With this geospatial analysis using census data, I was able to verify that the relationship between housing costs and number of dwellings is weak, and that the approach to addressing the housing crisis must go beyond simple additions of supply to include targeted and diverse approaches, including rental-only dwellings and subsidized housing. Using indicators of affordability and urban function derived from census data, areas of the city where there is a need for housing supply and where there is the opportunity to expand current housing stock were identified with a multi-criteria evaluation. Finally, the single and two-family zoned areas intersecting the aforementioned area of opportunity were identified to signal the next step that would follow from this kind of analysis: re-zoning.

Limitations

It is important to consider the appropriateness of census data for city-level analyses. The mandate of census data is to provide national and regional-scale indicators of change, and thus it may lack data that is pertinent to local issues. For example, in the housing crisis in Vancouver, a significant component of the challenge is homelessness, which is not accounted for in census data. While census data is capable of creating some useful indicators of housing need and urban capacity for increase of housing like income-to-housing cost ratios and rate of sustainable mode share, analyses like these in the future in cities like Vancouver and elsewhere should considering sourcing locally relevant data like incidence rates of homelessness. Further, temporal indicators that I originally intended to use for the multi-criteria evaluation (for example rate of change of unaffordability index) were ultimately taken out of the analysis due to the incidence of extreme outlying values in some years and not others, creating wide gaps of variations in data across DAs that inhibited the ability to generate smooth interpolation surfaces. To include these temporal indicators would have required more rigorous and careful polishing of the census data by, for example, setting extreme outlier values null.

Uncertainty in this analysis is present in two main areas: the quality of census data (especially from 2011), and the decision-rule uncertainty in the MCE. As for the latter, this is present in both the use of arbitrary weighted values for each criteria, and in the use of Kriging-based interpolation only. The equal weight sensitivity analysis MCE mitigated the uncertainty inherent to the arbitrary weights I assigned, but more thorough research into the relative importance of these criteria would have been most beneficial in reduction of uncertainty in the final results. Additional sensitivity analyses of the MCE using other interpolation methods would have also reduced uncertainty.

Next Steps and Recommendations

This analysis shows what is possible by using available social data to pinpoint spaces in the city that are both in need of action on housing to prevent displacement and promote well-being, and capable of absorbing additional population. By repeating such an analysis with more polished and more locally-relevant data, well-researched metrics for criteria of area selection, and additional measures to reduce uncertainty, it could be possible to make a convincing case for the re-zoning of specific low-density areas in Vancouver to develop social and rental housing as part of the implementation of the Housing Vancouver Strategy in the next 10 years.