AI Analysis: Canada Mortgage and Housing Corporation, newly completed and unoccupied housing in selected census metropolitan areas and census agglomerations of 50,000 and over

Executive Summary

Statistics Canada's Table 34100162 tracks newly completed and unoccupied housing units across 67 Canadian census metropolitan areas and agglomerations from January 1992 to February 2026, encompassing 186,438 records across 7 housing unit types. The data reveals a long-term decline in unoccupied newly completed housing, falling from an early-period average of ~54,574 units to a recent average of ~37,205 units, with a peak of 63,326 unoccupied units recorded in April 1995. The distribution is heavily right-skewed, with most region-type combinations reporting modest counts, while a small number of large metropolitan areas drive the most extreme values.

Key Findings

• The dataset spans 34+ years of monthly data (January 1992 – February 2026), containing 186,438 records across 67 geographic areas and 7 distinct housing unit types.
• Newly completed unoccupied housing peaked at 63,326 total units in April 1995, and has since trended downward to a recent 3-period average of approximately 37,205 units — a decline of roughly 32%.
• The distribution of unit counts is strongly right-skewed: the median is just 21 units while the mean is 284 units, with 75% of all records reporting 91 or fewer units and a maximum of 20,817 units in a single record.
• Approximately 15% of total-unit records (3,746 out of 24,918) were flagged as statistical outliers using the IQR method, with the upper bound threshold set at 558 units, concentrated primarily in Canada's largest metropolitan areas.
• The correlation heatmap across 26,634 geographic-date combinations shows that certain housing unit types — particularly multi-unit categories — tend to move together, suggesting shared regional market or policy drivers.
• The top 15 regions account for a disproportionately large share of cumulative unoccupied housing units over the full 1992–2026 period, while the bottom-ranked smaller census agglomerations report near-zero unoccupied completions.
• Zero-value records were identified in the dataset, potentially indicating reporting gaps or periods of full absorption in certain census areas, warranting caution in granular regional analyses.