Introduction
The Modernizing Housing Data Initiative—a collaboration between Housing, Infrastructure and Communities Canada; Canada Mortgage and Housing Corporation; and Statistics Canada—aims to generate timely and granular data that will inform housing policy and program development. Statistics Canada is expanding the Canadian Housing Statistics Program (CHSP) to produce annual data on dwellings at the municipal level. These data provide new information on housing stock, improving the understanding of Canada’s housing market.
Methodology
The derivation aims to produce a complete and accurate estimate of the number of private dwellings in Canada. It is the product of several reconciled administrative data sources: provincial and territorial property data from land registries and assessment rolls, and residential building data available in Statistics Canada’s Statistical Building Register (SBgR). The number of dwellings is estimated for each residential property covered by the CHSP.
The initial process involved data standardization and harmonization to consolidate any differences in concepts and data between the datasets in preparation for linkage. Multiple linkage methods were used, including deterministic geospatial location-based and probabilistic address-based linkages. Subsequently, a post-match resolution process was implemented to ensure that only high-quality property and building unit links were retained.
A final estimate was produced by reconciling inputs from the various sources and represents the total number of private residential dwellings at the property level.
Quality evaluation
Consultations were held with each of the data providers to ensure reference points were harmonized, with an emphasis on harmonizing the time period associated with dwelling attributes across the jurisdictions.
The content of the input data, including property characteristics, was then compared across vintages to ensure coherency over time.
Subsequently, steps were taken to consolidate and standardize variables originating from the various data sources to achieve the best match between records. Vintages of the administrative data sources were chosen to best reflect the target vintage date of the dwelling stock.
Linkage results were reviewed to ensure that the methods used were largely correct and appropriate.
The following is a selection of the data evaluations that were performed.
| Province | Area | 2023 | 2024 |
|---|---|---|---|
| percentage | |||
| Newfoundland and Labrador | Overall | 87 | 87 |
| CMA/CA | 97 | 97 | |
| Outside CMA/CA | 75 | 76 | |
| Nova Scotia | Overall | 98 | 97 |
| CMA/CA | 99 | 98 | |
| Outside CMA/CA | 96 | 95 | |
| New Brunswick | Overall | 92 | 92 |
| CMA/CA | 95 | 95 | |
| Outside CMA/CA | 88 | 88 | |
| Ontario | Overall | 98 | 97 |
| CMA/CA | 99 | 99 | |
| Outside CMA/CA | 88 | 88 | |
| Manitoba | Overall | 87 | 88 |
| CMA/CA | 97 | 97 | |
| Outside CMA/CA | 65 | 69 | |
| British Columbia | Overall | 97 | 96 |
| CMA/CA | 98 | 98 | |
| Outside CMA/CA | 86 | 85 | |
| Note: CMA/CA refers to census metropolitan area or census agglomeration. | |||
Table 1 presents linkage rates between residential properties from the CHSP and building units from the SBgR. The results show that the linkage rates are generally high and mostly stable across the two reference years. Rates are higher in census metropolitan areas (CMAs) and census agglomerations (CAs), compared with outside CMAs and CAs. These differences result from the higher presence of non-civic addresses, incomplete civic addresses and misaligned input data across the data sources in areas outside CMAs and CAs. This is demonstrated when examining the change in the linkage rate in rural Manitoba, where an improvement in the quality of civic addresses was the main contributor to greater linkage success. Additionally, differences in coverage of the input data, especially for newly constructed dwellings, also explain a portion of unlinked records.
For linked records, the alignment between dwelling count information from property assessment authorities and the SBgR is consistent and stable across time, as presented in Table 2. The results show that the alignment rate was slightly higher in 2024 compared with 2023, because of coverage improvements for residential building units among SBgR. Conversely, the alignment rate in British Columbia was slightly lower than that in other provinces, mainly because of challenges capturing secondary suites across urban communities.
For situations where linked records have differing dwelling counts across the two data sources, the higher dwelling count is taken to account primarily for the presence of secondary suites. When there is no link to the SBgR, the dwelling is derived from the property assessment source.
| Province | 2023 | 2024 |
|---|---|---|
| percentage | ||
| Newfoundland and Labrador | 92 | 92 |
| Nova Scotia | 93 | 94 |
| New Brunswick | 94 | 95 |
| Ontario | 94 | 95 |
| Manitoba | 97 | 98 |
| British Columbia | 87 | 89 |
Limitations
The CHSP database reflects the current content of the external data provider’s registry of residential properties as of the date of extraction, which varies by province and territory. In Newfoundland and Labrador, coverage is limited to a portion of the province. As a result, the total number of residential dwellings is not available at the provincial level.