![[logo] US EPA](https://webarchive.library.unt.edu/eot2008/20081107093955im_/http://www.epa.gov/epafiles/images/logo_epaseal.gif){kind=logo}
Final Report: An Integrated Framework for Estimating Long-Term Mobile Source Emissions Linking Land Use, Transportation and Economic Behavior
EPA Grant Number: R831450Title: An Integrated Framework for Estimating Long-Term Mobile Source Emissions Linking Land Use, Transportation and Economic Behavior
Investigators: Harrington, Winston , Nelson, Peter , Safirova, Elena , Shih, Jhih-Shyang
Institution: Resources for the Future
EPA Project Officer: Bloomer, Bryan
Project Period: January 1, 2004 through July 31, 2007
Project Amount: $749,080
RFA: Consequences of Global Change for Air Quality: Spatial Patterns in Air Pollution Emissions (2003)
Research Category: Global Climate Change
Description:
Objective:The purpose of this project was to develop a computable general equilibrium (CGE) model of an urban economy that links land use, transportation and other economic decisions at the firm and household levels and use that model to project mobile source emissions over a 50-year time horizon under a variety of economic and technological assumptions. The model we proposed had five submodels: a strategic transport model, a regional CGE model, a vehicle choice model, an emission model, and an air quality model, linked as shown in Figure 1.
Figure 1. Model Integration
We succeeded in developing and calibrating LUSTRE, a CGE model of the Washington metropolitan area capable of estimating the changes in mobile source emissions from a variety of policy initiatives and economic scenarios. It has a moderate level of spatial disaggregation—40 zones, much less than local transport planning models, but much more than most CGE models. Four industries are represented (agriculture, manufacturing, construction and retail), and four classes of worker/consumers, differentiated by skill level (and income). Both consumers and industry agents choose a mix of discrete and continuous variables to maximize utility or profits. For example, consumers choose their zones of residence and employment and their preferred housing type (detached or multi-family), and conditional on those choices, choose hours of work and goods purchases, which require shopping trips. Economic choices determine the demand for travel endogenously, and for each trip the traveler chooses time of day, destination, route, and mode. LUSTRE is one of the very few regional CGE models and, as far as we know, the only one with the following characteristics:
- Multiple classes of worker agents, who may choose work or unemployment
- Endogeneity of locally determined economic variables, including prices, rents, wages, and quantities of goods purchased, land and square footage of housing consumed, and hours worked. Essentially, the only exogenous variables are prices wholly determined in national markets (e.g. interest rates), initial endowments, local infrastructure, and government policies.
- A spatially disaggregated, multimodal transport network. The network contains a mix of stylized links representing ordinary street travel in the zones and a set of “special links Modes explicitly represented include single-occupancy vehicle (SOV), multiple occupancy vehicle (HOV), rail travel, bus travel, and non-motorized transport. Trip types represented include work trips, shopping trips, other personal trips, and freight. Both the vehicle and transit modes are congestible.
- A spatially disaggregated housing and real estate including explicit representations of construction and demolition.
LUSTRE falls short of the model described in the initial proposal in several ways:
- The agents of the model are individual workers, not households. A household model would be richer and more realistic because, for one thing, it would allow two workers to live in the same household, and residential choices would have to take into account two journeys to work.
- Without a household-based model, we were unable to model vehicle choice explicitly. Two person households would be able to choose between ownership of 0, 1 or 2 vehicles. Explicit modeling of the vehicle choice decision would allow to model response to a set of vehicle ownership policies that are now beyond our capability.
- The model is not dynamic; it is a comparative-static model. In any experiment using the model, we perturb the baseline configuration by changing one of the exogenous variables, then we run the model until it returns to equilibrium, or the point at which all markets clear. We observe the differences between the newly determined equilibrium and the baseline equilibrium. But markets do not clear at the same rates, in LUSTRE the transportation “market” clears daily while the real estate market could take decades.
In addition, we did not link the model to the CMAQ air quality model as promised in the proposal. This step was not required by the RFA, and in the event we needed to reallocate the resources that would have been devoted to the air quality model to the completion of the base model. We are confident, however, that LUSTRE could be linked to an air quality model if desired.
The failure to model households was due primarily to the great expansion in the size of the problem that that step would have represented. We found that a household level model would add so many endogenous variables, many of which are discrete, that we would not be able to fit the model on our existing hardware, it is not clear it would converge in any practical length of time. In future work we will look for ways to simplify other decisions in order to free up resources to accommodate household modeling. A more reduced-form model of shopping decisions is a leading possibility. In addition, we experimented with an explicit model of households in an more spatially aggregated model consisting of three zones.
Despite its limitations LUSTRE has proven to be very useful for policy analysis. Although each decision by each agent is simple and comprehensible, their combined effects and generate surprises and provide new insights into the consequences of policies. For example, it is a conclusion of simple models of local transport that modes should be priced at their marginal cost, and in particular that peak period pricing should be employed. However, we found that implementing peak period pricing on Metrorail has an adverse effect on welfare, because it drives transit riders back onto the roads. Indeed, we found that from a social welfare perspective, taxing everyone and using the revenues to subsidize transit riders has positive
benefits. For another interesting and surprising result, consider a policy of imposing congestion tolls and using the revenues to subsidize transit. This pair of policies is thought by many, both advocates and academics, to be “synergistic”: the gains generated by implementation of both policies together exceeds the sum of each implemented without the other. A so-called “virtuous circle” generates this happy outcome, for congestion fees induce a shift in rush-hour demand from highways to transit, and the greater investment in transit increases not only capacity but service levels, as waiting time is reduced. Increased service levels in turn draw even more commuters to transit. This, at any rate, is what is often predicted by simple models of transit and vehicle transport, which tend to represent road travel by a single link and transit on a separate and parallel link. In our model approximates a real linked road and transit network more closely than the simple models used to generate these expectations, we do not see a virtuous circle. We see at best that the combined effect is additive, and in some scenarios slightly subadditive. Among the more realistic features of our model that seem to account for this result are these: (i) bus transit does not have its own network in the model but competes for road space with other vehicles and (ii) congestion pricing is only implemented partially, on some or all freeways. Much of the policy explorations we have done with the model has been to examine the consequences of congestion pricing in a more realistic setting than it is usually examined.
The transportation component of the model START has been used to conduct policy simulations of gasoline taxes (Nelson et al. 2003), HOT lanes (Safirova et al. 2003), and congestion pricing (Safirova et al. 2004; Safirova et al. 2005), as well as to compute network-based marginal costs of transportation (Safirova et. al 2007) and evaluate the benefits of public transit (Nelson et al. 2007). After START and RELU were integrated into LUSTRE, the integrated model was used to test long-term effects of congestion pricing (Safirova et al. 2006a, 2006b), investigate energy implications of policies affecting spatial development (Safirova et. al. 2007) and compare the effects of an array of transportation policies on emissions outcomes (Harrington et. al 2007). The integrated model also has been used to look at economic and transportation implications of land-use policies (Coleman et al. 2007a; b)
Journal Articles on this Report: 4 Displayed | Download in RIS Format
| Other project views: | All 42 publications | 6 publications in selected types | All 4 journal articles |
| Type | Citation | ||
|---|---|---|---|
|
|
Fischer C, Harrington W, Parry IH. Should automobile fuel economy standards be tightened? Energy Journal 2007;28(4). |
R831450 (Final) |
|
|
|
Nelson P, Baglino A, Harrington W, Safirova E, Lipman A. Transit in Washington, DC: current benefits and optimal level of provision. Journal of Urban Economics 2007;62(2):231-251. |
R831450 (Final) |
|
|
|
Safirova E, Gillingham K, Harrington W, Nelson P, Lipman A. Choosing congestion pricing policy. Transportation Research Record 2005;1932:169-177. |
R831450 (2004) R831450 (Final) |
|
|
|
Safirova E, Gillingham K, Houde S. Measuring marginal congestion costs of urban transportation: do networks matter? Transportation Research Part A: Policy and Practice 2007;41(8):734-749. |
R831450 (2006) R831450 (Final) |
|
, Air, Scientific Discipline, RFA, Air Quality, climate change, Ecological Risk Assessment, Air Pollution Effects, Atmosphere, Environmental Chemistry, Economics, mobile sources, human activities, emissions measurement, engine exhaust, Global Climate Change, land use, air quality models, environmental monitoring, vehicle emissions
Progress and Final Reports:
2004 Progress Report
2005 Progress Report
2006 Progress Report
Original Abstract