The cost of congestion and the benefits of congestion pricing: A general equilibrium analysis
Introduction
We present a spatially detailed empirical general equilibrium treatment of the market and welfare effects of road congestion and of pricing the congestion at its social marginal cost. In the model, consumers and firms adjust to congestion and to its pricing in many margins. The model is static, not dynamic, and hence it cannot determine which adjustments by consumers and firms would happen quickly and which would take longer. Therefore, the model is used to compare the effects of congestion pricing after the markets have fully adjusted to such pricing. We compare market adjustments and welfare gains when the revenue from pricing is not recycled; and then when it is recycled by cutting distortionary taxes, while maintaining aggregate region-wide revenue neutrality from all tax sources including congestion tolls.
Economists have been interested in congestion pricing ever since Vickrey (1963) called for the pricing of road traffic according to the welfare economic principle of Pigou (1932). Since then, congestion pricing continues to become increasingly palatable politically because of the rising cost of congestion in the big cities of the world. London, U.K. (Leape, 2006); Stockholm, Sweden (Eliasson et al., 2009); and Milan, Italy (Rotaris et al, 2010) have introduced congestion pricing in their central areas, joining Singapore and some smaller cities in Norway that have had congestion pricing for many years. For a detailed review of these schemes, see Anas and Lindsey (2011).1
Congestion and congestion pricing have been studied theoretically in urban economics, using the monocentric city model. In this model, it is assumed that all jobs are located downtown (the center of a circular city) and cannot relocate, there is no public transit, all trips are commutes to work, buildings are not durable and, often, all consumers are identical. The earliest theoretical treatment of congestion in the monocentric city model was by Strotz (1965) and one of the most recent by Wheaton (1998). In these and other analyses, the congestion externality is internalized by tolling every mile of the radial commute to the downtown. In such a setting, consumers can blunt the impact of the toll, only by moving to a residence closer to the downtown to reduce miles traveled and tolls paid. Consequently, population densities near the downtown increase and the monocentric city becomes more compact. As a counterpoint to monocentric analysis, Anas and Kim (1996) demonstrated in a theoretical general equilibrium model that congestion and its pricing can cause the emergence of job centers or the decentralization of jobs from the downtown center to the periphery. With high congestion, firms move closer to their workers and customers, reducing their travel times, and benefit by paying lower wages or charging more for output.
On the empirical side, a likely magnitude of congestion pricing was first estimated by Keeler and Small (1977), who calculated optimal user tolls on a sample of suburban and central city highways in the San Francisco Bay Area. They found that the socially optimal rush-hour speeds were considerably higher than the observed speeds. But little is known about how and how much congestion pricing would realistically modify urban structure; how much rents, wages, prices and the location of jobs and residences would change; and how the benefits of congestion pricing would be distributed among consumer groups and sectors of the economy. Modeling these effects within a reasonably complete structural general equilibrium model remained a challenge.2 Some economists have expressed caution even about the relatively limited challenge of calculating congestion tolls on a network of roads. For example:
“…On the practical side, it may be computationally infeasible to estimate marginal congestion costs on every single link and intersection in an urban road network, particularly given that pricing at one point diverts traffic elsewhere within the network, ...” (Parry et al. 2007, page 393).3
In our model, the equilibrium determination of congestion tolls on all links of a road network is rendered feasible, and labor, housing and other real estate, production and travel markets including the road network are treated simultaneously. The spatial distribution of jobs and residences are not predetermined but are interdependent. Industries are also directly interdependent because they exchange intermediate inputs. These interdependencies give rise to pecuniary savings from the proximal locations of producers and their laborers and customers; and also from the proximal location of producers in different industries. These pecuniary savings, therefore, are a form of agglomeration economy despite the assumption of constant returns to scale in production. The current version of the model ignores technological (non-pecuniary) agglomeration economies, also known as Marshallian externalities. If such externalities were included, then the total factor productivity coefficient (scale factor) of the production function would change according to non-pecuniary externalities such as information flows between firms.
Microeconomic links between the travel decisions of consumers and their choices in the labor, housing and goods markets are central. The model treats the choices of routes on the road network, the choice between driving and public transit, commuting to work and the frequency and lengths of non-work trips. Stocks of buildings in the model are durable, and increase by new construction. The microeconomic relationships of the model are quantified from the data with calibration and econometrics. The modeling assumptions, and the calibration strategy are explained and discussed in Section 2.
Section 3 presents the welfare analysis of congestion pricing before the disposition of the congestion toll revenue, while Section 4 examines the effects of recycling the revenue by cutting the distortionary taxes. In Section 3, aggregate change in social welfare is measured as the sum of the compensating variations of consumers in the region and of outside consumers importing from the region, the real estate value changes accruing to landlords, changes in tax revenues from all sources and the unrecycled congestion pricing revenue that accrues to public funds. In one policy, all the roads in LA County, the center of the region, are charged Pigouvian congestion tolls, and in the other all the roads in the six Counties are. Pricing causes both jobs and population to decentralize when only LA County roads are tolled and to centralize in LA County under region-wide tolling. But these effects turn out to be modest under the empirically determined elasticity structure of the model. Most of the adjustments to congestion pricing are made in the routing of trips on the road network, and in curtailing the frequency and length of non-work trips, which, in the US. are 70%-80% of all person-trips (Nelson and Niles, 2000).
We find that, in the absence of congestion pricing and keeping regional population constant, the monetized congestion delay externality is $550 per consumer per year in the baseline year 2000.4 Pricing only LA County roads, the center of the region, the externality falls to $462, and the annual toll revenue is $254 per consumer internalizing 55% of the externality. Under region-wide pricing all roads are priced, and the annual congestion delay externality per consumer falls from $550 to $398 and is fully internalized by pricing the congestion. LA County pricing and region-wide pricing revenues are 0.83% and 1.31% of average year-2000 consumer incomes and amount to 0.34% and 0.53% of the regional gross product respectively. Under LA County (or region-wide) pricing, car miles traveled decrease by 3.7% (or 5.3%) and gasoline consumption decreases by 3% (or 4.5%). Tolls paid are no more than 34 cents per mile on any road. Per round-trip, the toll is as high as $4.20 and as low as almost zero and about 84 cents on average. The total annual economic benefit is $236 per consumer if only LA County is tolled or $350 if the region is tolled. The annual toll revenue is $3.02 billion when LA County roads are tolled, and $4.73 billion when all roads are; amounting to 3.54% and 5.55% respectively of the total taxes on income, sales, property and wages. All our calculations are for a value of time when traveling, set at half the average wage. With a value of time equal to the average wage, welfare and toll revenues are about 80% higher.
Section 4 examines the “second welfare dividend” from congestion pricing depending on alternatively substituting the toll revenue from pricing roads in LA County, for the existing taxes on the income of low income workers in LA County, on sales and on property, so that region-wide revenue from all tax sources plus congestion tolls remains the same as in the baseline case when there is no pricing. Under a revenue-neutral income tax cut for the low income consumers who work in LA County, there is a second welfare dividend of $409 per consumer, 1.73 times the $ 236 welfare benefit from tolling itself. Recycling the toll revenue by this income tax cut raises regional product by 1.34%.
In a simplified non-empirical partial equilibrium model with no spatial detail and no consumer heterogeneity, Parry and Bento (2001) attributed the second dividend of road pricing from an income tax cut, to a boost in the labor supply of the consumer. In our empirical and spatially detailed model which treats systematic and idiosyncratic heterogeneity among consumers, the second dividend of road pricing from the income tax cut is largely due to a rise in purchasing power that comes not only from the positive income effect of the tax cut, but also from the relocation of jobs and residences to LA County. There is a consequent decrease in labor supply in the peripheral Counties, and wages rise there due to the low labor demand elasticity. 66.4% of the total benefit of pricing accrues to consumers and 40.6% to landlords; while those who import from the region suffer a loss equal to 7% of the total benefit because product prices increase. Under the revenue neutral sales tax cut the benefits of tolling are shared more evenly by the region's consumers (36.4%), importers from other regions (38.4%) and the region's real estate owners (the remaining 25.2%). The property tax cut is nearly neutral because almost all of the cut is capitalized into property values. This is because the distortion from the substitution of land for structural capital in buildings occurs only in a small new construction margin, and not for durable structures inherited from the past. Recycling the toll revenue by making revenue-neutral cuts to sales or to property taxes instead of the income tax, yields a negligible second dividend. Under all revenue neutral tax cuts, we find that there are only small tax interaction effects between congestion pricing and the other taxes. Conclusions and possible extensions are discussed in Section 5.
Section snippets
Model structure
The LA version of RELU-TRAN, the Regional Economy, Land Use (RELU) and Transportation (TRAN) model is an extension of the Chicago version (Anas and Liu (2007)), and is fully documented in the Appendix of the current paper. Readers chiefly interested in the details of the equation structure and the solution algorithm of the LA model are directed to read the Appendix first.
Welfare analysis
We turn to the cost of the congestion externality, and the benefits of pricing it. The model does include the effect of congestion on gasoline as explained in the Appendix. But when calculating congestion tolls, we price only for the time delay which causes the bulk of the externality, ignoring the part stemming from the excessive use of gasoline. From Eq. (A.32), our Pigouvian toll on a model road a is the gap between the marginal social and the average private costs of travel:
Substituting the congestion toll revenue for distortionary taxes
We now turn to recycling the revenue from congestion pricing to reduce other taxes. Our target taxes are the income tax for the lowest income group who work in LA County, and the sales and the property tax in LA County. In each case congestion is tolled only in LA County in the manner explained in the previous section, but now the aggregate toll revenue replaces part of the revenue from another tax in LA County so that the aggregate tax revenue from all taxes plus the tolls in the entire region
Conclusions
We found that congestion pricing interacts little with other taxes. We also found that congestion pricing without recycling the toll revenue has relatively small effects on urban markets primarily because consumers elastically substitute travel routes on the road network, even when their ability to switch to public transit is limited by unavailability and demand inelasticity. Relocations of jobs and of residences are small under congestion pricing when the toll revenue is not recycled back into
CRediT authorship contribution statement
Alex Anas: Conceptualization, Funding acquisition, Methodology, Project administration, Writing - original draft, Writing - review & editing.
References (36)
Discrete choice theory, information theory and the multinomial logit and gravity models
Transport. Res. Part B
(1983)- et al.
An analytical approximation for the macroscopic fundamental diagram of urban traffic
Transport. Res. B
(2008) - et al.
The Stockholm congestion-charging trial 2006: overview of effects
Transport. Res. Part A
(2009) - et al.
Existence of urban scale macroscopic fundamental diagrams: some experimental findings
Transport. Res. B,
(2008) - et al.
New evidence on labor supply: employment versus hours elasticities by sex and marital status
J. Monetary Econ.
(1998) - et al.
The own-wage elasticity of labor demand: a meta-regression analysis
Eur. Econ. Rev.
(2015) - et al.
The urban road pricing scheme to curb pollution in Milan, Italy: description, impacts, and preliminary cost–benefit analysis assessment
Transport. Res. Part A
(2010) - et al.
City size, network structure and traffic congestion
J. Urban Econ.
(2013) Land use and density in cities with congestion
J. Urban Econ.
(1998)- et al.
General equilibrium models of polycentric urban land use with endogenous congestion and job agglomeration
J. Urban Econ.
(1996)
Reducing urban road transportation externalities: road pricing in theory and in practice
Rev. Environ. Econ. Policy
A regional economy, land use and transportation model (RELU-TRAN): formulation, algorithm design and testing
J. Reg. Sci.
A structural model of peak-period congestion: a traffic bottleneck model with elastic demand
Am. Econ. Rev.
Markovian traffic equilibrium
Math. Program.
Identifying employment subcenters: the method of exponentially declining cutoffs
Land
Environmental levies and distortionary taxation
Am. Econ. Rev.
Optimal environmental taxation in the presence of other taxes: general equilibrium analyses
Am. Econ. Rev.
Urban sprawl and the property tax
Int. Tax Pub. Finance
Cited by (34)
What affects commute cycling in Sydney: Access, infrastructure and demographics
2024, Transportation Research Interdisciplinary PerspectivesProductivity benefits of urban transportation megaprojects: A general equilibrium analysis of «Grand Paris Express»
2023, Transportation Research Part B: MethodologicalBayesian optimization for congestion pricing problems: A general framework and its instability
2023, Transportation Research Part B: MethodologicalDemand management for smart transportation: A review
2022, Multimodal TransportationTradable mobility permits in a monocentric city with pre-existing labor taxation: A general equilibrium perspective
2022, Transportation Research Part B: MethodologicalRepositioning of driverless cars: Is return to home rather than downtown parking economically viable?
2022, Transportation Research Interdisciplinary PerspectivesCitation Excerpt :This implies that in the case of empty vehicle repositioning, adverse traffic conditions would cause only a minor individual welfare loss. This contradicts the case of occupied trips where travel time savings would imply higher privately optimal speeds and as a consequence, a higher cost of congestion (Anas, 2020; Couture et al., 2018; TTI, 2021). Second, hourly parking with no possibility to choose a flat parking charge per day makes ‘return home’ quite attractive.
Earlier versions were presented at the 11th Urban Economics Association meetings in Minneapolis, November 9-12, 2016; at the Kraks Fond Workshop on City Structure, January 20, 2017, Copenhagen; as a keynote presentation at the 3rd Urbanics Workshop on Urban Dynamics, March 13-16, 2017 in Pucón, Chile; and at the 53rd annual conference of the American Real Estate and Urban Economics Association in Philadelphia, Pennsylvania, January 5–7, 2018. The research was supported by the University of California Office of the President's Multi-campus Research Project Initiative competition of 2009, award 142934 (January 2010 – July 2016). Alex Anas was the scientific director and author of the proposal that won the award and funded the development of the RELU-TRAN model for the Greater LA Region. A David E. Lincoln Fellowship in Land Taxation from the Lincoln Institute of Land Policy (January 2014 – December 2015) for Anas supported in part the tax analysis. The author is grateful to Huibin Chang, Tomoru Hiramatsu, Debarshi Indra and Ievgenii Kudko for their diligent research assistance on various phases of the project; to Richard Church, Michael Goodchild and Wenwen Li for their empirical work on the geographical zonal structure of the model which they performed within the University of California award; to the Southern California Association of Governments for sharing data; to the Center for Sustainable Suburban Development at the University of California at Riverside for hosting and administering the project and, especially to Shayna Conaway; to Richard Arnott for his role as project director of the California award, his encouragement throughout the project and his help with the assistance of Matthew Fitzgerald, in the gathering and evaluation of the tax data. The author and not the funding agencies or others mentioned here are responsible for the results and conclusions. The comments of three anonymous reviewers were helpful in making a more detailed exposition of the simulation results presented in the paper.