AI Analysis: Canada Mortgage and Housing Corporation, housing starts in all centres 10,000 and over, Canada, provinces, and census metropolitan areas, seasonally adjusted at annual rates, quarterly

Executive Summary

Canada's housing starts have followed dramatic cyclical swings over 35+ years, ranging from a low of ~83,641 units in Q1 1996 to a peak of ~260,316 units in Q1 2021, before cooling to ~235,394 units in early 2026. The dataset covers 47 geographic areas and 6 unit types across 145 quarters, revealing a long-term structural shift toward multi-unit housing driven by urbanization and densification. Despite high volatility, no statistical outliers were detected using IQR-based methods, indicating that even extreme values fall within historically normal bounds.

Key Findings

• Total housing starts peaked at 260,316 units in Q1 2021 — the highest level in the 36-year dataset — driven by pandemic-era demand and historically low interest rates, before declining to ~235,394 units by early 2026.
• The most severe housing slump occurred in Q1 1996 at 83,641 units, with additional cyclical dips visible around the early 1990s recession, the 2008–09 financial crisis, and the COVID-19 shock in 2020.
• Multiples (apartments, semi-detached, and row units) have grown as a share of total starts over time, with Multiples recording the highest mean (17,630 units) and widest spread (standard deviation: 28,390) of any unit type.
• All unit types exhibit heavy right-skewed distributions, with medians far below means — for example, Apartment units show a median of just 560 versus a mean of 12,540 — indicating that a small number of large urban centres dominate national totals.
• No statistical outliers were detected across 145 quarterly observations using the IQR method (bounds: 31,940–307,750 units), confirming that even the most extreme housing cycles remained within historically normal statistical ranges.
• Single-detached units show the most consistent tracking profile with a mean of 11,140 and median of 1,540, while Semi-detached and Row units have the smallest ranges and lowest averages, reflecting their niche role in the Canadian housing market.
• The dataset spans 47 geographic areas and 118 unique time series (VECTOR identifiers), with large provinces and major CMAs such as Ontario, Quebec, and British Columbia consistently ranking among the top regions for average housing construction activity.