Notes

On average, owner-occupied dwellings accounted for about 40% of household assets, and home mortgages accounted for 62% of household liabilities for households and unincorporated businesses in 2009 (CANSIM table 378-0051).

Renters do not incur the full cost of utilizing a housing asset, because rental contracts are unable to specify all the contingencies that could account for the increased wear and tear caused by a given tenant. Because tenants do not incur the full cost of utilizing the asset, this will lead to greater wear and tear on the same asset by renters than by owner-occupiers. In the standard user cost framework, this implies that rates of depreciation are higher for the landlord than for the owner-occupier of the same asset. Hence, ceteris paribus, it is more costly to rent rather than own the same dwelling. This is analogous to the classic "lemons" problem (Akerlof 1970).

See Pindyck (1991) for a standard treatment of the option value of investment under conditions of irreversibility and uncertainty and Burda (1995) for a (roughly) analogous application to the decision to migrate.

These estimates will be slightly different than those in Hou (2010) because of differences in age class and because the descriptive statistics are based on a matched sample of owners and renters where 5% of the sample is lost as a result of a poor match between owner-occupied and rental dwellings. The matching routine is discussed further in the Appendix.

As a result of possible economies of scale at the household level, dual-earning couples may find it easier to purchase a home (e.g., because it is easier for them to save for a down payment on a home). To test for this possibility, household income was divided by the square root of the household size. Homeownership rates and trends therein across income quintiles based on this adjusted income were not qualitatively different from those based on unadjusted income levels.

The level of the price-to-rent ratio is not used because, by construction, it will be highly correlated with the relative cost of renting versus owning.

For a given census year, the change in the relative price-to-rent ratio is measured by using the percentage change in its value over the subsequent five years. As its value is not available for 2011, the percent change in the ratio for 2006 is estimated by means of an autoregressive model, which is a function of the percent change between 2001 and 2006.

These costs can be substantial. Real estate fees and closing costs can amount to as much as 7% of the value of the home. If a household stays in a home for five years, this amounts to an annual cost of $5,600 for a home valued at $400,000 when sold.

Goodman (1988) and Dusansky and Koç (2007) included transitory income in their final housing demand model. This approach will not be followed here as including this residual term does not seem appropriate in this context.

Following Bourassa and Hoesli (2010), we assume that the household’s wealth constraint requires a minimum 20% of the value of the house in equity.

When the model was applied to the SFS, 74% of the observations were accurately predicted.

As the tenure choice of households was regressed against independent variables aggregated to the regional level, the standard errors were initially adjusted for the potential correlation of errors within regions. However, because there was no qualitative change in the standard errors, the econometric results are presented without this adjustment.

The sample is restricted to households for which the cost of renting relative to owning an equivalent home is greater than 0.5 and less than 4. The sample is restricted in this way in order to minimize the effect of measurement error on the estimates. Exploration of the micro data suggested that very low or very high relative costs were the result of errors on the file regarding rental rates, rather than genuine differences in the cost of renting relative to owning.

It may seem as though income is controlled for twice here, since the regressions are done across income quintiles and permanent income is included in the model. It should be kept in mind, however, that individuals’ current income may differ from their permanent income, particularly for those in the bottom income quintile.