AI Analysis: Canada Mortgage and Housing Corporation, housing starts, under construction and completions in selected census metropolitan areas, monthly

Category: housing

Executive Summary

This Statistics Canada dataset (Table 34100154) tracks over 50 years of monthly housing activity across 37 Canadian census metropolitan areas, covering nearly 250,000 records from January 1972 to February 2026 across three pipeline stages: starts, under construction, and completions. The data reveals a heavily right-skewed distribution — with a mean of 916 units but a median of just 68 — driven by large metropolitan markets that dominate cumulative totals. All three housing metrics are exceptionally tightly correlated (r > 0.94), confirming that starts, construction activity, and completions move in near-lockstep as sequential stages of Canada's housing supply pipeline.

Key Findings

  • The dataset contains approximately 250,000 monthly records spanning January 1972 to February 2026, organized across 37 census metropolitan areas and 555 unique data vectors representing combinations of geography, estimate type, and unit type.
  • Housing values are heavily right-skewed: the mean is 916 units versus a median of just 68, with individual records ranging from 0 to a maximum of 165,167 units in a single entry.
  • Units under construction carry the highest average at 2,230 units — nearly nine times the averages for housing starts (260) and completions (258) — reflecting the longer duration of the construction phase.
  • All three housing pipeline metrics are exceptionally correlated, with the strongest relationship between completions and starts (r = 0.963), followed by completions vs. under construction (r = 0.949) and starts vs. under construction (r = 0.943).
  • The middle 50% of all monthly values fall within a relatively tight band of 10 to 301 units (IQR = 291), despite the full range spanning 165,167 units, underscoring how a small number of large markets drive aggregate totals.
  • No statistically anomalous months were detected using the IQR method, as all monthly housing start totals across CMAs fell within the wide but normal bounds of 912 to 43,506 units — reflecting the natural variability absorbed over 54 years of data.
  • Large metropolitan areas such as Toronto, Montreal, and Vancouver are expected to dominate cumulative housing rankings, while smaller CMAs in the bottom 10 reflect lower population density and significantly reduced construction volumes over the same period.

This AI-generated analysis covers 8 analytical sections of Statistics Canada Table 34100154.

Source: Statistics Canada — Open Government Licence Canada