This project was conceived as a complement to an existing body of research that has analyzed the Toronto metropolitan region as a whole. Looking deeply at smaller pieces of the region was intended to add nuance and support to these studies by revealing "micro" phenomena that are not visible at the "macro" level. The result is, in a sense, a conversation between original research and academic and professional studies undertaken at similar and different scales.

Many of the conclusions usefully accord with previous findings; others are surprising. This concluding section draws together the broad themes and their implications with an emphasis on providing insights into how to improve planning policies in the Toronto region and elsewhere.

Density should be supplemented by other measures in planning practice.

At the outset, this report questioned the use of density in plans as a proxy for desired urban form characteristics and land use outcomes. Extensive investigation and manipulation of densities and density values in this study leads to the conclusion that while density is a useful indicator of the efficiency of infrastructure and service provision, especially for public transit, it tells us little or nothing about other important attributes of urban form: housing type mix, the degree to which uses are mixed, the contiguity of the urbanized area, and the connectivity of street systems. All of these attributes have been shown to influence matters of central concern to planning policy: transportation behaviour and the preservation of natural heritage systems, not to mention economic and social processes.

Dwelling unit density is only a loose proxy for housing type mix, and only when calculated on the net residential land base. Given the wide variation in average household size, population density is an even poorer indicator of housing type mix, and vice versa.

The introduction of combined population-plus-employment density in the Toronto region context is an important advance insofar as it provides a better sense of the viability of public transit than population density alone. But rolling people and jobs together obscures the balance between the two, and therefore is a poor indicator of the degree of mix of use on the land base in question.

The utility of measuring or setting targets for densities over broad areas is limited. First, density numbers are averages. Densities can be lower in some locations if they are offset by higher densities in others. The larger the land base, the less the overall density number represents the parts. The Growth Plan's minimum density target applies to the designated greenfield areas of municipalities, land bases which comprise thousands of hectares. The purpose of the target is to reinforce policies that, in part, are intended to encourage urban form that supports public transit and the efficient provision of infrastructure. As these goals require a minimum density threshold throughout the urbanized area, it would be sensible to pair the overall density target with minimum density targets for smaller areas as well. Indeed, the Growth Plan already does this for certain classes of nodes, which it refers to as "urban growth centres" (MPIR 2006a; Filion 2007).

Such an approach has been pursued in the United Kingdom. Since 2002, local planning authorities in rapidly growing parts of the country have been required to consult the national government before permitting individual developments of less than 30 dwelling units per hectare on greenfield land (Dept. of Communities and Local Government 2006: s. 47). The government has indicated that it will intervene if this threshold is not met. The policy has been effective: between 2001 and 2006, the average density of development on greenfield land rose from 25 to 41 units per hectare of developable land (Dept. of Communities and Local Government 2007:7).19 The U.K. approach has a further benefit: its effects can be observed immediately. By comparison, it will not be possible to evaluate whether the Growth Plan's target has been met in a municipality until the designated greenfield area has been fully built out. Although the provincial government has expressed its intention to monitor progress at five-year intervals, it has not stated a method for doing so (MPIR 2006b).

Second, given the large variation in the jobs density of different types of facilities located in single-use employment areas (office, industrial, and retail parks), the Growth Plan's inclusion of spatially segregated employment zones in the envelope to which the greenfield density target is applied is problematic. Its impact on the overall density average being unknown, the inclusion of employment land may lead to incorrect inferences as to how well the remaining mixed community areas can support transit or other infrastructure. Setting separate or additional targets for the density and other characteristics of employment lands would allow for greater precision and create an additional incentive to mix employment into residential areas. The recently released background paper on employment (MPIR 2008) indicates that the Province is seeking to better understand the characteristics of employment land, but not necessarily to set thresholds and targets.

Third, different definitions of "gross" density should be used with care. This study has differentiated between "gross area" (all land) and "developable area" (all land net of protected natural heritage features, transportation and infrastructure corridors, and hazard lands). The Growth Plan's minimum density target is intended to apply to a land base similar to the developable area used in this study. A developable area density, however, does not in itself reveal how much land is designated undevelopable by conservation authorities and other public agencies. This study could not reveal whether changes in environmental protection and other standards have served to reduce the envelope of land available for urban development more in recent years than in the past. It is clear, however, that such reductions produce lower gross densities and may undermine the contiguity of urban areas and therefore the efficient provision of public transit and other infrastructure.

The Growth Plan contains many policies that, if fully reflected in municipal plans and actual urban development patterns, will create built environments that are more mixed, accessible, dense, and serviceable by public transit (MPIR 2006a: s. 2.2.7). Achieving these goals might be more effective if, in addition to minimum density, other measures, such as the degree of urban contiguity, mix of use, and neighbourhood accessibility, were also tracked.

An already changing housing type mix is likely to deliver higher densities.

The densities of the districts examined in Section 2 tend to be lower the more recently they were planned and built out. All of those built in the 1980s and 1990s have combined population and employment densities lower than the Growth Plan's target of 50 residents and jobs combined per hectare for future greenfield development. This fits with other research that indicates that much of the established urban area surrounding the City of Toronto falls short of the target, including areas planned or developed as recently as the late 1990s (Blais 2000; Mitra 2007).

There are signs, however, that a shift in the type of housing being built is leading to higher overall densities. Although this phenomenon was not observed in the analysis in Section 2, both Blais (2000) and Gordon and Vipond (2005) have found that planned developments from the late 1990s appear to have higher densities than those built previously. More generally, data indicate that the share of housing construction consisting of single-detached dwellings declined in suburban areas in the late 1990s. The "compact" growth forecast contained in the provincial government-commissioned Growth Outlook (Hemson 2005), which is incorporated into the Growth Plan (MPIR 2006a: Schedule 3), assumes that this trend will continue.

The analysis in Section 2 suggests that the most powerful predictor of net residential density is the proportion of single-detached dwellings in the housing type mix. The smaller lot sizes that accompany the move from detached to attached housing appear to be more decisive in producing higher densities than increasing the proportion of apartments. This aligns with Diamond's (1976) earlier findings. At the same time, however, 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 mix. These need not be in high-rise form, however. The sketch modelling in Section 3 suggests that a shift in the housing type mix will contribute significantly to the achievement of the minimum density target.

The changing composition of households could have a profound effect on the future viability of services.

To plan is to try to anticipate future events. The future often has a way of turning out differently from what we expect, however. Population growth, housing demand, and employment projections are subject to national and international political, economic, and environmental changes and forces. We may try to anticipate some of these factors -- climate change, immigration levels, and rising energy costs, for example -- but others will no doubt take us by surprise.

One phenomenon that is studied by demographers is the long-term trend of changing household composition. In recent decades, social change has led to an increasing number of one- and two-person households, while the proportion of large and multi-generational or multi-family households has declined. The average size of households has therefore declined over time, with the effect that more dwelling units are required to house the same number of people and the population density of established urban areas is declining.

This trend poses a challenge to planners, who must organize urban development not only to accommodate today's population, but also the population that is expected to exist in the future. For example, as noted in Section 2.2, who would have predicted in 1951 that Riverdale's population density would decline by over 60% over the subsequent half-century? Who would have predicted that between 1951 and 2001, the number of dwellings in the pre-amalgamation City of Toronto (corresponding more or less to the prewar city), would increase by 85%, even as the population remained virtually the same? Only extensive redevelopment has maintained the levels of population -- and population density -- on the same territory.

Declining household size has important implications not only for meeting the minimum density targets in the Growth Plan, but also for achieving their purpose. Should average household size continue to decline, a subdivision planned and occupied today at 50 residents and jobs combined per hectare will very likely have a lower population density in the future. Lower densities will reduce the efficiency of infrastructure investment and service provision, and undermine the use and cost-effective provision of public transit. One possible solution is to overshoot the target now in anticipation of future population declines. Another solution is to "loosen up" planning by encouraging flexible building forms and configurations of uses that can be adapted to different potential futures.

Greater mix of use may actually reduce densities measured at the district scale.

One of the surprising findings of the sketch modelling in Section 3 was that greater mix of use -- that is, increasing the amount of employment land in a district's developable area -- reduced gross density, because jobs density on employment lands tends to be lower than the population density of residential areas. Increasing the amount of employment land will result in an overall density increase only if the jobs density on employment lands equals or exceeds the population density in residential areas. Redistributing land uses at the metropolitan regional scale to promote greater local-area mix of use may therefore frustrate the goal of increasing local-area densities. And if higher-density housing were built, thereby increasing population density, it would further unbalance the local-area ratio of jobs to population, by increasing the number of residents relative to jobs in the area.

While the claim that land for public facilities is increasing at the expense of land for private development can be neither confirmed nor rejected, it is clear that smaller and smarter allocations for public facilities would increase densities.

The analysis in Section 2 could not corroborate the claim made in other studies that, in aggregate, increasingly generous standards for public facilities such as parks, schools, roads, and environmental protection areas have reduced the amount of land available for private development, and therefore density. In fact, the division between public and private property varies little across the 16 study areas. Disaggregating public land use categories reveals that the proportions of developable land area accounted for by parks, schools, and roads vary considerably both within and between era groups. It may be that the economics of private-sector land development has held the overall proportions of public and private land steady over the years, while ever-changing ideas of "good" planning have resulted in quite different arrangements of land within the public component.

A claim about rising standards for public facilities is difficult to prove one way or another, as in the postwar period, public facilities such as parks and schools have typically been allocated in proportion to the expected resident population. The characteristics of the resident population have changed over time, however, as household sizes have declined. A case for rising standards could be made only by comparing, for a variety of neighbourhoods built out in different time periods, planned public facilities allocation levels and populations to the actual population at the time the areas were first built. The one trend observed in the analysis is that parkland per capita and per household tends to be higher the more recently an area was developed, but this does not prove that parkland allocation standards have increased over time.

While the search for trends may be inconclusive, it is clear that smaller and smarter allocations for public facilities could increase density by expanding the amount of land available for private residential and commercial development. It is already common practice to plan dual-use park and schoolyard facilities. Other measures, such as locating playing fields on flood plains and integrating parks into natural heritage systems traditionally considered external to subdivisions, would also increase the density of the developable area. In Section 3, it was found that a 20% reduction of per-capita standards for public facilities increased combined population and employment density by about 12-15%. This is, however, greater than the 7-11% increase produced by a 20% rise in net residential dwelling unit density.

While meeting the Growth Plan's minimum density target is feasible, the promise of "complete communities" will likely be less easily fulfilled.

While the notion of a "complete community" is couched in the language of meeting a diverse population's daily needs, the primary functional objective is to change personal travel behaviour. Putting jobs, services, and public facilities close to each another and to housing is expected to cause people to make fewer trips and travel shorter distances. Embedding a broad mix of uses in an accessible urban form served by a high-quality and integrated public transit system is expected to lead people to make more trips on foot, bicycle, and transit, and reduce trips by automobile.

Sections 2.6 and 2.7 suggest that this outcome will not be easily achieved. The most difficult nut to crack is the location of employment. The case analysis and literature review indicate that workplaces are largely segregated from residential areas. Most jobs and services are located in specialized zones -- business and industrial parks, and shopping and retail power centres -- that are primary accessible by automobile and poorly connected to residential neighbourhoods. Most residents' daily destinations therefore cannot be reached in reasonable time on foot, by bicycle, or even by surface public transit. Whether or not a transit system can be devised that would efficiently connect the majority of trip origins and destinations is open for debate. Today's pattern of "everywhere-to-everywhere" inter-suburban commuting would seem to militate against this goal (Soberman et al. 2006; Miller & Shalaby 2000; Mitra 2007).

The postwar districts examined in Section 2 are far from "complete" in the sense of containing a diverse mix of land uses and a balance of population and jobs. Most of the study areas contain very little employment land; those that do are located near highway or rail corridors, to which the employment lands are oriented. Little land for employment translates into fewer jobs in these districts, both in absolute terms and in proportion to the resident population. The ratio of jobs to resident employed labour force -- in other words, jobs-housing balance -- is very low in most of the study areas analyzed, meaning that the residents of these areas do not have the option of working near where they live.

It should be recognized that not all land uses are compatible; separation of uses is often necessary. But while heavy manufacturing or logistics and residential neighbourhoods are not compatible, if only because of the traffic volumes they generate, offices, retail, and some kinds of services are certainly compatible with neighbourhoods. If the observed pattern of thoroughgoing segregation of land uses at the metropolitan regional and neighbourhood scales continues, there is little potential to bring about significant change in the way people travel through the city.

The location of jobs is important, as visits to other destinations typically occur as part of work-related trip chains -- in other words, on the journey to and from work. For this reason, the automobile mode share for shopping trips observed in Section 2.7 equals or exceeds those for journeys to work. Simply put, if people drive to work, they tend to buy their groceries on the way home, at least partly because retail and services are increasingly located on large-scale automobile- or highway-oriented sites, often on formerly industrial land, catering to large market areas. While it may be possible to relocate retail and services to within the residential urban fabric, trip chaining means that this change alone may not be enough to alter travel behaviour. The exceptions to this pattern are trips to school and for childcare. In most of the cases analyzed in Section 2, walking, cycling, transit, and trips by school bus together account for more than half of all trips to school or for childcare, competing favourably with the automobile.

Greater mix of use and higher density are necessary but not sufficient conditions for shifting travel behaviour, however. The cost and convenience of alternative means of travel -- socio-economic factors -- are also important. The observed automobile mode shares for shopping trips are high, not just because of trip chaining, but also because the car is a more convenient way to transport goods. As Dieleman (2002) puts it, people who own cars will use them. And they do, even in mixed and dense metropolitan core areas like Riverdale. Moreover, even if there were enough employment land to support one job for every member of the resident labour force, there is no guarantee that residents will choose to work locally, even if local jobs matched their skills and expectations. Members of multi-worker households very often work in disparate locations.

Given all of this evidence, the impact on transportation behaviour of the creation of more dense, mixed, and "complete" communities is likely to be incremental rather than transformative. This does not mean that creating "complete communities" is an invalid goal. Seemingly small reductions in automobile use can produce significant declines in road congestion today (Miller & Shalaby 2000). More walkable and transit-supportive neighbourhoods, combined with increased proximity of homes to jobs, services, shopping, and amenities, will be better able to adapt to a future in which energy, and therefore automobile use, is more expensive.

The immediate challenge will be to translate the Growth Plan's general principles into real places that truly embody them. This study suggests that this may require more robust monitoring of outcomes at a variety of scales and significant changes in the day-to-day practice of planning and land development.

Existing postwar suburban areas will be hard to retrofit.

The Growth Plan can be read as a response to the perceived failures of postwar urban development patterns, which are too low-density and too segregated by land use, with street networks that are incompatible with pedestrians, cyclists, and transit. Analysis of the case studies in Section 2 suggests that the belt of lower-density suburbs comprehensively planned in the 1960s, 1970s, and 1980s may be difficult to "retrofit" to create the characteristics specified in the Growth Plan.20 Existing street networks, and therefore neighbourhood accessibility levels, are unlikely to change, and highly segregated land use patterns are also not easily reversed. While site-by-site redevelopment may introduce additional jobs and people into the established segregated urban fabric, a generalized increase in local-area mix of use and density would take years, if not decades. In fact, intensification must offset the effects of declining average household size before a net increase in density occurs.

Even if future greenfield development fulfils the vision in the Growth Plan, existing lower-density and "incomplete" suburbs will remain. In the long run, the result may be a reverse-doughnut pattern in which a dense metropolitan core is surrounded by lower-density suburbs, which is in turn surrounded by a newer, higher-density band. The challenge of how to raise the performance of the middle realm, and to efficiently connect the three urban realms by transit, is formidable, and will not be easily resolved.

Change will take time.

Like the proverbial ocean liner, which cannot be turned on a dime, it will be years before the Growth Plan produces demonstrable change. While all development applications must immediately conform to the plan (Places to Grow Act, 2005, s. 14(1)), municipalities have until June 2009 to bring their official plans into conformity (Places to Grow Act, 2005, s. 12). It will probably be several years into the next decade before the Growth Plan's policies are reflected in the full hierarchy of planning documents: from upper- and lower-tier municipal official plans to secondary plans and zoning bylaws. It will be later still before a visible portion of the built environment reflects the impact of the Growth Plan. Indeed, there are tens of thousands of dwellings "in the pipeline" -- planned and approved under previous rules -- that must be absorbed first (CMHC & MMAH 2003). All of this means that it will be years before the impact of the Growth Plan can be assessed.

19. For a description of the land base used to calculate these density values in the United Kingdom, see Dept. of Communities and Local Government (2006: Annex B).
20. For an interesting collection of articles on "retrofitting" the suburbs, see Plan Canada 36:4 (1996).