts). The permit clock measures the city's issuance process, but application quality and applicant response times are inside that clock and cannot be separated with this file. The investor-by-vintage staircase confounds vintage with unit mix: newer condos are smaller on average, and smaller units attract investors, so the staircase describes the stock honestly without isolating why.
The same study in SQL
As applied relational-database practice at scale, all 1,085,803 permit revisions also live in a SQLite database, where the deduplication that pandas does imperatively is stated once, declaratively, as a view:
CREATE VIEW permits_current AS
SELECT * FROM (
SELECT pr.*,
ROW_NUMBER() OVER (
PARTITION BY permit_num
ORDER BY rev DESC, row_id DESC
) AS rn
FROM permit_revisions pr
)
WHERE rn = 1;
-- 1,085,803 revision rows read as 988,069 current permits
Every query downstream selects from the view as if the table had one row
per permit. Seven queries reproduce this page's published numbers,
including the window-function medians behind the permit clock and a
percentile computed with the same linear interpolation pandas uses, and an
automated checker verifies all 32 values against the committed
results.json and permits_annual.csv: 32 of 32
match. The rebuild also caught the 557-versus-560 distinction now
documented in the data note above, which is what reproduction exercises
are for.
Reproducibility
Three scripts (get_data.py, run_analysis.py,
make_charts.py) on pandas. Development-application context comes
from study 05's committed outputs. The SQL rebuild
adds build_db.py, queries.sql, and
run_checks.py in the project's sql/ folder.
References
- City of Toronto. (2026). Building permits: active and cleared permits [data sets]. City of Toronto Open Data Portal. https://open.toronto.ca/dataset/building-permits-cleared-permits/ (retrieved August 2026). Open Government Licence, Toronto.
- Canada Mortgage and Housing Corporation. (2026). Housing starts, by type of dwelling unit and market type [table 34-10-0148]; housing starts, under construction and completions [table 34-10-0154]. Statistics Canada. https://www150.statcan.gc.ca/t1/tbl1/en/tv.action?pid=3410014801 (retrieved August 2026).
- Statistics Canada. (2024). Canadian Housing Statistics Program: residential properties by investment status [table 46-10-0070]. https://www150.statcan.gc.ca/t1/tbl1/en/tv.action?pid=4610007001 (reference year 2022; retrieved August 2026).