This section assesses the commonly held assumption that the density of development has decreased over time by separately considering population, dwelling unit, and combined population and employment density calculated on different land bases.
Literature review
Recent research in the Toronto region has linked density to the period during which subdivisions were planned and built. Lehman and Associates et al. (1995) and Blais (2000: fig. 3.11) found that developable area dwelling unit density is considerably lower in most post?Second World War subdivisions than in those developed earlier. While gross residential densities in prewar parts of Toronto range from 28 to 36 units per hectare, the densities of plans of subdivision registered in adjacent municipalities in the late 1990s range from 10 to 15 units per hectare. (See Fig. 10.)
Fig. 10: Residential densities in the GTA
Data compiled by Blais show that in general, developable area dwelling unit density in the GTA is lower the more recently an area was built. Some of the highest-density areas have the lowest proportion of units in apartment form. (Reprinted from Blais 2000: Fig. 3.11.)
There are signs, however, that the density of new development has increased in recent years. Examining registered plans of subdivision in the Regional Municipalities of York, Durham, and Peel, Blais (2000:11-13) found that since the 1970s, developable area dwelling unit densities have increased in urban lower-tier municipalities and that net dwelling unit densities also rose in the 1990s, though they remain significantly lower than in pre?Second World War neighbourhoods. Reports by GHK (2002:31) and Hemson Consulting (2003b:18) show that the increase in net residential density is the result of smaller lot sizes and a greater proportion of higher-density housing forms. This is corroborated by Gordon and Vipond (2005:41-54) who found that in Markham, neotraditional plans of subdivision achieved considerably higher developable area densities than adjacent conventional subdivisions built in the 1970s and 1980s (61 vs. 36.6 persons per hectare).11 They concluded that the higher density of neotraditional plans is due to a combination of factors, including a higher proportion of denser housing types such as townhouses and apartments, smaller lot sizes, and the integration of population-serving employment into mixed-use buildings. Note, however, that the Gordon and Vipond study relies on secondary plans, which indicate what is approved, not necessarily what will be built. Approved development plans may be underbuilt for economic or political reasons. Densities calculated from registered plans of subdivision may therefore be overstated. Only one systematic comparison of built areas to their plans has been performed in the Toronto region -- a 1993 study of Ajax, which found that built densities fell short of those planned (Malone Given Parsons 1993a, 1993b).
There are several possible explanations for a decline in density over time. First, changing professional norms and practices, especially as land use planning was formalized following the Second World War, favoured the production of an urban form dominated by detached single-family houses on larger lots than was previously the norm. Second, the general increase in wealth in the postwar period altered people's preferences, changing the character of housing demand. Third, the government promoted suburban development by supporting access to mortgage capital, enabling households to purchase larger houses on larger lots. Fourth, older areas located closer to the urban core have experienced substantial infill and redevelopment that has increased their dwelling unit densities. More recently built areas located near the edge of the contiguous urbanized area are less likely to have undergone intensification.
With respect to the Growth Plan's target of 50 residents and jobs combined per hectare, Mitra (2007:75-76) and Mitra & Gordon (2007) suggest that the majority of existing urbanized land in the GTA outside of the City of Toronto falls short of this target. Moreover, plans for a major urban expansion area -- North Oakville -- approach but do not exceed the target. The North Oakville East Secondary Plan (Town of Oakville 2007) projects that at full build-out sometime after 2021, 45,000 to 55,000 residents and 25,000 jobs will occupy a gross land area of 2,300 hectares, of which 600 hectares will be a "natural heritage system."12 This results in a gross density of between 30 and 35 residents and jobs combined per hectare, and a developable area density of between 41 and 47 residents and jobs combined per hectare, depending on the resident population.
The relationship between population and dwelling unit densities depends on household size. Neighbourhoods containing the same number of dwellings may have very different populations. For a variety of social and economic reasons, average household size in industrialized countries has been in decline for several decades. Canada-wide between 1971 and 1981, the number of rooms per dwelling increased by slightly less than 6%, even as the number of people in each household decreased by 20% (Blumenfeld 1991). Between 1971 and 2001, average household size across Canada declined from 3.5 to about 2.6 persons per household (Engeland et al. 2005:28). This phenomenon is mirrored at the local level. The average household size in the City of Toronto declined from about 4.0 in 1951 to slightly more than 3.2 in 1971 (Metropolitan Toronto Planning Board Research Division 1974: table 19), and to 2.6 in 2001 (Statistics Canada 2001a). The 2006 Census shows that this trend continues, with the proportion of large households declining while one-person households increased (Statistics Canada 2007).
Research questions
1. Is dwelling unit density lower the more recently a study area was developed?
2. Are the net residential dwelling unit densities of subdivisions built in the 1980s and 1990s higher than those built in the 1960s and 1970s?
3. Do recently developed areas have combined population and employment densities that meet or exceed the Growth Plan's target of 50 residents and jobs combined per hectare?
Findings
Dwelling unit density
Fig. 11: Net residential dwelling unit density
Fig. 11 shows the net residential dwelling unit density of the study areas by era group. Riverdale's density is considerably higher than that of the other pre-1960 study areas, reflecting its small lot sizes and substantial proportion of attached housing forms. Indeed, if Victorian and Edwardian Riverdale is removed, the average of the pre-1960 study area densities is virtually identical to that of the 1970s-80s study areas: 30.8 units per net hectare.
The densities of the 1960s-70s study areas vary considerably. With the exception of Cachet, the density of which is reduced by the presence of a large-lot "estate" subdivision, the densities of the post-1980 study areas are strikingly consistent, ranging from 20.2 to 22.3 units per hectare. Overall, the five post-1980 areas have lower net residential densities than the majority of those developed previously. As will be discussed in Section 2.4, it appears that this is largely a product of the industry's convergence on a limited range of housing types.
Population density, dwelling unit density, and average household size
When the era groups' population and dwelling unit densities are compared, an interesting pattern emerges. On average, all measures of dwelling unit density are lower the more recently a study area was developed, but population density does not follow the same trend. The average population densities of the pre-1960 and 1960s-70s study areas are similar, while the post-1980 study areas are more than one-third lower. (See Fig. 12.) This is because of variations in average household size.
Fig. 13 suggests that the earlier a study area was developed, the smaller its average household size in 2001. The larger average household size in the 1960s-70s study areas compensates for their having lower dwelling unit densities than the pre-1960 study areas. Comparison of the 1960s-70s to the post-1980 groups, however, reveals that the latter group's larger average household size is not sufficient to counter lower dwelling unit density. Only larger household sizes in the newer areas raise the population density to the level observed. (For comparison, the average household size for the GTA as a whole is also shown -- 2.9 persons. This is comparable to the 16-district average of 3.0.)
Fig. 12: Average population and dwelling unit density, by era group
Fig. 13: Average household size, by era group
Why do the post-1980 districts tend to have larger households? Fig. 14 shows average household size, average number of rooms per dwelling, and average number of bedrooms per dwelling for the three era groups in 2001. On average, values for each variable increase the more recently a study area was developed. All things being equal, it may be that larger households are attracted to newer areas located at the metropolitan fringe because they offer larger dwellings. This is only part of the equation, however. Research indicates that the cost of housing also helps determine household location decisions (Will Dunning Inc. 2006; Miller et al. 2004). More generally, a long-term trend towards the construction of larger houses -- that is, those with more rooms or floor space per resident -- is well documented, and appears to reflect increased general wealth. For example, the U.S. Census Bureau (2007) estimates that 44% of new single-family houses were 2,400 ft2 or larger in 2006, up from 12% in 1973.
Moreover, although dwellings in older areas had fewer rooms, they also used to have higher average household sizes and, therefore, higher population densities. Riverdale is a case in point -- while the built form has changed little between 1951 and 2001, its gross population density has changed significantly. In 1951, Riverdale had a gross density of 57,510 people per square mile, or 222 per hectare.13 Fifty years later, its gross density was 85 people per hectare.
Fig. 14: Average household size and average rooms and bedrooms per dwelling, by era group
Combined population and employment density
Combined population and employment densities were also calculated for each study area. On both gross and developable land bases, each succeeding era group has a lower average density. (See Figs. 15 and 16.) Despite their large household sizes, the combined densities of the post-1980 study areas are among the lowest, in part because of their small amounts of employment. Whether calculated on a gross or developable area land base, only half of the cases exceeded the Growth Plan's target of 50 residents and jobs combined per hectare. None of the 1980s-90s study areas meet or exceed the target.
Population and employment do not contribute equally to density values. There are fewer jobs than residents in each study area. Across all 16 study areas, employment density makes up 22% of combined population and employment density. The proportion declines with succeeding era groups, from an average of 33% for the pre-1960 study areas, to 21% for the 1960s-70s study areas, to 13% for the post-1980 study areas. This finding reflects a lower mix of uses in more recent developments. This trend will be discussed further in Section 2.6 (see Fig. 41).
Fig. 15: Gross combined population and employment density
Fig. 16: Developable area combined population and employment density
Summary of findings
1. Is dwelling unit density lower the more recently a study area was developed?
Yes. Whether calculated on a gross, developable area, or net basis, dwelling unit densities tend to be lower the more recently a study area was developed. This supports findings by Blais (2000) and Lehman and Associates et al. (1995).
However, population density does not follow the same trend because, on average, household size is higher in more recently built areas, partly offsetting the lower dwelling unit density. As a result, the population densities of the pre-1960 and 1960s-70s era groups are similar, even though the dwelling unit density of the pre-1960 group is higher. The population density of the post-1980 era group is lower, however, because higher average household size does not compensate for lower dwelling unit density. If average household size continues to decline in the future, population density will also decline.
Higher average household sizes in more recently developed areas may be related to dwelling size. On average, dwellings in more recently developed areas have more rooms and bedrooms than those in older areas. All things being equal, larger households may be attracted to newer areas at the metropolitan fringe because they offer larger dwellings.
2. Are the net residential dwelling unit densities of subdivisions built in the 1980s and 1990s higher than those built in the 1960s and 1970s?
No. All five of the post-1980 study areas have net dwelling unit densities of 22 or less per developable hectare, with four having approximately 20. On average, this is lower than the six 1960s-70s study areas, four of which had densities of greater than 30 dwelling units per developable hectare. The findings of Blais (2000) and Gordon and Vipond (2005), who report that densities in the neotraditional developments of the late 1990s are higher than those in conventionally planned subdivisions of the 1960s and 1970s, are not borne out here. Since only two cases were predominantly constructed in the late 1990s -- Vaughan and Richmond Hill -- this may be an artifact of case selection.
3. Do recently developed areas have combined population and employment densities that exceed the Growth Plan's target of 50 residents and jobs combined per hectare?
No. All of the post-1980 study areas have combined population and employment densities lower than the Growth Plan target of 50 residents and jobs combined per hectare, calculated on either the gross or developable area land base. Densities in the post-1980 study areas ranged from 23 to 47 residents and jobs combined per hectare, with an average of 35.
Some of the 1980s-90s study areas -- Cachet, Richmond Hill, and Vaughan -- contain some vacant land that will likely be developed (although not necessarily for residential use) in the future. These lands account for less than 8% of the gross land base in each case. The presence of vacant land means that the gross and developable area densities at full build-out will likely be somewhat higher than those reported. The net densities, however, will not change, as they were calculated exclusive of vacant land.
Implications for policy
It is noteworthy that none of the five study areas built out after 1980 met the Growth Plan's minimum density target of 50 residents and jobs combined per hectare in 2001. This supports Mitra's (2007:75-76; Mitra & Gordon 2007) finding that the majority of existing urbanized land in the GTA outside the City of Toronto falls short of the target.
There is evidence, however, that the densities of present and future developments will be higher than those of past developments. Gordon and Vipond (2005) found that the expected "mature" developable area population densities of neotraditional subdivisions in Markham are considerably higher than those of existing adjacent conventionally designed neighbourhoods (which include the Markham Northeast study area). If these areas are built out as planned, they will exceed the Growth Plan's target on the basis of population density alone. If these urban development patterns were to be replicated for all greenfield development in the metropolitan region, the target would be met.
Long-term planning must take declining household size into account. Planners must recognize that neighbourhoods will house not only today's population, but also the population expected in the future. If average household size continues to decline, as broader trends suggest may occur, the population densities of existing neighbourhoods will also decline, potentially undermining the Growth Plan's minimum density target of 50 residents and jobs combined per hectare.
Notes
11. The term "neotraditional" is used to describe urban design principles associated with New Urbanism -- narrower streets, garages confined to back lanes, smaller front setbacks, and gridded streets. Note that what Gordon and Vipond (2005) call "gross" density excludes hazards lands, utility corridors, employment lands, expressways, and arterial roads from the land base. Their gross density is therefore analogous to what this study refers to as developable area density.
12. Population and employment values are from the Town of Oakville (2007) ss. 7.3.6 & 7.3.7. Approximate land areas are from Rusk (2007) A9 and <http://www.oakville.ca/nr-07aug13.htm>.
13. The value for 1951 is from Metropolitan Toronto Planning Board Research Division (1974) Table 5. The 1951 value pertains to a slightly larger land base, bounded by the Don Valley Parkway to the west, the East York municipal boundary to the north, Coxwell Avenue to the east, and Eastern Avenue to the south.