Housing record/Canada · 1948 – 2025/Layer 1 — verifiable fact · one projection, marked/Built 2026-08-13

Homes, started.

Canada started 259,028 homes in 2025 — 6.2 for every thousand people living here. In 1973 it started 12.0.

That gap is 52 years old and it has never closed. But the story is not the one the number invites: 2025 was a better year per person than any year of the 2010s, and the worst year on this chart is 1995, not a recent one. What has happened is not a collapse. It is a plateau — 5.1 to 6.3 homes per thousand people, every decade since 1990, through five governments.

Homes started per 1,000 people

Every year from 1948 to 2025. Housing starts across all areas of Canada, divided by the average of that year’s four quarterly population estimates.

peak12.01973 trough3.81995 20256.2-48% vs peak years781948–2025

The axis starts at zero. The marked point is 1973. The dashed segment and open point are the 2026 projection — its arithmetic is under “The year underway” below, where you can change it.

By decade
Decade	Per 1,000 people	Per new resident	Years
1940s	6.91	0.25	2
1950s	7.36	0.28	10
1960s	7.92	0.45	10
1970s	10.24	0.77	10
1980s	7.00	0.60	10
1990s	5.13	0.48	10
2000s	6.27	0.63	10
2010s	5.65	0.52	10
2020s	6.30	0.51	6

The 1940s and 2020s are partial decades — 2 years and 6 years — and are shown with their counts rather than left to look like the others.

The 1970s were the anomaly, not the baseline. Canada averaged 10.2 homes per thousand people through that decade and has never come close since. Some of that was policy and some of it was arithmetic: the baby boom reached household-forming age all at once, and a country whose 25-year-olds outnumber its 45-year-olds builds more houses than one where the reverse is true. Treating 1973 as the target Canada has failed to hit is a way of blaming the present for not being a demographic accident.

The other denominator

Per-person is the fair way to compare 1948 to 2025. It is not the way to ask whether Canada is keeping up, because it divides this year’s construction by everyone who already has somewhere to live. This chart divides it by the people who arrived.

Homes started per net new resident

Housing starts divided by the year’s increase in population. The dashed line at 0.42 is one home per 2.4 people, the average Canadian household in the 2021 census — roughly the rate at which construction houses population growth and does nothing else.

20250.54homes per newcomer 20240.213rd lowest of 78 all-time low0.171949 break-even0.421 ÷ 2.4

The axis starts at zero. The dashed line is break-even at 0.42. The 2026 projection appears here only when the projected population grows — a ratio with a negative denominator is not drawn.

The 2025 recovery is the denominator. On this measure 2023 and 2024 were the two lowest years since 1952 — 0.23 and 0.21 homes per new resident, about half of break-even. 2025 came in at 0.54, which reads as a 137% improvement in two years. It is not one. Starts went from 240,267 to 259,028 over the same period — +8%. Population growth went from 1,063,106 to 483,757 — -54%. Canada did not build its way back to break-even. It admitted fewer people. Whether that is the right policy is not this page’s question; that it is what happened is arithmetic, and a chart of this ratio without this paragraph beside it would be a lie told with true numbers.

Break-even is a floor and not a target. 0.42 houses population growth and nothing else. It assumes no one already here forms a new household, that nothing is demolished, and that no one currently living somewhere too small ever leaves. It also assumes household size holds, and it has not: the census average fell from 2.9 people in 1981 to 2.4 in 2021, which means the same population needs more homes each decade simply to stand still. A country genuinely keeping up would sit above this line, not on it.

The year underway

Everything above this line is measurement. This section is not. It takes the 2 reported quarters of 2026 and projects the rest of the year from them — a judgment, not a fact, which is why its arithmetic is printed here in numbers you can change. A projection whose assumptions are hidden is an opinion wearing a number’s clothes. The tables on this page carry measured years only.

2026, projected from 2 of four quarters

Reported so far: 51,211 in Q1, 66,788 in Q2 — 117,999 starts. The population estimate stands at 41,417,056 as of Q2, down from 41,575,585 at the end of 2025. Those are facts. Everything below the line inside this box is arithmetic on two assumptions, both editable.

Over the last 10 full years, the first 2 quarters carried per cent of the year’s starts — the year-to-year range was 41.9 to 49.1. Hold that share and 2026 finishes at … starts — … per 1,000 people, drawn as the open point on the first chart.

The population estimate has moved persons per quarter so far this year. Extend the 2 missing quarters at that pace and the four-quarter average is … — … against 2025, which makes the projected homes per net new resident ….

The defaults are recomputed on every build: the share is the ten-year mean, the quarterly change is this year’s own reported pace. Edits live in your browser and are written nowhere.

The numbers

Every figure on this page is in this table, and every figure in this table is a division of two columns that are also in it. A rate whose inputs are not published is not checkable. The headings sort.

Year	Starts	Population	Per 1,000	Growth	Per new resident
2025	259,028	41,601,578	6.23	483,757	0.54
2024	245,367	41,117,821	5.97	1,181,462	0.21
2023	240,267	39,936,359	6.02	1,063,106	0.23
2022	261,849	38,873,253	6.74	648,421	0.40
2021	271,198	38,224,833	7.09	227,034	1.19
2020	217,880	37,997,798	5.73	442,584	0.49
2019	208,685	37,555,214	5.56	545,875	0.38
2018	212,843	37,009,339	5.75	514,999	0.41
2017	219,763	36,494,340	6.02	441,811	0.50
2016	197,915	36,052,529	5.49	376,062	0.53
2015	195,535	35,676,467	5.48	287,233	0.68
2014	189,329	35,389,234	5.35	356,996	0.53
2013	187,923	35,032,238	5.36	368,090	0.51
2012	214,827	34,664,148	6.20	365,618	0.59
2011	193,950	34,298,531	5.65	339,053	0.57
2010	189,930	33,959,478	5.59	377,300	0.50
2009	149,081	33,582,177	4.44	382,694	0.39
2008	211,056	33,199,483	6.36	350,829	0.60
2007	228,343	32,848,654	6.95	319,376	0.71
2006	227,395	32,529,278	6.99	326,372	0.70
2005	225,481	32,202,906	7.00	303,546	0.74
2004	233,431	31,899,360	7.32	297,520	0.78
2003	218,426	31,601,840	6.91	293,977	0.74
2002	205,034	31,307,864	6.55	336,572	0.61
2001	162,733	30,971,291	5.25	323,891	0.50
2000	151,653	30,647,400	4.95	280,349	0.54
1999	149,968	30,367,051	4.94	243,177	0.62
1998	137,439	30,123,875	4.56	256,303	0.54
1997	147,040	29,867,572	4.92	297,697	0.49
1996	124,713	29,569,875	4.22	306,868	0.41
1995	110,933	29,263,007	3.79	302,943	0.37
1994	154,057	28,960,064	5.32	308,602	0.50
1993	155,443	28,651,462	5.43	327,308	0.47
1992	168,271	28,324,154	5.94	337,043	0.50
1991	156,197	27,987,111	5.58	354,751	0.44
1990	181,630	27,632,360	6.57	417,458	0.44
1989	215,382	27,214,902	7.91	463,428	0.46
1988	222,562	26,751,474	8.32	353,604	0.63
1987	245,986	26,397,870	9.32	330,384	0.74
1986	199,785	26,067,487	7.66	254,287	0.79
1985	165,826	25,813,200	6.42	236,465	0.70
1984	134,900	25,576,735	5.27	240,784	0.56
1983	162,645	25,335,951	6.42	253,006	0.64
1982	125,860	25,082,945	5.02	298,391	0.42
1981	177,973	24,784,554	7.18	313,839	0.57
1980	158,601	24,470,715	6.48	300,551	0.53
1979	197,049	24,170,164	8.15	234,704	0.84
1978	227,667	23,935,460	9.51	241,525	0.94
1977	245,724	23,693,935	10.37	279,581	0.88
1976	273,203	23,414,354	11.67	311,344	0.88
1975	231,456	23,103,010	10.02	330,967	0.70
1974	222,123	22,772,043	9.75	318,301	0.70
1973	268,529	22,453,742	11.96	266,647	1.01
1972	249,914	22,187,095	11.26	439,776	0.57
1971	233,653	21,747,319	10.74	459,819	0.51
1970	190,528	21,287,500	8.95	293,250	0.65
1969	210,415	20,994,250	10.02	302,250	0.70
1968	196,878	20,692,000	9.51	328,250	0.60
1967	164,123	20,363,750	8.06	366,250	0.45
1966	134,474	19,997,500	6.72	364,000	0.37
1965	166,565	19,633,500	8.48	356,250	0.47
1964	165,658	19,277,250	8.59	358,250	0.46
1963	148,624	18,919,000	7.86	348,250	0.43
1962	130,095	18,570,750	7.01	346,250	0.38
1961	125,577	18,224,500	6.89	369,250	0.34
1960	108,858	17,855,250	6.10	387,750	0.28
1959	141,345	17,467,500	8.09	405,250	0.35
1958	164,632	17,062,250	9.65	482,750	0.34
1957	122,340	16,579,500	7.38	509,250	0.24
1956	127,311	16,070,250	7.92	389,000	0.33
1955	138,276	15,681,250	8.82	411,750	0.34
1954	113,527	15,269,500	7.43	436,500	0.26
1953	102,409	14,833,000	6.90	396,250	0.26
1952	83,246	14,436,750	5.77	431,750	0.19
1951	68,579	14,005,000	4.90	301,500	0.23
1950	92,531	13,703,500	6.75	348,500	0.27
1949	90,509	13,355,000	6.78	538,750	0.17
1948	90,194	12,816,250	7.04	271,500	0.33

Method

All areas, not “centres 10,000 and over”. The 10,000-and-over tables are the ones quoted in the news, because they are released monthly and fast. They also omit every home built in a small town or in the country. This page uses the quarterly all-areas table, which is slower and complete.

Unadjusted, not seasonally adjusted at annual rates. A SAAR figure answers “what pace is this quarter running at” and cannot be added into a year without saying something false four times over.

A start is not a home. It is an excavation and an intention. The number that would check it is completions, and it is not available: CMHC’s all-areas completions series returns null for 14 consecutive quarters, every one after 2022. Statistics Canada serves those as empty rather than as zero, which is correct and which this page’s own reader nearly got wrong — an early build of it read null as zero and reported that Canada finished no homes at all in 2023. Absence is not a quantity. It is disclosed here because a reader deciding how much to trust a starts chart deserves to know which cross-check was missing.

Partial years are projected, not drawn short and not dropped. A year with 2 quarters reported plotted beside years with four is a bar shorter than its neighbours, and every reader on earth reads that as a decline — so an earlier rule dropped the running year from this page entirely. The rule is now one step less blunt: the reported quarters are printed as facts under “The year underway”, the full year is projected from them by arithmetic the page shows and lets you change, everything derived from the projection is drawn dashed and open rather than solid, and the tables carry measured years only.

Population is the four-quarter mean. Starts accumulate across a year, so the denominator should describe the whole year rather than its last day.

Fetched live, not from a snapshot. Every other page on this site names one downloaded file and one hash and can stand behind it indefinitely. This one asks Statistics Canada at build time, so it can only tell you the day it asked, the tables it asked for, and the exact coordinates within them. Those are below. Statistics Canada renumbers table members between vintages; the generator reads each cube’s metadata back on every run and refuses to build if a coordinate no longer points at the member it was written for.

Sources

Table 34-10-0135 — Canada Mortgage and Housing Corporation, housing starts, under construction and completions, all areas, quarterly.
coordinate 1.1.1.1.0.0.0.0.0.0 · vector 730416 · through 2026-04-01
Table 17-10-0009 — Population estimates, quarterly.
coordinate 1.0.0.0.0.0.0.0.0.0 · vector 1 · through 2026-04-01
Table 98-10-0039 — average household size in private dwellings, 1981 to 2021 censuses.
1981: 2.9 · 1986: 2.8 · 1991: 2.7 · 1996: 2.6 · 2001: 2.6 · 2006: 2.5 · 2011: 2.5 · 2016: 2.5 · 2021: 2.4

Statistics Canada’s Web Data Service needs no key and no account. That is worth saying out loud: a source that requires a credential is a source that can be revoked, and a chart whose provenance can be revoked is a chart this site should not publish.