This section considers the relationship between density and residential built form. Statistics Canada distinguishes between six housing types: single-detached, semi-detached, row house, duplex apartments, apartments in buildings under five storeys, and apartments in buildings of five or more storeys (Statistics Canada 2004). In this analysis, apartments and duplexes are considered to be non-ground-related housing -- that is, individual dwelling units have no direct access to the street. Single-detached, semi-detached, and row houses are considered ground-related, meaning that units have direct access to the street. The study areas are ranked and classified by dwelling unit and population density.
Literature review
Many studies have sought to quantify the relationship between density and built form. Some are based on analyses of hypothetical cases, while others examine existing built areas.
Housing type mix and net residential dwelling unit density
In an analysis of 99 hypothetical housing schemes reflecting a range of combinations of dwelling types, unit sizes, lot sizes, and block configurations, Alexander (1993:192-96) found a "clear association of certain parts of the range of possible densities with specific dwelling forms," with single-detached, row house and low-rise apartments, and high-rise apartments each occupying distinct ranges of net dwelling unit density.
In a 1976 study, Diamond found that high-rise redevelopment does not necessarily increase site density. Assuming constant floor area per unit, he found that the relationship between density of built form and land consumption is non-linear and that the greatest reduction in land consumption per unit occurs between 0.75 and 1.5 FAR, corresponding to a shift from row housing to walkup apartments. Densities of more than 1.5 FAR provide little additional advantage in terms of efficiency of land use. (See Fig. 27.)
Fig. 27: Land consumption per dwelling unit: a non-linear relationship
Higher-rise forms produce diminishing returns with respect to land consumption. Adapted from Diamond (1976):15.
Housing type mix and gross dwelling unit density
Studies of existing built areas indicate that the same array of housing types can produce different gross dwelling unit densities through different street configurations, block sizes, lot sizes, site layouts, and designs. Conversely, neighbourhoods with different mixes of housing types can have the same density. For an analysis of housing and density, see, for example, Hemson et al. (1993:18); BLGDG (1995); Urban Design Advisory Service (1998:29); Design Center for American Urban Landscape (n.d.); Campoli and MacLean (2007); and CMHC (n.d.).
While net dwelling unit density is closely related to floor area and therefore built form, gross density depends on the amount of undevelopable and public land within the gross land area. (See Section 2.3.) Since some types of public land are allocated in proportion to population, increasing net density through higher-rise forms produces diminishing returns in terms of gross density. The maximum carrying capacity of a given land area is reached when the amount of public land required to serve the population begins to compete with the land required to house it. Beyond this point, one can increase only at the expense of the other.
Research questions
1. Do study areas with higher net residential dwelling unit densities have more non-ground-related dwellings in their housing type mix? Conversely, do study areas with lower net residential dwelling unit densities have more single-detached dwellings in their housing type mix?
2. What role does housing type mix play in determining population density?
Findings
Net residential dwelling unit density
Fig. 28 shows the housing type mix of each study area in ascending order of dwelling unit density calculated in relation to the net residential parcel area. In general, the higher the density, the lower the proportion of single-detached dwellings. The proportions of the different types of attached dwellings varies greatly from one study area to another, however.
While the presence of apartment buildings is important to achieving higher net residential densities -- indeed, no study area with more than 30 units per net hectare contains less than 30% apartments -- they need not be in high-rise form. While containing a much higher proportion of high-rise apartments than the other cases, classic "tower-in-the-park" neighbourhoods such as the Peanut and Mississauga Valleys have only three-quarters the density of Riverdale. Riverdale achieves the highest net residential dwelling unit density with fewer apartments in buildings over five storeys than any study area with more than 30 dwelling units per net hectare. At the same time, Riverdale has about the same proportion of attached dwellings as the Richmond Hill study area and ground-related dwellings as the Whitby study area, both of which have relatively low densities.
Fig. 28: Housing type mix of study areas by net residential dwelling unit density
Fig. 29 tests the degree to which net residential dwelling unit density is associated with the prevalence of different housing types. Fig. 29A shows that higher net density correlates with a lower proportion of single-detached dwelling units. The relationship is weaker with respect to units in apartment and duplex form (Fig. 29B) and the proportion that are ground-related (Fig. 29C). Fig. 29D indicates that there is no meaningful association between higher density and the proportion of units in attached form. The proportion of detached dwellings in the housing type mix appears to be the largest determinant of net residential density.
Fig. 29: Correlation of net residential dwelling unit density and housing type mix
A. % Single-detached
B. % Apartments and duplexes
C. % Ground-related
D. % Attached (semis and rowhouses)
The closer the R2 value is to 1, the more the independent variable (housing type) explains the variation in the dependent variable (density).
Net residential population density and average household size
Variation in average household size has a profound effect on the impact of housing type mix on population density. Neighbourhoods with low dwelling unit densities but larger households have higher net population densities than they would if household size was constant across all neighbourhoods.
Fig. 30: Housing type mix and average household size of study areas by net residential population density
Fig. 30 ranks the study areas in order of increasing net residential population density. Compared to Fig. 28, some areas with low dwelling unit densities rank higher in order of population density than dwelling unit density because of variations in average household size. For example, Malvern, which at 3.7 persons per household has a higher-than-average average household size, shifted from eighth to fourth place in the density ranking. The Old Oshawa and Oshawa West study areas, which average 2.2 and 2.3 persons per household respectively, slip from fourth and fifth place to eighth and ninth place.
Summary of findings
1. Do study areas with higher net residential dwelling unit densities have more non-ground-related dwellings in their housing type mix? Conversely, do study areas with lower net residential dwelling unit densities have more single-detached dwellings in their housing type mix?
The answer is a qualified yes to both questions. The proportion of detached dwellings in the housing type mix is the most significant determinant of net residential dwelling unit density. In general, the higher the proportion of detached dwellings, the lower the density. The proportion of apartments and duplexes in the housing type mix is only loosely associated with higher net residential density. All cases with a net residential density of over 30 units per hectare have a housing type mix in which non-ground-related housing accounts for more than 30% of the housing stock. Nevertheless, non-ground-related dwellings need not be in high-rise form for high densities to be achieved. This finding supports Diamond's (1976) assertion that, beyond a certain threshold, high-rise development does not necessarily increase site density.
2. What role does housing type mix play in determining population density?
The impact of housing type mix on population density is mediated by average household size. Larger average household size amplifies the impact of housing type mix on population density; smaller average household size diminishes it. Areas with higher-than-average household sizes therefore have higher population densities than would otherwise be the case if household sizes were constant everywhere. Many of the lower-density study areas have above-average household sizes. Were household sizes to decline in these areas, their already low population densities would decrease further.
Implications for policy
If average household size continues to decline, the population density "bonus" observed in newer neighbourhoods may fade, reducing the viability of public transit and efficient provision of other services. Given the correlation between average number of rooms and bedrooms per dwelling and average household size demonstrated earlier (see Fig. 14), policies to encourage family-oriented dwelling forms and configurations of space within buildings that will attract and retain larger households, regardless of their relationship to the ground, may be required. Indeed, this is what the Growth Plan calls for in its vision for "complete communities."