AI Analysis: Canada Mortgage and Housing Corporation, housing starts, by type of dwelling and market type in centres 10,000 and over, Canada, provinces, census metropolitan areas and large census agglomerations

Executive Summary

Statistics Canada's Table 34100148 tracks monthly housing starts across Canadian urban centres (10,000+ population) from June 1988 to February 2026, encompassing 4,225 records across five dwelling types and five market types. The data is heavily right-skewed — a median of just 54 units versus a mean of 897 units — reflecting the outsized influence of large census metropolitan areas on national totals. Homeowner markets dominate construction volume, while rental and condo starts show the strongest co-movement, and outlier detection flags recognizable economic events such as the 2008–2009 financial crisis and post-pandemic construction surges.

Key Findings

• The dataset spans nearly 38 years of monthly data (June 1988 – February 2026), containing 4,225 records with housing starts ranging from 0 to 13,416 units and a mean of ~897 units.
• The distribution is strongly right-skewed, with a median of only 54 units versus a mean of 897 units (a 17x gap) and a standard deviation of 1,945, driven by a small number of large metropolitan markets.
• Homeowner markets are the dominant segment, averaging 2,892 units per record (median 717), far exceeding Rental (avg 578, median 86) and Condo (avg 953, median 554) market types.
• The strongest correlation in the dataset exists between Rental and Condo housing starts (r=0.727), suggesting these two segments share common economic drivers, while Homeowner starts are the most independent (max r=0.280 with Condo).
• Co-op and Other market types have medians of 0, meaning the majority of observations for these categories record no housing starts, indicating they are niche or regionally concentrated segments.
• Outlier detection using a ±2 standard deviation threshold successfully flagged historically significant periods of abnormal construction activity, including the 2008–2009 financial crisis downturn and post-pandemic construction surges.
• Total units records the highest average starts at 2,242 units and the widest range (0–13,416), as expected given it aggregates all dwelling subtypes, while single-detached units show the largest variability with a standard deviation of 2,596 despite a median of just 8 units.