3. Sustainable Transportation and Urban Form: Principles and Evidence Base
3.3 Urban Form and Sustainable Travel Relationships: a Review of the Empirical Evidence
3.3.4 Micro-Scale Analysis of Travel Patterns, Accessibility and the Built-Environment
While city comparison studies provide a useful broad overview of urban travel patterns, the approach does not allow the consideration of heterogeneity within cities, or the travel behaviour decisions of individuals. Disaggregate analysis is required to analyse this variation. At a basic level it is clearly apparent that travel patterns vary spatially at intra-urban scales. Suburban residents are typically frequent car users, whilst inner-city populations are more likely to travel by public transport and non-motorised modes. We would expect correlations to exist between intra-urban form and travel patterns. As with the previous section however, the challenge is to explain the complex cross-correlations that exist between built-environment, accessibility and socio-economic variables and highlight the most significant factors in these relationships. Disaggregate studies have the advantage that more detailed measures of urban form, land use and accessibility can be considered, and in micro-level studies socio-economic variables can be controlled for at the level of the individual trip maker. This section discusses micro-scale individual studies and the following section considers meso-scale intra-urban analysis.
The evidence base from disaggregate analysis is extensive, with studies coming to mixed and conflicting conclusions on the importance of urban form in travel behaviour (Badoe and Miller, 2000; Banister, 2005). Studies vary in scales of analysis employed, statistical modelling methods used, the areas of study and the variables included. The latter issue of variable choice is important as the high degree of cross-correlations in urban dimensions can lead to false
correlations where significant variables are absent from studies. We mentioned in the previous section the importance of accessibility and socio-economic variables in travel behaviour, and disaggregate analysis presents an opportunity to model these dimensions alongside built-environment variables. An overview of potential variables that could be included is presented in Table 3.4. The
independent variables are classified into socio-economic, urban form/land use, and accessibility/transport supply factors. Accessibility bridges between physical geography and socio-economic perspectives by considering the travel costs for populations of reaching urban facilities. Multivariate modelling that includes these factors can assess their relative significant in influencing travel patterns, and answer key questions such as whether socio-economic factors are the dominant consideration in travel patterns; and whether urban form
relationships are a result of accessibility influences.
Table 3.4: Overview of Variable Types in Disaggregate Urban Form -Travel Behaviour Studies.
Independent Variables b) Private Transport ownership
i) Driving licence
ii) Number of cars / car availability iii) Cycle availability
c) Employment i) Income
ii) Employed / unemployed iii) Occupation type d) Behavioural
i) Personality type / attitudes ii) Cultural factors, trends
2) Urban Form / Land Use
a) Residential/trip origin measures i) Density population / floorspace ii) Housing type / car parking
b) Non-residential/trip destination measures i) Density employees / floorspace ii) Building type / car parking c) Land use measures
i) Function
ii) Mix-of-uses / diversity / jobs-housing balance d) Design
i) Street layout
ii) Pedestrian provision, severance etc.
iii) Cycling provision iv) Transit integration
3) Accessibility / Transport Supply a) Potential measures
i) Automobile
(1) Regional accessibility
(a) Origin access to facilities/opportunities (b) Destination access to facilities/opportunities (2) Local accessibility / Connectivity
(a) Origin parking availability/cost (b) Destination parking availability/cost ii) Transit
(1) Regional accessibility
(a) Origin access to facilities/opportunities (b) Destination access to facilities/opportunities (2) Local accessibility / Connectivity
(a) Origin access to transit station / services (b) Destination access to transit station /
services iii) Walking / Cycling
(1) Regional accessibility
(a) Origin access to facilities/opportunities (b) Destination access to facilities/opportunities (2) Local accessibility / Connectivity
(a) Origin access to pedestrian routes, cycle lanes
(b) Destination access to cycle parking b) Journey specific
i) Generalised cost/ travel time of journeys by available modes.
Dependent Variables 4) Travel Behaviour
a) Work travel / non-work travel / all travel i) Trip Frequency
ii) Distance, Time iii) Mode-choice
The dependent travel behaviour variables that models try to predict are a further key variable choice. It is common to analyse frequency, distance and mode-choice separately, either in isolation or through multiple linked models (Ewing and Cervero, 2001). Additionally work travel and non-work travel can be modelled separately. As discussed previously, trip frequencies are generally a product of socio-economic factors, therefore urban form studies generally focus on distance, mode-choice and car ownership as the dependent variables, or combinations of these such as total vehicle miles travelled. The number of independent and dependent variables in Table 3.4 is high, and the list is by no means comprehensive. Research over the last fifteen years has begun to model a more comprehensive range of these variables and control for the most
significant factors in travel behaviour (Ewing and Cervero, 2010). A consensus has emerged around a relatively weak, but not insignificant, influence for urban form factors once socio-economic and accessibility factors have been controlled for.
Controlling for socioeconomic factors is required to provide a rigorous methodology for the analysis of the built-environment and travel patterns.
Current travel demand modelling methods (e.g. activity based models) have been developed at the disaggregate level of the individual trip-maker to include the diversity of behavioural responses which occur amongst different types of people. Generally the most significant socio-economic factor in influencing travel patterns is car ownership (Banister, 2005; Cervero, 1996b). Car owners invest in their vehicles financially (with purchase costs greatly exceeding running costs in current ownership structures) and to a varying extent behaviourally and psychologically, and therefore make use of their cars once purchased. Non-car owners in contrast are clearly much more restricted in terms of car availability and subsequently use. The decision to own a car is in turn interrelated with residential and workplace location decisions, as well as individual and household socio-economic factors. Households owning fewer cars tend to drive less and use public transport and non-motorised modes more often. One of the simplest means of considering socio-economic factors is to include car ownership as an independent variable. This approach does not attempt to understand the dynamics between car ownership preferences,
residential location and subsequent activity/travel decisions. Alternatively car ownership can be modelled as an endogenous function, linked to other socio-economic factors and to travel patterns (Chen et al., 2008).
The connection between travel preferences and residential location is an important consideration. Populations to an extent choose housing locations based on the lifestyles they wish to lead (Kitamura et al., 1997), and travel patterns are a component of these lifestyles. Therefore it is possible for relationships between urban form and travel behaviour to work both ways: i.e.
populations do not necessarily choose to travel a certain way because of where they live, they can choose where to live depending on how they want to travel (a process known as residential self-selection). Therefore personal attitudes
towards lifestyles acting through residential location can lead to correlations between urban form and travel patterns. While it is possible to include socio-economic variables in micro-level studies to control for such effects, more behavioural and attitudinal aspects that can effect travel patterns are rarely included in built-environment studies (Kitamura et al., 1997). Some caution must be taken on this issue however as it is possible for the importance of personal attitudes to be overstated. Two-way relationships between attitudes and behaviour are clearly part of human nature. The international comparison discussion in Section 3.3.3 clearly illustrates shared city-wide urban cultures are shaping individual attitudes towards transport modes, as for example in the cycling cities of Copenhagen and Amsterdam.
After controlling for socio-economic considerations, studies can then gauge the influence of accessibility and urban form factors. The following discussion considers the various accessibility and urban form measures that are possible.
Accessibility describes the opportunities available to a population in a specific location depending on travel cost. As discussed previously, it has been argued that the built-environment influences travel patterns by influencing
accessibility. If this is the case, then studies that include both accessibility and
built-environment variables should find a significant role for accessibility variables, and low or insignificant relationship for built-environment variables1. Accessibility measures can be classified as general potential measures, which consider opportunities to a range of populations/facilities depending on travel cost, or alternatively can be journey specific measures, which consider the travel costs of a particular journey (typically by several modes). The latter type relates only to studies where the origin and destination of a trip is known and is used for mode-choice modelling.
Accessibility measures can be calculated from the perspective of trip origins (typically residences) and trip destinations (typically activity/employment centres). This distinction also applies to urban form and land use measures.
Badoe and Miller (2000) note that there is a strong tendency in theory and practice to focus on the residential side of land-use transportation relationships, while the trip-end side may have a more direct relationship with travel
behaviour and furthermore be more susceptible to successful planning measures. The preoccupation of the transit-orientated sustainability literature with residential density and neighbourhood design may then be limited if the trip destination context is more significant (Ewing and Cervero, 2001).
The scope of accessibility measures can vary depending on trip purposes and modes. The distinction between regional and local accessibility measures is one means of defining scope (Handy, 1993). Regional accessibility covers the activity space of common medium distance journeys such as commuting, comparison shopping and leisure activities; while local accessibility refers to facilities accessible within walking distances. Local accessibility measures are essentially connectivity measures, for example a measure of local accessibility to public transport services is a measure of transit connectivity, and can be
1 Note it is not always straightforward to differentiate accessibility and built-environment measures. For instance street network design measures essentially consider both physical structure and pedestrian accessibility simultaneously.
considered for both trip origins and destinations. Regional accessibility relates to the position of a residential district in relation to accessing major
employment, population and services centres. An example of a regional accessibility measure is jobs within 45 minutes travel time. Where detailed accessibility data is unavailable, proxies are often used such as distance to the city centre (the calculation of more accurate network accessibility measures is discussed in Section 4.7). Regional accessibility is considered to be the most important geographical factor for car owners on total vehicle miles travelled (Ewing and Cervero, 2010). Effectively this conclusion follows the common sense logic that residents in more remote locations need to travel further to access facilities. This conclusion indicates that focussing on local design without considering the regional context is a flawed approach to sustainable travel planning. Additionally parallels can be drawn with studies that show the decline in vehicle miles as settlement sizes increase (ECOTEC, 1993), as larger settlements support a greater scale of services and thus have higher regional accessibility. Local accessibility can also have a more modest impact on travel distances, by reducing trip lengths for purposes such as convenience shopping (Handy, 1995).
Once socio-economic and accessibility variables have been included, then the influence of urban form variables independent of these factors can be assessed.
There are a range of built-environment measures that can be used, including density (both of residents, employees, and built form), land use mixes and street design. (Note there are a wide range of spatial analysis issues that occur in the measurement of these variables as discussed in Section 4.4). There is no universal consensus on the importance of these variables on travel patterns, though some trends are evident from the comparison of studies. Residential density has been a focus of sustainable travel studies following Newman and Kenworthy‟s (1989) research and debates around compact city policies. The empirical evidence at disaggregate levels is very mixed. A number of studies have found an association with higher densities and more sustainable mode-choices (Cervero, 1996b; Frank and Pivo, 1994) but generally these studies have not fully accounted for socio-economic and accessibility factors. The role
marginal in many studies once other socio-economic and accessibility factors are accounted for (Badoe and Miller, 2000; Ewing and Cervero, 2010).
Densities can also be measured at trip destinations, typically in the form of employment densities for workplaces. It has been argued that findings show that increased employment concentrations have significant impacts on more
sustainable mode-choice (Badoe and Miller, 2000). On the other hand, it is likely that destination-based accessibility measures will also account for much of this variation. Additionally one of the most important factors in determining private vehicle accessibility is car parking cost and availability. This is seldom modelled as it is likely to be closely connected to destination density.
One means of quantifying the relative importance of the many variables is to calculate elasticities from multivariate modes. These describe how the dependent variable in the model would respond to a 1% change in the independent variable. Ewing and Cervero (2010) conducted a meta-analysis combining many research studies of the influence of the built-environment on vehicle miles travelled as shown in Table 3.5. The results point to regional accessibility variables (job accessibility by auto and distance to downtown) having the strongest elasticities. The relatively high values for land use mix and street network measures also indicate that local accessibility could play an important role. The values for density measures are notably low. The meta-analysis approach is a useful means of summarising relationships. There are some shortcomings as significant variation in elasticity values exists between the studies used to form the weighted averages. Furthermore the number of studies using advanced residential self-selection techniques is low and the elasticity approach cannot capture the potential synergistic and non-linear relationships that potentially exist in travel pattern relationships.
The analysis of mode-choice rather than vehicle miles travelled can produce somewhat different results than analyses of total vehicles miles. A recent study by Chen et al. (2008) of commuting mode-choice in New York, notable for including a range of accessibility measures and controlling for self-selection, found several built-environment and accessibility variables were significant in predicting private vehicle commuting, including employment density at work,
connectivity to transit at both home and work, job accessibility at work by transit and commute travel time and cost. It is likely the importance of
workplace density is connected to car parking availability/cost, which was not included in the study (Chen et al., 2008).
Table 3.5: Weighted Average Elasticities of VMT with Respect to Built-Environment Variables.
Source: Ewing and Cervero (2010).
Total number
Intersection/street density 6 0 -0.12
% 4 way intersections 3 1 -0.12
Overall, it is clear that while relationships between the built-environment and travel patterns are complex and an ongoing research area, a more in-depth understanding of relationships is possible with micro-scale analysis. This research summary indicates that socio-economic variables, in particular acting through car ownership and residential location, are major drivers of travel demand. Of the built-environment related factors, it is accessibility measures that have been found to have the strongest relationships. Regional accessibility has a significant influence on total vehicle miles travelled, while high local accessibility through mix-of-uses and pedestrian focussed streets can also have an impact on travel distances. Correlations with density are largely a product of accessibility factors. For mode-choice analysis, accessibility factors at the trip end may be more significant than trip origin measures. This includes transit connectivity measures and employment density, which is related to car parking costs. Overall both a regional and a local perspective is required to encourage sustainable travel patterns, considering trip origins and destinations and the connections between localities and their regional context.
3.3.5 Meso-Scale Analysis of Journey-to-Work Patterns