The literature suggests that density, diversity of land uses, and neighbourhood design -- Cervero and Kockelman's "3Ds" (1997) -- influence travel behaviour. By putting residents and jobs closer to one another, a high-density, mixed-use urban fabric is believed to decrease the length and number of automobile trips in favour of walking and cycling. As discussed in Section 2.5, this effect is believed to be amplified if the street network is more "accessible."
Building on the previous section, which found that land uses are highly segregated at both the neighbourhood and metropolitan region scales, this section examines the travel behaviour of study area residents. The section concludes with a discussion of potential implications for policies that seek to redirect people from automobiles to public transit, walking, and cycling.
Literature review
A wealth of empirical research on factors influencing travel behaviour has been produced in recent decades. As Crane and Crepeau (1998), Boarnet and Crane (2001), Zhang (2004), and Williams (2005) note, this work has been uneven in its methods and often inconclusive and contradictory in its findings. Where correlations and causal relationships have been discerned, researchers disagree about their strength. Still, they have identified a variety of factors that influence travel behaviour.
Newman and Kenworthy's (1999:101) widely cited correlation between density and energy intensity of transport (a loose proxy for automobile use) was based on a comparison of aggregate values for metropolitan regions. As one moves from "macro" to "micro" analysis, however, the findings become more ambiguous.
A variety of built environment variables have been shown to influence travel behaviour, including density, mix of use, and accessibility. In a statistical analysis of land use and travel behaviour in the Puget Sound region, Frank and Pivo (1994) found that travel mode choice was related to the population and employment density of both the origin and destination of trips, though the relationship is not linear. The number of pedestrian shopping trips increased only when density surpassed 32 residents per hectare at trip origin. Trips to work by single-occupant vehicle decreased only when the density at trip destination was higher than 185 jobs per hectare. Frank and Pivo also found that increased mix of use also correlated positively with walking. In a study of Portland-area neighbourhoods, Greenwald (2003) found that pedestrian trips increased at the expense of automobile trips in neighbourhoods featuring accessible street patterns, but that transit use was unaffected.
Similarly, Cervero and Kockelman (1997) used 1990 travel diary data for San Francisco Bay Area residents in conjunction with indicators of density, diversity of uses, and street configuration for 50 census tracts to demonstrate that higher density, use mixture, and street network connectivity lead to small but statistically significant declines in automobile use and trip length. Kulash et al. (1990) found that more connected street systems may reduce vehicle distance travelled within neighbourhoods, but noted that intra-neighbourhood trips constitute only a small proportion of total trips. Distinguishing between travel to local and distant destinations, Handy (1992) explored the linkage between local-area and metropolitan regional urban form and land use patterns in determining travel behaviour.
Others have demonstrated that socioeconomic and demographic variables such as age, gender, income, household structure, and participation in the labour force are also important influences on travel behaviour. On the basis of an Atlanta survey, Helling (1996) found that men's and women's travel behaviour differs. Dieleman et al. (2002) found that in Holland, the presence of children in the household is a more powerful predictor of automobile use than labour market participation.
Some studies link physical and socio-demographic variables. In a study of the Greater Toronto Area, Riekko (2005) found that the propensity to use public transit increased with higher densities, proximity to the metropolitan core, grid street patterns, shorter blocks, greater mix of use, and proximity to rapid transit. At the same time, however, individual and neighbourhood socioeconomic and demographic variables are of central importance: age, gender, dwelling tenure, employment status, and income all affect transit use. Riekko concludes that sociodemographic variables have stronger explanatory power than urban form variables. In a study of 31 cities in the United States, Canada, and Europe, van de Coevering and Schwanen (2006) also found that socioeconomic variables are more important than urban form variables in determining average commuting distance and time and automobile mode share. Scheiner and Kasper (2005) and Schwanen et al. (2005) both investigated how the connection between land use and transportation behaviour is mediated by household composition and associated lifestyles.
"Supply-side" variables also play a role. Proximity of trip origins to potential destinations (including local area jobs-housing balance) clearly matters. Cervero and Duncan (2006) found in a San Francisco Bay Area study that jobs-housing balance is much more effective than retail-housing balance at reducing vehicle distance and hours travelled. Dieleman et al. found that car ownership was the single most important determinant of automobile use: "If people own a car, they use it" (2002:524). Schimek (1996) found that in the United States, average household income (and therefore automobile ownership) was a much stronger predictor of automobile travel than density at trip origin. While automobile ownership is related to income, it is also related to need, indicating complex relationships among density, income, mix of use, and auto ownership and access (Schimek 1996; Vandersmissen et al. 2004). Badoe and Miller (2000:254-55) found that frequency of public transit service is also important, as transit is a viable alternative to the automobile only if it efficiently and conveniently connects origins and destinations.
In a study of Boston and Hong Kong, Zhang (2004) found that land use variables and the relative cost of travel have roughly equivalent effects. He suggests that to maximize the impact of built environment factors, policymakers must also "get the prices right."
Still others have focused on attitudes and perceptions, finding that predispositions, expectations, and self-selection play an important role in determining the propensity to walk or cycle to destinations rather than drive. On the basis of a survey of residents in six neighbourhoods in Austin, Texas, Handy et al. (1998) found that residents' propensity to walk was influenced not only by urban form and proximity to destinations, but also by perceptions of safety, shade, time limitations, and the attractiveness of the visual environment. In another study, Eid et al. (2006) found that rather than suburban form (commonly characterized as sprawl) "causing" obesity, obese people may choose to live in low-density environments; conversely, people who choose to live in older areas closer to the metropolitan core may be more inclined to walk, cycle, or use transit than those living in more peripheral areas.
In a synthesis of more than 50 different empirical studies of the relationship between travel behaviour and the built environment, Ewing and Cervero (2002) conclude that trip frequencies, regardless of mode, are more a function of socioeconomic variables than the built environment. For trip lengths, however, the reverse is true: the configuration of the built environment is more determining. Mode choice appears to depend on both, although socioeconomic variables may be more important than built form.
In short, the research suggests that both local-area urban form at the origin and destination of trips (e.g., street configuration, density, neighbourhood design, and use mixture) and the proximity of different land uses at the metropolitan region scale must be appropriate if people are to make fewer and shorter trips by automobile and more trips on foot, by bicycle, or on public transit. Moreover, the relative cost and convenience of the alternatives must be competitive with the automobile. Regardless of urban form, however, people tend to use cars if they own them.
Research questions
1. How do the travel mode shares for each study area differ by purpose of trip?
2. What are the likely causes of the observed patterns?
Fig. 43: Motorized (automobile, taxi, and motorcycle) mode share for journeys to work, school and childcare, and shopping
Findings
To get a sense of residents' transportation behaviour for particular purposes, the mode shares for journeys to work, school and childcare, and shopping were retrieved from the 2001 Transportation Tomorrow Survey (TTS). (See Figs. 43-45.) The 2001 TTS surveyed travel behaviour in the GTA, as well as the cities of Hamilton, Guelph, Orangeville, Barrie, Orillia, Peterborough, and Kawartha Lakes, the Regional Municipality of Niagara, and the counties of Wellington, Simcoe, and Peterborough. Because of boundary mismatches with the TTS data, Old Oshawa and Meadowvale are not considered in this section. Shopping trips by walking and cycling may be underrepresented due to the survey's methodology.17 See Appendix B for more information about the TTS.
Fig. 44: Non-motorized (walking and cycling) mode share for journeys to work, school and childcare, and shopping
n all study areas, more than half of all journeys to work were by automobile, taxi, or motorcycle. In the post-1980 study areas, over 80% of journeys to work were by automobile, taxi, or motorcycle. Public transit accounted for most of the remainder. Riverdale and Leaside had the highest proportion of journeys to work by walking and cycling, at 12% and 8%, respectively. In these areas, the high non-motorized mode share for the journey to work may reflect the large number of jobs close to residential areas and an urban form conducive to reaching them.
School and childcare trips show a different story. Most journeys to school or childcare were made by walking, cycling, or public transit in all but three of the study areas. Walking and cycling together account for between 24% and 45% of journeys to school in the pre-1980 study areas. In the five post-1980 study areas, however, the highest value is 25%, and for Cachet and Richmond Hill the values are less than 2%. (The latter value is not surprising, as the Richmond Hill study area contains no schools.)
Fig. 45: Combined public transit (local and regional) and school bus mode share for journeys to work, school and childcare, and shopping
In all cases, automobile, taxi, and motorcycle combined accounted for the vast majority of shopping trips. Even in dense and highly mixed Riverdale, the automobile accounted for 70% of all shopping trips, while walking and cycling together accounted for only 8%. Since all of the study areas contain shops of one kind or another, there are two possible reasons for this. First, convenience. It is much easier to transport goods, especially in large quantities, by automobile than by other means. Second, shopping and the use of personal services, even in the mixed-use inner city, often occurs on journeys with multiple destinations. Shopping is often incorporated into trips to and from work. If a household already owns and regularly uses a car to travel to and from work, it will usually be used for other purposes, even if intermediate stops are accessible by other means.
On average, the mode share values for the study areas are similar to those of the metropolitan region as a whole. According to the TTS, 94% of shopping trips and 88% of journeys to work were made by automobile in the region as a whole in 2001. Only the City of Toronto study areas -- Riverdale (30%), Leaside (17%), the Peanut (22%), and Malvern (20%) -- have higher mode shares for public transit.
Fig. 46: Relating motorized mode share to neighbourhood accessibility
Lower composite neighbourhood accessibility scores indicate greater neighbourhood accessibility. Old Oshawa and Meadowvale are omitted due to lack of TTS data.
The remainder of this section will explore the potential relationships between travel behaviour and two variables -- neighbourhood accessibility and density. As a valid statistical analysis is not possible with a sample of 16, the discussion assesses the degree to which the characteristics of the cases correspond to expectations established by the literature.
Potential influences on travel behaviour
The literature suggests that neighbourhood accessibility has a modest influence on travel behaviour. Fig. 46 plots the study areas' motorized mode share (automobile, taxi, and motorcycle) against their rank on the composite neighbourhood accessibility indicator developed in Section 2.5. As might be expected, the chart indicates that the districts with the lowest motorized mode shares also tend to score better in terms of neighbourhood accessibility. The wide spread of accessibility scores among the cases with similar mode shares suggests that other factors are more decisive in determining automobile use, however. Location within the City of Toronto appears to be a stronger predictor of the use of non-automobile modes of transportation than either the era of development or the neighbourhood accessibility ranking.
Fig. 47: Relating non-motorized mode share to neighbourhood accessibility
Lower composite neighbourhood accessibility scores indicate greater neighbourhood accessibility. Old Oshawa and Meadowvale are omitted due to lack of TTS data.
Fig. 47 plots the combined walking and cycling mode share for the study areas against the composite neighbourhood accessibility scores. The relationship is very weak. Again, being located in the City of Toronto appears to be a much stronger predictor of greater walking and cycling than accessibility.
In Fig. 48, the motorized mode share of 14 of the study areas is plotted against gross combined population and employment density. A coherent pattern is visible: as density increases, motorized mode share decreases. The study areas fall into two discrete clusters (marked with circles). The first cluster, with densities of less than 50 residents and jobs combined per gross hectare, has motorized mode shares between 80% and 90%. The second cluster, with densities of between 61 and 83 residents and jobs combined per hectare, has mode shares between 70% and 80%. Riverdale stands alone, with a considerably higher density and lower motorized mode share than all other cases. Density seems to have greater explanatory power than neighbourhood accessibility.
Fig. 48: Relating motorized mode share to density
Old Oshawa and Meadowvale are omitted due to lack of TTS data.
Location within the metropolitan region also appears to have a significant, even decisive effect. Fig. 49 ranks the study areas by distance from Toronto's central business district. In general, the greater the distance, the higher the automobile mode share and the lower the mode share for transit, walking, and cycling. With few exceptions, the study area mode shares correspond to the values of their parent municipalities.
The four study areas with the lowest automobile mode shares are in the City of Toronto. It may be that the long-standing, frequent, and integrated service provided by the Toronto Transit Commission, combined with planning policies that have traditionally linked land use and transportation objectives, are responsible for the City of Toronto's high performance (Miller & Soberman 2003:35). All of the 1980s-90s study areas are in the high motorized share cluster. The Milton, Whitby, and Oshawa West study areas, located in self-standing towns at the periphery of the metropolitan region, have the highest motorized mode shares and lowest transit mode shares.
Fig. 49: Mode shares of study areas and municipalities, ranked by regional location
Study area (municipality) | Auto | Transit | Walk & cycle | Approximate distance from Toronto CBD |
City of Toronto | ||||
Riverdale (Toronto) | 57% (68%) | 30% (22%) | 13% (8%) | 2 km |
Leaside (Toronto) | 74% (68%) | 17% (22%) | 8% (8%) | 6 km |
Peanut (Toronto) | 70% (68%) | 22% (22%) | 8% (8%) | 15 km |
Malvern (Toronto) | 72% (68%) | 20% (22%) | 9% (8%) | 22 km |
Greater Toronto Area outside of Toronto | ||||
Mississauga Valleys (Mississauga) | 79% (84%) | 15% (8%) | 6% (5%) | 21 km |
Richmond Hill (Richmond Hill) | 87% (86%) | 12% (8%) | 1% (3%) | 22 km |
Cachet (Markham) | 86% (87%) | 11% (7%) | 3% (4%) | 25 km |
Vaughan (Vaughan) | 86% (87%) | 10% (5%) | 4% (4%) | 27 km |
Markham Northeast (Markham) | 87% (87%) | 7% (7%) | 6% (4%) | 28 km |
Glen Abbey (Oakville) | 80% (87%) | 14% (5%) | 6% (4%) | 37 km |
Bronte (Oakville) | 86% (87%) | 7% (5%) | 7% (4%) | 39 km |
Whitby (Whitby) | 88% (76%) | 7% (5%) | 5% (6%) | 42 km |
Milton (Milton) | 90% (92%) | 3% (1%) | 7% (5%) | 43 km |
Oshawa West (Oshawa) | 90% (88%) | 6% (4%) | 4% (6%) | 48 km |
Source: TTS, all trips.
Summary of findings
1. How do the travel mode shares for each study area differ by purpose of trip?
The combined mode share of automobile, taxi, and motorcycle for journeys to work and shopping is high in all study areas -- even those developed prior to the Second World War. As schools are embedded within the residential urban fabric, only journeys to school and childcare show a higher mode share for walking and cycling than for the automobile, although this is not the case in areas developed in the 1980s and 1990s.
2. What are the likely causes of the observed pattern?
The most decisive contributor to lower automobile mode share appears to be the study area's proximity to the metropolitan core and, more directly, location within the City of Toronto. Densities tend be higher and automobile mode shares lower the closer the area is to Toronto's central business district. This supports the general finding in the literature of a negative relationship between density and automobile use.
Interestingly, the degree of local-area use mixture, expressed as the contribution of jobs density to combined population-plus-employment density, appears to have little
effect (see Fig. 41). The Oshawa study areas contain a high proportion of jobs relative to resident population (about 0.68 jobs per resident), yet their automobile mode shares are among the highest. Riverdale and Leaside have the lowest automobile mode shares, yet their jobs-population balance is more modest (0.28 and 0.37, respectively). Moreover, no definitive relationship was found between neighbourhood accessibility and mode share. Location within the metropolitan area appears to be more decisive than higher mix of use and greater accessibility within the cases.
The literature suggests that automobile use is increased by segregation of employment uses at the metropolitan scale, which places jobs and shopping opportunities beyond the reach of walking or cycling. In addition, the low-density form of specialized employment districts may reduce the ability to efficiently service them with public transit.
The relatively high transit mode share in the City of Toronto study areas is likely a function of the integrated and frequent service offered by the Toronto Transit Commission, which is of higher quality than transit service in neighbouring municipalities. This supports Miller and Soberman's (2003) contention that the supply of transportation alternatives is important. The integration of transit systems may also play a role. While only 20% of Torontonians work outside the city, 40% of workers in the surrounding regional municipalities cross municipal boundaries, and often transit service districts, to get to work (Miller & Shalaby 2000).
Despite these variations, the automobile has the majority share for most trips throughout the region. This fits with the findings of Dieleman (2002), Ewing and Cervero (2002), Riekko (2005), and van de Coevering and Schwanen (2006), who suggest that socio-economic variables -- principally, wealth and therefore automobile ownership -- play a determining role. In present conditions, the low cost and higher convenience of the automobile trumps all alternatives.
Implications for policy
This discussion leads to two implications for policy. First, the segregation of jobs and many shops and services from residential areas, both at the local and metropolitan regional scales, promotes reliance on the automobile. But even if all shopping and services were fully integrated into the residential fabric of future subdivisions, most shopping trips would likely continue to be links in automobile trip chains -- that is, intermediate stops on the way to and from work. To reduce the length and frequency of trips, destinations of all types -- jobs and shopping) -- must be concentrated in highly accessible nodes or radically decentralized into residential neighbourhood areas. Both nodal development and greater mix of use are encouraged in the Growth Plan and municipal plans and policies. As discussed in Filion (2007), however, there are many barriers to the creation of nodes that change travel behaviour at the metropolitan region scale. At the same time, a return to a prewar pattern in which mixed-use, small-format employment and retail predominates is unlikely.
In the face of contemporary socio-economic realities -- high levels of low-cost automobile ownership and shrinking households -- a combination of both policies may produce a significant but not transformative shift in travel behaviour. Both IBI Group's (2003) and Riekko's (2005) models indicate that major change would be needed to produce a substantive redirection of transportation behaviour.
Second, it is not possible to reduce the number and length of trips and increase the propensity to walk, cycle, or use public transit solely by manipulating urban form at the local and metropolitam regional scales. The relative supply, cost, and quality of transportation alternatives must also be addressed. If transit is to be a competitive alternative to the automobile, it must be convenient, frequent, integrated, and competitively priced, and must efficiently connect people to destinations. The regional integration of and investment in public transportation systems heralded by the launch of Metrolinx (formerly the Greater Toronto Transportation Authority) and the announcement of multi-year capital funding for transit projects are important first steps.
Notes
17. The TTS collects the travel behaviour of every household member over the age of 11 for the preceding weekday. As a result, it does not capture shopping trips that may occur on weekends. Respondents may also be more likely to recall trips by automobile than those made on foot or by bicycle.