Purpose-built rental housing and household formation among young adults in Canadian cities, 1991–2016

Abstract

In recent decades, several Canadian cities have experienced a significant restructuring of their housing demand as young adult aspirations and expectations of maintaining independent households have been confronted by soaring housing costs, falling vacancy rates, and a rental supply system that is characterized by an overreliance on rental condominiums and other provisional forms of ‘secondary’ rental housing. The study presented in this paper focuses on intra-metropolitan variation in household formation among young adults and attempts to explain it from a macro perspective, emphasizing the importance of a community’s ‘primary’ or purpose-built rental sector. Following a review of the international literature on the impacts of housing supply and availability on household formation, panel regression models are estimated to examine the effects of changes in the relative supply and real cost of purpose-built rental housing on changes in young people’s household formation in Canadian metropolitan areas between 1991 and 2016. The results indicate that a 1% increase in the relative supply of purpose-built rental housing leads to 0.24% and 0.14% increases in the rates of family and non-family household formation of young adults. These estimates are generated using instrumental variables, controlling for potential simultaneity between household formation and purpose-built rental supply.

Keywords:

• Household formation
• young adults
• purpose-built
• rental housing
• Canada
• panel analysis
• regression

View correction statement:

Correction

Disclaimers

The views and analysis presented in this article are solely those of the author and not those of the City of Toronto.

Data on purpose-built rental housing in the analysis is adapted from Canada Mortgage and Housing Corporation, Rental Market Survey, 1991-2016. This does not constitute an endorsement by Canada Mortgage and Housing Corporation of this article.

Acknowledgements

The author would like to thank Alan Walks, Larry Bourne, Rishab Mehan, and Lindsay Stephens for their valuable feedback and comments on earlier drafts of this paper. The author also thanks the anonymous reviewers for their helpful comments and suggestions.

Funding

This work was supported by the Social Sciences and Humanities Research Council of Canada (SSHRC).

Note on contributor

Keir Matthews-Hunter is a graduate of the Master of Science in Planning (2018) program at the University of Toronto and currently works as a housing planner in Toronto, Canada. His research interests include housing markets and housing policy analysis, history of housing and community planning, economic and intergenerational inequalities in housing, and the interactions between housing and other areas of social and economic policy.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 For example, the City of Toronto – Canada’s largest municipality – has a Residential Rental Property Demolition and Conversion Control by-law and Official Plan policies that protect rental properties with six or more units from being demolished or converted into condominiums or non-residential uses. For the most part, the City will not approve new development that would result in the loss of six or more rental housing units unless at least the same number, size, and type of rental housing units are replaced and maintained with rents similar to those in effect at the time a development application is made. Similar by-laws and policies to protect against rental housing demolition and/or conversion exist in other major Canadian municipalities, including Vancouver (Rental Housing Stock Official Development Plan), Montréal (Québec’s Act respecting the Régie du logement and Montréal’s By-law concerning the conversion of immovables to divided co-ownership), Mississauga (Rental Housing Protection By-law), and Hamilton (section 3.2.5 of the Urban Hamilton Official Plan).

2 Due to major boundary conflicts between former municipalities and their respective 2016 CMAs/CAs, this process resulted in the exclusion of one smaller CMA (Greater Sudbury) and three CAs (Kamloops, North Bay, Prince George). One additional CA (Chilliwack) was then also excluded because of missing rental market data for the year 2006. This yielded a total of 45 metropolitan areas for analysis.

3 There are two CMAs – Montréal and Saguenay – in which the primary rental universe, as published publicly by CMHC, experienced inordinate increases between 2012 and 2015 that were not the result of new construction (as can be confirmed by data on annual rental completions). A third CMA – Sherbrooke – experienced similarly inordinate increases in purpose-built rental housing units between 2005 and 2006. The author directly confirmed with data specialists at CMHC that such increases were the result of ‘universe reviews’, whereby the Corporation conducted a review of the universe in each CMA, discovered rental structures that had previously existed but had not been captured by earlier Rental Market Surveys, and began enumerating these newly discovered structures in each subsequent survey year. In order to overcome these discrepancies, the Corporation supplied the author with additional custom data on ‘permanent removals’ (units permanently removed from the rental universe, i.e. demolitions, condominium conversions) and units in ‘new structures’ (i.e. units added through new construction) for the purposes of producing estimates of the primary rental universe in each CMA while accounting for units added through ‘universe reviews’. These estimates were produced for each of the three CMAs by taking the rental universe in the most recent survey year (2016) and then adding the sum of annual permanent removals and subtracting the sum of units in ‘new structures’ for each of the previous survey years.

4 Separate models were tested using an independent variable for the total immigrant population, as it could be argued that levels of immigration across all age cohorts (and not only young adults) may affect headship rates. However, the results of the models using the entire immigrant population did not significantly diverge from those obtained using the variable representing young adults. For consistency, only the 25 to 34-year age cohort is used for the immigration variable in this study.

5 Some studies have also treated socioeconomic variables such as income and education as endogenous to household formation at the individual level (while others have not), although most studies have treated such variables as exogenous at the macro scale. For the purposes of this study, IVs were created only for the key purpose-built rental housing variable of interest and all control variables, including the average total individual income of young adults and percentage of young adults with a bachelor’s degree, serve as covariates in the 2SLS models.

6 The author thanks one of the anonymous reviewers for this helpful suggestion.

7 Diagnostic plots for the equation one (1) regression models suggest relatively constant error variance, although results from Breusch-Pagan (BP) tests confirmed the presence of heteroskedasticity. The null hypothesis of the BP test is homoskedasticity against the alternative hypothesis that heteroskedasticity is present, and was rejected at the 5% level for the family (p-value = 0.012) and non-family (p-value = 0.011) headship rate models, and at the 1% level for the overall headship rate model (p-value = 0.009). Accordingly, I report robust standard errors (which are robust to, and valid in the presence of, heteroskedasticity of unknown form) in my estimation output for all models. Results are consistent and no important conclusions are overturned when using either regular or robust standard errors.