![Sample our Built Environment journals, sign in here to start your access, Latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5926/TOC-Subject-Sample-2021-Built-Environment.jpg"})
Abstract
Problem: Congestion pricing and land use planning have been proposed as two promising strategies to reduce the externalities associated with driving, including traffic congestion, air pollution, and greenhouse gas emissions. However, they are often viewed by their proponents as substitutive instead of complementary to each other.
Purpose: Using data from a pilot mileage fee program run in Portland, OR, we explored whether congestion pricing and land use planning were mutually supportive in terms of vehicle miles traveled (VMT) reduction. We examined whether effective land use planning could reinforce the benefit of congestion pricing, and whether congestion pricing could strengthen the role of land use planning in encouraging travelers to reduce driving.
Methods: VMT data were collected over 10 months from 130 households, which were divided into two groups: those who paid a mileage charge with rates that varied by congestion level (i.e., congestion pricing) and those who paid a mileage charge with a flat structure. Using regression models to compare the two groups, we tested the effect of congestion pricing on VMT reduction across different land use patterns, and the effect of land use on VMT reduction with and without congestion pricing.
Results and conclusions: With congestion pricing, the VMT reduction is greater in traditional (dense and mixed-use) neighborhoods than in suburban (single-use, low-density) neighborhoods, probably because of the availability of travel alternatives in the former. Under the same land use pattern, land use attributes explain more variance of household VMT when congestion pricing is implemented, suggesting that this form of mileage fee could make land use planning a more effective mechanism to reduce VMT. In summary, land use planning and congestion pricing appear to be mutually supportive.
Takeaway for practice: For policymakers considering mileage pricing, land use planning affects not only the economic viability but also the political feasibility of a pricing scheme. For urban planners, congestion pricing provides both opportunities and challenges to crafting land use policies that will reduce VMT. For example, a pricing zone that overlaps with dense, mixed-use and transit-accessible development, can reinforce the benefits of these development patterns and encourage greater behavioral changes.
Research support: This project was supported by the Mineta Transportation Institute, where the authors are research associates.
Acknowledgments
The authors thank the Mineta Transportation Institute National Transportation Finance Center for funding and staff support for this research. In addition, thanks are due to James Whitty, Betsy Imholt, and Don R. Crownover at the Oregon Department of Transportation for providing data; to Anthony Rufolo of Portland State University for answering many questions; James Heckman, William Greene, Rajeev Dehejia, and Francis Vella for advice on selection bias; Kory Kroft for advice on behavioral economics; Sarah Oliver, Melissa Reese, and Megan Quirk for research assistance; and finally to three anonymous reviewers for their constructive comments.
His research interests include public transit, pedestrian behavior, parking policy, and road pricing.
Her research interests include transportation finance, planning for pedestrians and bicyclists, urban street design, and transportation history.
Her research interests focus on individual and organizational decision making related to travel behavior, land use, the environment, and health.
Notes
1. Developing assumptions and parameters from past experience for the land use planning and congestion pricing comparison is challenging. Even if they are correct, past experience may not fit in the future development, especially when there is a paradigm change in the policy field. Scenarios are not forecasts of the future, but alternative paths given certain conditions (Hopkins & Zapata, Citation2007). Many times, these scenarios “represent extreme ends of the policy-implementation spectrum” (C. Rodier et al., 2009, p. 2), and have limited applicability in the real world. For example, in the Dortmund and Naples studies, the land use planning strategy designated all new residential units over a 20-year period as transit-oriented development. In some of the congestion pricing strategies, car operating costs are increased by 100%, cordon pricing is set up to equal 60 minutes time value, and parking charge is set to equal 60 minutes time value in the city center (Lautso et al., Citation2004). Sometimes these multiple pricing policies are combined into one scenario, which might be too aggressive to be feasible in reality, at least in the short term (C. Rodier et al., 2009).
2. Synergy is defined as: Benefit (A 1 B). Benefit A 1 Benefit B. In this article, we treat complementary and synergy differently.
3. The research by Rufolo and Kimpel (Citation2009) differs significantly from this study in terms of unit of analysis, research design, and the research objectives. For example, one key difference is that we compared the VMT reduction between two household groups in order to control for the seasonal travel pattern.
4. It is well known that VMT in winter is generally less than VMT in summer, so we could not assume any change in VMT between Phase 1 and Phase 2 was due solely to the mileage fee program. Data from nine traffic monitoring stations on the main highway intersections in the Portland metro area indicated that the average monthly VMT in the five months of Phase 2 was 7% lower than that in the five months of Phase 1 (ODOT, Citation2010). The Automatic Traffic Recorder Stations are 26-004, 26-002, 26-014, 26-015, 26-016, 26-018, 26-022, 26-024, and 26-027.
5. Only two other states have done similar experiments. A Seattle, WA, program tested hypothetical charges on selected expressways (Puget Sound Regional Council, Citation2008), and a Minneapolis, MN, program added the mileage fee as a supplement to (not replacement for) the gas tax (Abou-Zeid, Ben-Akiva, Tierney, Buckeye, & Buxbaum, Citation2008). To date, the Oregon program represents the most sophisticated and realistic mileage fee design in the United States. Another nationwide pilot program, currently being conducted by the University of Iowa, has a much larger sample size and covers six metropolitan regions, but it does not have a pricing component (Public Policy Center, Citation2011).
6. The Portland program cost about $3 million for less than 200 households, almost 80 times more expensive than a typical household travel survey at a per-household base. Other mileage fee pilot programs have a similar level of cost per household.
7. In this research, we defined land use broadly, equivalent to the terms “urban form” or “built environment” used in other studies. Here, urban form and land use are exchangeable.
8. The other variables tested include (under access) distance to the nearest frequent bus stop, number of total bus stops within a buffer, and distance to closest freeway exit; (under density) population density, job density, etc.; (under mixed use) share of seven land use types within a buffer, and square footage of commercial space within a buffer; (under street pattern) street density within a buffer, density of four-way intersections within a buffer, average block size, and bike lane length within a buffer (Guo et al., Citation2010). Most of these variables were tested under multiple measures, for example, network distance versus dummy variables, or various buffer sizes (0.25, 0.5, 1.0, and 2 miles). These variables were not included because they were either insignificant in all models, unstable due to a small number of observations, or highly correlated with each other. We did not use factor analysis to consolidate all of these variables because we decided to keep the estimation results straightforward and easy to interpret.
9. The assignment process was as follows. First, households were ranked according to their Phase 1 peak hour mileage in Zone 1, per vehicle. Going down the list, three households were assigned to the peak-charged group and one to the flat-rate group until half the households on the list were in the peak-charged group. The remaining households were assigned to the flat-rate group (Rufolo, 2010). This design was a trade-off between the objective of keeping a sufficient number of households in the peak-charged group who would travel in peak hours within the UGB, on the one hand, and the objective of decreasing the selection effects, on the other (Rufolo & Kimpel, Citation2008). A similar approach has been used by Arentze and Timmermans (Citation2006) in their stated preference study on job and household relocations under different pricing schemes. A comparison of the two groups in the Portland pilot shows that the peak-charged group contains more households with a full-time employee, but there is no statistical difference between the groups (at the 5% level) in any other characteristic (number of drivers, vehicles, children, and adults, income levels, and all eight program design and land use attributes). This assignment process generated a nonrandom assignment of households to the two groups. However, we believe that the problem is not as severe as it might appear, because the households did not self-select into groups. Had they done so, their preferences would have become a missing variable that correlated the base VMT to the error term. In the Portland program, households did not know the selection rule beforehand, so this assignment process did not present a self-selection problem. Therefore, including the Phase 1 (base) VMT in the models and estimating a percentage change model could largely control for the effect of this assignment process.
10. Although the peak-charged households tend to have slightly more full-time workers and drive more than the flat-rate households in Phase 1, we could not find a good theoretical explanation of why these factors, instead of congestion pricing, should explain the different contributions of urban form variables between the two groups. However, we understand that findings based on this analysis are limited due to the sample size and selection process. A larger, more random dataset is necessary for more robust results.
