Does Accessibility Require Density or Speed?

Abstract

Problem, research strategy, and findings: Advocates of accessibility as a transportation performance metric often assert that it requires higher density. Conversely, traditional transportation planning methods have valued speed per se as an indicator of success in transportation. In examining these claims, we make two methodological innovations. The first is a new intermetropolitan gravity-based accessibility metric. Second, we decompose the impact of density on accessibility to highlight the distinct opposing influences of speed and proximity in a manner that illustrates different families of relationships between these two factors. This reveals that denser metropolitan regions have slower travel speeds but greater origin-destination proximity. The former effect tends to degrade accessibility while the latter tends to enhance it. Despite theoretical reasons to expect that the speed effect dominates, results suggest that the proximity effect dominates, rendering the denser metropolitan areas more accessible.

Takeaway for practice: Having destinations nearby, as when densities are high, offers benefits even when the associated congestion slows traffic. Where land use policy frequently seeks to support low-development densities in part in an attempt to maintain travel speeds and forestall traffic congestion, our findings suggest that compact development can often improve transportation outcomes.

Research support: Environmental Protection Agency project RD-83334901-0, FHWA Cooperative Agreement Number: DTFH61-07-H-00037, and the Graham Environmental Sustainability Institute at the University of Michigan.

Keywords:

• accessibility
• mobility
• speed
• proximity
• transportation planning

Notes

1. In other contexts, accessibility focuses on the needs of people with disability. The concept is used more broadly here.

2. Complete data required for the calibration of a regional β include zone-to-zone travel times and trip flows for home-based work trips. Of the 38 MPOs, 22 provided either none or incomplete trip flow data.

3. The 16 metropolitan regions are reasonably representative of the full set of 38 metros in terms of geography, density, and population: Bridgeport-Stamford, Chicago, Cincinnati, Dallas, Detroit, Hartford, Indianapolis, Los Angeles, Minneapolis–St. Paul, Philadelphia, Phoenix, Portland, San Diego, San Francisco, Seattle, and Washington, DC.

4. As explained more fully in Grengs et al. (2010), the transformation is accomplished by calculating a z-score for each value in a zone-to-zone travel speed matrix from metro A. This z score matrix is then applied to the mean and standard deviation of speeds from Metro B to transform the speed distribution of Metro A into that of Metro B.

5. The R ² here is the product of the R ² values from the two structural equations as shown in the path diagram: 0.26 × 0.37 = 0.11. That this relationship is weak is not surprising given the complexity with which traffic congestion is produced (Downs, 1992).

6. The speed limit is an aggregate of all types of roads, including highways and local roads, and is intended to capture the effect of roadway speed limits as they contribute to the average transportation speeds of a metropolitan region.

7. In the widely used gravity model formula in travel demand forecasting (United States Department of Transportation, Federal Highway Administration and Urban Mass Transportation Administration, 1977), the number of trips between two areas is determined by the productions of the origin area and the attractions of the destination area, weighted by the impedance between the two areas. The productions and attractions are estimated based on household income and travel information collected. Here, in this context, we simply use the population of the origin area to approximate the productions and use the job of the destination area to approximate the attractions. Since the imputed travel volume share here is only used as a weighting factor of travel speed, such approximation, although not as accurate as what the travel demand forecasting procedure yields, should not yield any significant error.