AI Analysis: Canada Mortgage and Housing Corporation, housing starts, under construction and completions in centres 10,000 and over, Canada, provinces, selected census metropolitan areas

Executive Summary

Canada's housing dataset spanning 77 years (1948–2025) reveals a strongly right-skewed construction landscape dominated by units under construction, which average ~40,271 units per month — roughly 7x higher than starts or completions — reflecting increasingly complex, long-duration projects. All three core metrics (starts, under construction, completions) are strongly correlated (r = 0.73–0.80), confirming a predictable construction pipeline, though the data captures dramatic cyclical swings including a 1970s boom, a 1990s bust, and a post-2020 backlog surge. The all-time monthly starts peak of 25,014 units occurred as recently as November 2021, underscoring the intensity of Canada's current housing demand pressures.

Key Findings

• The dataset contains 11,790 records across 936 months (1948–2025), with values ranging from 0 to a peak of 174,929 units and a mean of 12,749 — more than double the median of 5,928, confirming a strongly right-skewed distribution.
• Housing under construction dominates in scale with a mean of ~40,271 units and a maximum of 174,929 units recorded in October 1973, compared to starts and completions which average approximately 5,856 and 5,881 units respectively.
• All three housing metric pairs show strong positive correlations (r = 0.73–0.80), with completions vs. under construction being the strongest relationship (r = 0.802), reflecting a consistent and predictable construction pipeline.
• The all-time monthly housing starts peak of 25,014 units occurred in November 2021, while the historical low of 1,298 units was recorded in February 1948, illustrating the vast range of activity over the dataset's 77-year span.
• Three months were flagged as statistical anomalies (Z-score > 2.5) in the under-construction series, with the 1970s boom, 1990s prolonged slowdown, and post-2020 construction backlog surge identified as the three most structurally significant macro-patterns.
• Apartment and multi-unit categories consistently dominate the highest cumulative value segments, while semi-detached and specialized unit types appear among the lowest-ranking contributors across the full historical record.
• The standard deviation of 22,263 units — nearly twice the mean — reflects extreme variability across regions, time periods, and housing types, highlighting the uneven nature of Canadian housing construction activity.