Reducing the total number of vehicle kilometres travelled (vkt, sometimes abbreviated VkmT) in a city is a new objective of land use planning that addresses not only the cost of transportation infrastructure, but also the emission of greenhouse gases.
There is a small but statistically significant effect of urban form on transit use. What urban form around a transit station will minimize vehicle use? There is widespread agreement that residential density around transit stations should be higher than otherwise, but what housing mix is best? Some Canadian cities, like Edmonton, Calgary, Ottawa, and the Golden Horseshoe, plan for high density housing to support transit, while the Congress for the New Urbanism calls for a range of housing types.
Effect of Housing Mix
On average, increasing the dwelling density of a neighbourhood tends to decrease both the number and length of automobile trips per household in that neighbourhood, as does decreasing the distance from a transit station. However, increasing residential density, especially around transit stations, may not decrease the total vkt for the city as a whole, because of some self-selection effects that may actually increase vehicle use elsewhere in the city.
There is such a thing as too low a density for Transit-Oriented Development (TOD). For any given mode of public transit, there is a minimum transit-supporting density, a residential density below which the transit system does not get enough customers to justify frequent service. However, there is also a density beyond which the decreased vkt of people near the transit station may be offset by increased vkt of people further from the station.
In general, when the residential density is higher, people drive less. But we also know that when comparing neighbourhoods of different densities, low-density neighbourhoods tend to be dominated by single-family houses, while high-density neighbourhoods tend to be dominated by apartment buildings. How much of the decrease in driving that we see in higher-density neighbourhoods is attributable to a difference in housing mix? In other words, do people who live in higher density actually drive less, or does higher density simply concentrate in one place more people who tend to drive less, without reducing anyone’s driving? This effect, where local changes in transportation behaviour are influenced by where people choose to live, is known as “residential self-selection”. The demographic profile attracted by higher-density TOD is quite different from the general population.
To help elucidate whether density actually reduces driving, we can graph the total amount of driving done per person against the local housing density (see figure), separating detached houses plus townhouses in one line, and apartments plus duplexes in the other. The result is surprising and very important.
Using data from the 2001 National Household Travel Survey (NHTS) carried out in the U.S., on average, people tend to use their personal vehicles less as the density around them increases. Broken down by housing type, both on a per-household and on a per-person basis, people who live in single houses, townhouses and row houses drive significantly less when the local density around their house is higher. However, this general rule does not apply as much to apartment and duplex dwellers. They use their vehicles less on average, but the number of kilometres they drive does not seem to vary as much with density. At higher density, apartment dwellers drive as much as house dwellers if not more. This graph excludes rural areas and New York City, whose high density lowers the average a bit. Density is at the block-group level. When using census-tract level densities, the difference between the two groups is less pronounced.
Figure 1: People who live in houses drive less if the house is in higher density, but apartment dwellers are less sensitive to neighbourhood density. Plotted against the lowest densities in the range.
This graph points to an interesting conclusion. Whether apartments are located in high or in low density areas has a smaller than average effect on the vehicle use of apartment dwellers. The location of houses, on the other hand, does have a significant effect. A possible conclusion to be drawn from this is that reducing a city’s total vehicle use through TOD is more effectively achieved by locating single houses and townhouses in higher density areas near transit stations.
Consider briefly scenarios at the two extremes: One is to maximize density by putting apartments in the central area and near transit, and putting single houses and townhouses in lower density, further away. The second is to put houses in higher density areas, and apartments in lower density further away. Assuming that people’s preference for one housing type over another is greater than their preference for a neighbourhood density, the difference between scenario 1 and scenario 2 is that house dwellers will drive 8,500 km less per person per year, while apartment dwellers (ignoring the rarer low-density urban apartments) will drive 2,700 km more, making scenario 2 better as long as apartment dwellers are less than 76% of the population.
Of course, scenario 2 is an unrealistic extreme for two reasons: Euclidean geometry and economics. There is less land near the centre and near transit than further away, and the value of that land makes apartments more profitable, all other things being equal. Also, the economics of individual transit stations is such that the raw numbers of people living near the station is what makes it financially viable, and those numbers are most easily achieved through apartment-dominated density and self-selection.
The scenario 1 strategy will show measurable effects in the immediate vicinity of transit stations, with nearby residents driving less, but will likely not reduce total driving city-wide. Unfortunately, house dwellers, whose driving is most dependent on density, on average end up in lower density and further from transit when applying this strategy.
This may help explain why Portland, Oregon, for instance, with its 25 years of light rail investments and priority to multiple-unit buildings, has seen significant positive changes within individual compact neighbourhoods but no improvement in transit ridership overall. The improvements have been in the outer rail corridor, where single family homes pay a premium for transit access. Similar trends hold true in other cities
Apartment buildings should play a secondary role in TOD when the objective is to reduce total kilometres driven. They certainly help to generate good transportation statistics for the immediate vicinity of the transit station, by packing in more self-selected households than would fit otherwise, but apartments are not as effective as houses at reducing the total kilometres driven by the total population.
Of course TOD also has other objectives, including economic viability of the transit service and of new development, which may translate into more apartments. Other non-residential features of successful TOD also contribute to vkt reduction, and apartments may complement them well, for instance above retail.
Some TOD projects try to put in as many units as the market will allow. Instead, a good rule of thumb is to put in as many houses and townhouses as the market will allow, then adding apartments where appropriate. A detailed land use and transportation model would be more useful than rules of thumb, given the interaction with land prices and the household and business relocations that result.
To some degree this rule of thumb runs counter to conventional wisdom, which tends to place high-density apartment buildings near transit stations and particularly near downtown. But since apartment dwellers are not the major part of a region’s total driving, they are also not a major part of the solution.
Martin Laplante, PhD, is Vice-President of RES Policy Research Inc. He can be reached at email@example.com
 Bento, Antonio M., Cropper, Maureen L., Mobarak, Ahmed Mushfiq , Vinha, Katja, The Impact of Urban Spatial Structure on Travel Demand in the United States, Working Paper EB2004-0004, University of Colorado at Boulder, 2004
 Fort Road Old Town Master Plan, City of Edmonton, 2002
 Transit Oriented Development Policy Guidelines, City of Calgary, 2004, amended 2005
City of Ottawa Official Plan, A Component of Ottawa 20/20, the City's Growth
Publication 1-28, May 2003
 Technical Backgrounder - Intensification and Density Targets, annex to Proposed Growth Plan for the Greater Golden Horseshoe, November 2005, Government of Ontario, Ministry of Public Infrastructure Renewal
 Eash, R. Incorporating Urban Design Variables in Metropolitan Planning Organizations’ Travel Demand Models. Proc., Conference on Urban Design, Telecommuting, and Travel Behavior, Travel Model Improvement Program, Oct. 1996.
 Robert Cervero, Michael Duncan, Residential Self Selection and Rail Commuting: A Nested Logit Analysis Working Paper, University of California Transportation Center, December 2002
 Dill, J. (2003) Travel Behavior and Attitudes: New Urbanist vs. Traditional Suburban Neighborhoods, presented at the 83rd Annual Meeting of the Transportation Research Board, Washington, DC January 11-15, 2004.
 Anthony Flint, The Density Dilemma: Appeal and Obstacles for Compact and Transit-Oriented Development, Lincoln Institute of Land Policy Working Paper, September 2005
 Raw NHTS 2001 Version 4.0 data sets and statistical weights provided by U.S. Department of Transportation, Bureau of Transportation Statistics
 Dueker, K. J. and M. J. Bianco (1998) Effects of Light Rail Transit in Portland: Implications for Transit-Oriented Development Design Concepts. Portland : Center for Urban Studies, College of Urban and Public Affairs, Portland State University, Discussion Paper 97-7
 Renne, John L, Thirty Years of Trends in Transit-Oriented Development Across America, Transit Oriented Development - Making It Happen conference, July 2005, Fremantle, Australia
 Kitamura R, Mokhtarian P, Laidet L. A micro-analysis of land use and travel in five neighborhoods in the San Francisco Bay Area Report prepared for the California Air Resources Board, Sacramento; November 1994