Open Collections

UBC Undergraduate Research

Is Vancouver’s Proposed Road Toll Regressive? : An Empirical Analysis of Congestion Pricing See-Fernandez, Justine; Sriram, Aditi 2021-04

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata


52966-See_Fernandez_J_et_al_ECON_492_Proposed_road_toll_2021.pdf [ 2.61MB ]
JSON: 52966-1.0397111.json
JSON-LD: 52966-1.0397111-ld.json
RDF/XML (Pretty): 52966-1.0397111-rdf.xml
RDF/JSON: 52966-1.0397111-rdf.json
Turtle: 52966-1.0397111-turtle.txt
N-Triples: 52966-1.0397111-rdf-ntriples.txt
Original Record: 52966-1.0397111-source.json
Full Text

Full Text

Is Vancouver’s Proposed Road Toll Regressive?An Empirical Analysis of Congestion PricingBy Justine See-Fernandez and Aditi Sriram*AbstractVancouver is proposing to implement transport pricing in its Metro Core in an attempt toreduce pollution, however there has been pushback because of the hypothesis that theburden of this toll would fall most heavily on low-income populations. This researchstudy tests this hypothesis empirically by understanding at a Census Tract (CT) levelwhich communities would be disproportionately impacted. Results from this study showthat the toll burden falls on people from higher income CTs who have the ability tosubstitute their means of transportation with public transit. These results are similar to theexisting literature on transport pricing in other metropolitan cities, mitigating concernsthat this toll would be environmentally progressive but economically inequitable.* Climate Justice Research Collaborative, University of British Columbia. Emails: We are deeply grateful to Dr. Werner Antweiler and Andrew Figueiredo for invaluable feedback,discussion, and encouragement.Table of ContentsIntroduction 3Literature Review 3Data Collection 8Estimation Strategy 10Data Tables 13Discussion and Conclusion 18References 212I. IntroductionThe Climate Emergency Action Plan was approved for the Greater Vancouver Area (GVA) on November17, 2020 (City of Vancouver, 2020). Included in this plan is the introduction of transport pricing withinVancouver’s Metro Core by 2025, which will effectively lobby a flat toll on anyone accessing heavilycongested roads within a defined boundary in downtown. The introduction of transport pricing seeks toreduce driving during the most congested time, thereby reducing pollution from vehicles, creating moreroad space for other modes of transport and providing revenue to finance other transportation options.However, there has been pushback from some actors who claim that the toll might be combating climatechange at the expense of disadvantaged groups (Zeidler, 2020). This is because low-income families aremore likely to live farther away from the downtown area and thus, be disproportionately tolled incomparison to their higher income counterparts. The following analysis seeks to deconstruct thishypothesis quantitatively, and arrive at an understanding of which demographic groups will actually bemost impacted by this toll and if they have adequate options to substitute their transportation by usingtransit. We find that toll incidence is predicted to be higher among higher-income commuters living inVancouver’s city-centre. Additionally, we find that toll incidence increases with commuters’ margin ofsubstitution between driving and taking public transit. These findings suggest that Vancouver’s proposedtransport pricing policy is progressive along both income and transit substitution dimensions. Theremainder of the paper proceeds as follows: Section II presents relevant literature on transport pricingpolicies implemented in other cities. Section III describes our final dataset and its sources. Section IVpresents the estimation strategy we use to conduct our empirical analysis. Section V presents ourestimation results and discusses implications and section VI concludes.II. Literature ReviewMany cities around the world have been grappling with an increase in traffic volume and congestion. Thiscongestion comes with large environmental costs, causing air pollution and increased greenhouse gas3(GHG) emissions, alongside economic costs such as increases in the prices of goods and services for allconsumers (Ragan et al., 2015). To counter rising congestion levels, cities have begun implementingmobility pricing and other transport policies to price the externalities of road usage.As these transport policies are implemented in more cities, an increasing amount of research has beendedicated to understanding their implications and effectiveness in combating climate change. Notableresults from mobility pricing in cities such as London, Stockholm, and Singapore show that they areeffective in managing congestion, advancing environmental goals, and promoting public transportationover driving (Smith et al., 2010). Despite the literature confirming the positive environmental andcongestion impacts of road pricing (Jones et al., 2019, Smith et al., 2010), an obstacle that often impedesadoption of road pricing policies are equity concerns. For example, there are concerns about whetherhigher incidence of tolls and taxes could fall on already disadvantaged communities, potentially tradingoff fairness for efficiency. To explore potential equity effects, several authors have taken empiricalapproaches in determining if tax incidence or exposure disproportionately impacts the poor by studyingcities that have implemented mobility pricing policies.The literature has found that it is actually high-income households that are most impacted by congestionpricing, as they travel most frequently into charged zones. Linn, Wang, and Xie (2016) find that withBeijing’s congestion charge, impacted commuters are on average wealthier than those who are notaffected by the road pricing scheme. In addition, the authors compute the Suits index, a measure ofprogressivity, and find that the congestion charge is slightly progressive. Stockholm’s congestion chargeis of particular renown in the literature because it initially faced political and public hostility. However,following the city’s 7-month trial, over ⅔ of the population ended up voting in favour of the congestioncharges’ continued implementation. Eliasson and Mattsson (2006) develop a model for the quantitativeassessment of equity effects in Stockholm’s context, simultaneously taking into account travel patternsand individual characteristics that determine distributional effects. They find that male, high-income, and4inner city inhabitants will be most affected by the charge. In collaboration with Uber Technologies,Kitchen (2019) analyzes congestion pricing in Seattle as a solution to urban traffic. Kitchen uses traveltime and speed data from Uber to model the congestion pricing policy in Seattle’s downtown context. Hefinds that households that will be most frequently crossing the toll area are, on average, more affluent thanthe region’s households in general and that the top 10% of income earning households pay nearly 25% ofthe tolls. Households in the lower 40% of income earners pay under 20% of the tolls (Kitchen, 2019).These findings are in line with the results from Beijing and Stockholm, where the majority of road pricingexposure falls on higher-income commuters.While high-income households have borne the brunt of congestion pricing on aggregate, there is evidencethat there could be a higher relative financial burden on low-income households (Wiginton et al., 2018).Wiginton (2018) found that in Toronto and Hamilton, a highway toll would amount to 23.2% of thelowest household income quartile’s annual transport expenditure before accounting for redistributionpolicies, whereas the cost would only be 4.6% for households in the highest income quartile. Theseempirical findings from Toronto’s context show that the cost of congestion pricing as a percentage oftransportation expenditure decreases as a household’s income level increases. However, onceredistribution plans for revenues generated by the toll are considered, net equity effects are overallpositive and toll incidence ends up being higher for higher-income commuters rather than low-income.Wiginton’s analysis highlights an integral piece of congestion pricing policies: redistribution plans for therevenues generated by road tolls.The literature has shown that the key determinants of equity effects from road pricing are the structure ofthe policy and how the revenues are used (Eliasson, 2014, Smith et al., 2010, Wiginton, 2018, Schweitzer,2009). In a survey of Stockholm’s congestion charge, Eliasson (2014) finds that the equity effects leaveindividuals across all income categories better off after being compensated by either a lump sum or a taxreduction, which he describes as “bounds” for the uses of toll revenues. The lump sum is a deliberately5progressive refund, dependent on income, and the tax reduction is deliberately regressive, with a constantincome tax rate. Both refund scenarios yield net positive effects after redistribution, suggesting that awide range of uses for toll revenues can achieve a net positive effect for all income categories, with someuses being more progressive than others. Schweitzer (2009) finds similar conclusions about thedeterminants of equity effects. Her paper concludes that variation in equity impacts of transport policiesderives from differences in policy design and revenue usage.The empirical models used to estimate equity effects differ across the Beijing, Stockholm, Seattle andToronto contexts, suggesting that there are multiple methods of measuring the equity effects of mobilitypricing. Our approach to studying the Vancouver context will be different from those of theaforementioned authors. Due to the time and data constraints on our research, particularly the limited dataon time preferences and travel behaviour for Vancouver commuters, we are unable to fully replicate mostof these models. However, Wiginton’s methodology to analyze congestion pricing in Toronto andHamilton can be used as a template because it is similar to our project in assessing a policy that has notyet been implemented using travel and income data from the Canadian census.II.A Data SummaryCity Policy Tool Measured EquityImpactsResults of Study Paper(s)Beijing Congestioncharge -- Driversgoing to Beijingdowntown areaHigher incomehouseholds mostaffected (suggestingthose who drive towork in Beijing arerelatively wealthy)Congestion charge isslightly progressivePolicy effectivelyreduced congestionbecause majority of tripsinto Beijing werediscretionary and notcommutes to workLinn, Wang,and Xie(2016)Stockholm Congestioncharge -- Cordoncrossing charge,High income residentsdriving in the inner citymake more car tripsNet equity impacts aredetermined by how therevenues are spentEliasson andMattsson(2006)6varying ratesdepending ontime of dayand thus pay more ofthe taxCity centre inhabitantswill lose more from thecharges than residentsin other parts of thecountry20% reduction in trafficcongestion in city centre10-14% decrease ofemissionsEliasson(2014)Karlströmand Franklin(2009)Lee (2018)Smith et al.(2010)London CongestionchargeN/A Traffic reductions from19-25%Net revenues went totransportationimprovements in GreaterLondonLee (2018)Smith et al.(2010)Singapore Electronic RoadPricing (ERP)Achieved free-flowroad speed (desiredspeed withoutcongestion) of45-65kph onexpressways, and20-30kph on arterialsNet revenues returned tovehicle owners throughtax rebate afterinvestment into transitand highway systemsLee (2018)Smith et al.(2010)Seattle CongestionchargeHigh-income andaffluent individualswill face the highesttoll exposureTo achieve net positiveequity effects, revenuefrom congestion taxshould go towardimproving publictransportationKitchen(2019)Wu, Yin,Lawphongpanich, andYang (2012)Toronto/Hamilton(GTHA)Congestioncharge -- Flathighway toll of $2on two majorcommuterhighwaysNot yet implemented,but looked at existingcommuting routes andthe household incomeof those affectedAlso analyzed the % oftransport expenditurewould be spent on thistoll by different incomeclassesBefore consideringredistribution plans:Low-income commutersface highest toll exposureAfter consideringredistribution plans:High income householdswill be most affected bythe congestion chargeWiginton etal. (2018)7III. Data CollectionOur empirical strategy uses a unique dataset constructed from transportation surveys included in StatisticsCanada’s 2016 census data and a digitized boundary of the proposed road toll around Vancouver’s MetroCore. We first used OpenRouteService to create a distance (route) matrix for Vancouver’s census tracts(CTs), such that the travel route data obtained would allow us to know if the route between an origin anddestination CT passes through a tolled area. Using a PERL script to retrieve JSON files fromOpenRouteService, the output from this exercise was a dataset describing all potential CT pairs withinVancouver, the longitude and latitude points between origin and destination CT, and the array ofwaypoints within a route. We faced a data limitation at this point, because routes could not be computedfor a number of CT pairs. There are 13, 101 commutes based on our transportation survey data, but dataon only 10, 113 commutes were retrieved from OpenRouteService. We attribute these missing data to thepossibility that some CTs are not densely populated enough for OpenRouteService to find routes to andfrom them. Due to the time constraints of our project, we were unable to address this limitation further butexpect that with more time, a more detailed investigation into these missing routes could be achieved.To determine if any point along a route fell within a tolled area, we created a geocoded version of theproposed toll boundary to put through a point in polygon algorithm. Using a map of the proposed tollarea, we hand-coded a polygon from taking the toll area’s longitudes and latitudes from Google Maps andrunning a point in polygon algorithm in Python that would simply indicate with “True/False” whichroutes between CT pairs pass through a tolled area. We merged the “True/False” indicator data fromPython with the CT pair data from OpenRouteService to end up with a simple dataset with CT origin, CTdestination, waypoints between origin and destination CTs, longitude and latitude points, and a“True/False” indicator of whether or not the route in between passes through the tolled area. This datasethas 10, 113 observations for each CT pair for which we have data.8Our main transportation and income data comes from Statistics Canada’s 2016 transportation and nationalcensus surveys: “Long commutes to work by car”1 and “Earnings of immigrants and children ofimmigrants in official language minority populations”2, respectively. The transportation survey ismeasured at the CT pair level and the income dataset is measured at the CT level. From the transportationsurvey, we use commuting data indicating the number of commuters that drive, take public transit,bicycle, walk, or carpool in census tracts. We dropped all observations not from the Greater VancouverArea and merged this commuting data with the aforementioned OpenRouteService dataset on theproposed tolled area. The Canadian census samples one-third of the entire population, so to scale up ourtransportation data, we multiplied all values by 3 and then collapsed this data onto origin CTs,aggregating the commuting data for all destination CTs to one origin CT. After collapsing our data ontoorigin CTs, we merged this data with the income data from the 2016 census.Our final dataset has 501 observations, measured at the origin CT level, with toll exposure indicators,commuting, and income data. The table below summarizes our data sources and transformations.Dataset Source Purpose2016 Census: “Longcommutes to work bycar”StatisticsCanada,Government ofCanadaUse: Provides demographic statistics for census tracts inVancouver. All the commutes that run through thegeographic area where there is proposed to be a toll musteventually be linked back to census tract data. Keyinformation from these data are the CT coordinates.Manipulation: We first merged the commutes data with theCT data retrieved from OpenRouteService as both weremeasured at the CT pair level. To aggregate up to totalvalues of commuting data, we multiplied these data by 3(the Canadian census samples one-third of the population).2016 Census:“Earnings ofimmigrants andStatisticsCanada,Government ofUse: Provides income statistics for census tracts inVancouver. This data provided the values for ourindependent variable: household median income.2 Full article title for this specific 2016 census survey: “Results from the 2016 Census: Commuting within Canada’s largestcities”. Statistics Canada Catalogue no. 75-006-X.1 Full article title for this specific 2016 census survey: “Results from the 2016 Census: Long commutes to work by car”. StatisticsCanada Catalogue no. 75-006-X.9children ofimmigrants in officiallanguage minoritypopulations”Canada Manipulation: After merging the commutes data with thedata retrieved from OpenRouteService, we merged thatdata with this income data to create our final dataset.Centroids Derived fromStatisticsCanadageographicboundary filesfor CTsUse: This dataset provides the latitude and longitude of thecensus tract’s centre point, and will be used as thepreliminary data to calculate curther datasets.Manipulation: Using a PERL Script andOpenRouteService, these centroids data were used tocalculate the commute path between each CT pair.Toll BoundaryIndicatorCreated usingcommuting data(StatisticsCanada) onPythonUse: Each row represents the route between a pair of CT.There are a variable number of data points per rowdepending on the length of the route, with each entryrepresenting the latitude and longitude of every waypointalong the route. Out of the 10,713 routes that wererequested, 600 came back as unsuccessful navigationrequests.This file was created using a Python script whichcompared all points along a route between CT pairs to thepolygon of the toll area. This script outputted a 1 or 0 foreach CT pair to identify two aspects of the route: whetherit passed through the toll zone or ended in the toll zone(Vancouver’s downtown area).Manipulation: We merged this data with the commutesdata, collapsed on origin CT, and then merged with theincome data to create our final dataset.IV. Estimation StrategyOur empirical strategy takes a linear regression (OLS) approach in estimating different relationshipsbetween toll incidence and income. Our final dataset contains identification variables for each census tractpair (CT origin and CT destination), a binary variable indicating whether the route in between passesthrough a charge zone, and income and demographic statistics. This binary variable represents ourtreatment. We can then analyze each CT to determine how much of the toll incidence falls on itscommuters.10We are primarily curious if the proposed road toll is progressive or regressive. Specifically, we ask if tollincidence falls more on higher-income commuters. To answer this question, we run the followingregressions:(1) 𝑙𝑛(𝐶𝑖𝑇𝑐𝑜𝑚𝑖) = β0+ β1𝑙𝑛(𝑌𝑖) + ε𝑖(2) 𝑙𝑛(𝐶𝑖𝐷𝑇𝑐𝑜𝑚𝑖) = β0+ β1𝑙𝑛(𝑌𝑖) + ε𝑖where is the share of cars tolled over all types of commuters in census tract i, is the share of𝐶𝑖𝑇𝑐𝑜𝑚𝑖𝐶𝑖𝐷𝑇𝑐𝑜𝑚𝑖cars whose trips end in downtown over all types of commuters in census tract i, and is the median𝑌𝑖income of census tract i. There are 139 observations for which , meaning when we take the log𝐶𝑇𝑖𝑇 = 0form of this variable in our regression, these 139 observations are not included in the analysis. Thecoefficient of interest is , which we can interpret as an elasticity representing the relative income effectβ1of the road toll.The second part of our analysis focuses on another dimension of the road toll’s progressivity: the marginof substitution between driving and taking public transit among those affected by the toll. We are curiousif toll incidence is higher for commuters who have commuting options other than driving, which we proxyby the census tract’s access to transit. We run the following baseline regression for this analysis,specifying across census tracts that have greater and lower access to transit:(3) if𝑙𝑛(𝐶𝑖𝑇𝑐𝑜𝑚 ) = β0 + β1𝑙𝑛(𝑌𝑖) + ε𝑖, 𝑙𝑛(𝑇𝑖𝐶𝑖) > 0. 0503(4) if𝑙𝑛(𝐶𝑖𝑇𝑐𝑜𝑚 ) = β0 + β1𝑙𝑛(𝑌𝑖) + ε𝑖, 𝑙𝑛(𝑇𝑖𝐶𝑖) < 0. 050311where is the share of cars tolled over all types of commuters in census tract i and is the median𝐶𝑖𝑇𝑐𝑜𝑚𝑖𝑌𝑖income of census tract i. To specify between groups who have greater and lower access to transit, wecondition the above regression on observations that have above or below , which is the 25th𝑇𝑖𝑐𝑜𝑚𝑖0. 0503percentile value of our relative transit access identifier, .𝑇𝑖𝑐𝑜𝑚𝑖Because accounts for cars that can either be passing through the toll area or ending their trip𝐶𝑇𝑖𝑇downtown, we are also curious about the distribution of the toll exposure among commuters who are ableto avoid crossing the toll area since their trips do not end downtown (destination CT lies outside of the tollarea, but route passes through toll area). For this, we use a negative binomial regression to estimate therelationship between the median income of census tracts and the share of cars who pass through thecharge zone but do not end their trips downtown over total commuters in each CT. This specific form ofregression and Poisson distribution is well suited for this analysis because possible values for in𝐶𝑖𝑃𝑐𝑜𝑚𝑖equation (5) below are non-negative integers. The following equation shows the specification we used forthis disaggregated analysis:(5) 𝑙𝑛(𝐶𝑖𝑃𝑐𝑜𝑚𝑖) = β0+ β1𝑙𝑛(𝑌𝑖) + ε𝑖where is the share of cars whose commute route passes through the charge zone but does not end up𝐶𝑖𝑃𝑐𝑜𝑚𝑖downtown over all types of commuters for census tract i. is log median income of census tract i.𝑌𝑖12V. Data TablesTo establish the context of Vancouver’s commuting population, we summarize commuting data for municipalities in the Greater Vancouver Area.Tables 1 presents this data below. Column (1) presents the number of origin CTs in the corresponding city. Column (2) presents the share of tolledcars of the total number of cars commuting from a census tract in the corresponding city. We computed these data from the transportation surveyby multiplying the number of trips taken by car by 3 to aggregate up to total numbers, since the census samples one-third of the entire population.Column (3) presents the total number of cars commuting from a census tract in the corresponding city. Columns (4) and (5) present the number ofcars that are tolled and the number of cars whose trips end in downtown Vancouver, respectively, that originate their commute in a census tract inthe corresponding city.Table 2 presents the same data as Table 1 with values adjusted for the city’s population for ease of comparison. Column (5) presents medianhousehold income for the corresponding city.Table 1 Toll Impact: Unweighted by census tract population(1) (2) (3) (4) (5)City N Proportion of tolled carsout of total cars (%) Total cars Number of tolled carsNumber of cars with tripsending in downtownVancouverVancouver 113 60.56 166, 670 78, 485 62, 125Burnaby 40 7.80 79, 425 15, 870 14, 805Richmond 39 3.71 101,159 10, 904 10, 319North Vancouver 24 4.39 61, 110 10, 755 9, 945Coquitlam 23 2.05 59, 444 5, 789 5, 384West Vancouver 8 2.03 17, 340 3, 195 3, 045Surrey 90 1.80 187, 544 5, 235 5, 05513Delta 18 1.24 54, 824 4, 035 3, 720New Westminster 13 0.95 27, 885 2, 519 2, 519Port Moody 6 0.82 13, 605 2, 010 1, 950Port Coquitlam 8 0.61 30, 885 2, 490 2, 205Musqueam 2 1 0.19 315 60 60Maple Ridge 13 0.14 36, 329 420 420Langley 27 0.14 69, 044 420 359Capilano 5 1 0.10 885 90 90White Rock 5 0.05 7, 170 90 90Pitt Meadows 4 0.02 9, 930 45 45Note: Values are not weighted by population. Areas in the Greater Vancouver Area also included in the transport survey data but for which we were unable to retrieve toll datafrom OpenRouteService: Barnston Island 3, Burrard Inlet 3, Coquitlam 1 and 2, Katzie 1 and 2, Langley 5, Matsqui 4, McMillan Island 6, Mission 1, Semiahmoo, Seymour Creek2, Tsawwassen, Whonnock 1Table 2 Toll Impact: Weighted by census tract population(1) (2) (3) (4) (5)City NProportion of tolledcars from total cars(%)Share of tolled cars percapita (%)Share of cars with tripsending in downtownVancouver per capita (%)Median householdincome ($)Vancouver 113 53.6 12.7 10.3 69, 370.57West Vancouver 8 25.4 9.10 8.70 102, 759.75Burnaby 40 19.5 6.80 6.30 70, 031.36Musqueam 2 1 19.0 3.60 3.60 81, 408North Vancouver 24 18.3 8.60 8.00 96, 876.30Port Moody 6 13.7 6.50 6.30 96, 702.43Capilano 5 1 10.2 3.10 3.10 38, 933Richmond 39 9.50 5.00 4.80 70, 707.97Coquitlam 23 8.90 4.10 3.90 79, 958.42Port Coquitlam 8 7.60 4.50 3.90 91, 242.56New Westminster 13 7.30 2.90 2.90 67, 038.92Delta 18 6.90 4.00 3.70 93, 224.26Surrey 90 2.00 0.90 0.09 82, 307.74Maple Ridge 13 1.10 0.60 0.06 89, 193.33White Rock 5 1.00 0.04 0.04 66, 763.20Langley 27 0.05 0.02 0.02 87, 325.10Pitt Meadows 4 0.00 0.02 0.02 87, 876Note: Values are weighted by population. Areas in the Greater Vancouver Area also included in the transport survey data but for which we were unable to retrieve toll data fromOpenRouteService: Barnston Island 3, Burrard Inlet 3, Coquitlam 1 and 2, Katzie 1 and 2, Langley 5, Matsqui 4, McMillan Island 6, Mission 1, Semiahmoo, Seymour Creek 2,Tsawwassen, Whonnock 1.14Figure 1 Scatterplot: Toll Incidence and IncomeTable 3 Outcome Variable: Share of Toll ExposureTo explore whether or not the proposed road toll is progressive along income dimensions, we estimatedthe effect of log income on the log transformations of the shares of cars that pass the toll area and cars thatend their trips downtown. These results are displayed in Table 3 above. Because of the log specification ofour regression, we can interpret these coefficients as elasticities. A 10% increase in income is associatedwith a 3.94% increase in the share of cars that will face the road toll across all types of commuters.Similarly, a 10% increase in income is associated with a 5.07% increase in the share of cars that end their15trips downtown and, by definition, unavoidably face the toll. These results suggest the proposed road tollis progressive as toll incidence increases with income. Additionally, in Figure 1 above, we see that tollincidence is concentrated among commuters with household incomes that are higher than what isconsidered low-income.3Table 4 Outcome Variable: Share of Toll ExposureWe are also interested in the distribution of the toll exposure across communities with varying access totransit. Another dimension of progressivity would be if tolled commuters are those with a margin ofsubstitution for other modes of transportation. To test for such a margin of substitution, we estimate therelationship between income and toll incidence, stratifying across commuters with varying access totransit. The cutoff we use to condition these specifications is the 25th percentile of the proportion oftransit accessibility over all modes of transportation. These results of equations (3) and (4) are displayedin Table 4 above. For commuters who live in census tracts with greater access to transit (above 25thpercentile), a 10% increase in income is associated with a 6.59% increase in the share of cars that face thetoll. This result suggests that toll incidence is higher for commuters who do have a margin of substitutionbetween driving and taking transit. For commuters who live in census tracts with lower access to transit(below 25th percentile), a 10% increase in income is associated with a 22.9% increase in the toll burden,but we do not attribute much economic significance to this coefficient since only a small number ofobservations were used in this estimation. We attribute the small sample to geographical explanations,where these commuters likely live in areas with little access to transit.3 $28,400 = LIM (Low Income Measure) calculated by taking half of household after-tax median income in 2016 ($56,800) as perStatistics Canada’s definition of “low income”: individuals are considered to be living in low income if their household after-taxincome falls below half of the median after-tax income16Table 5 Outcome Variable: Commuters who can change route around tollFinally, we are interested in observing the relationship between income and toll incidence across adisaggregated sample, for commuters who only pass through the toll area but do not necessarily end theirtrips downtown. These commuters therefore may have the option of adjusting their commute to go aroundthe toll area, as their final destinations lie outside the toll boundary. We use two specifications to test forthe relationship between income and toll exposure for commuters who have the option to avoid the tollonce it is in place: our regular linear regression and a negative binomial regression to allow for zerovalues in our dependent variable. These results are displayed in Table 5 above. Both specifications yieldnegative and insignificant estimates. We consider two explanations for these results. The first is that thereis no significant relationship between income and toll incidence for commuters who have the option toavoid paying the toll. The second is that there are limitations to our data.Overall, the results of our empirical analysis point to a positive relationship between income and tollexposure. These results are displayed in Figures 2 and 3 below, both created using QGIS, to provide avisual understanding of the communities that are impacted by the toll and their relative affluence. Theseresults suggest the proposed congestion pricing policy is progressive along relative income dimensions,where toll incidence increases with income. In addition, the policy is progressive along a transitsubstitution dimension, where the toll area increases as a census tract’s accessibility to transit increases.17Figure 2 Map: Geographical Distribution of Toll Incidence in each Census TractFigure 3 Map: Household Median Income in each Census Tract18VI. Discussion and ConclusionThe goal of this empirical study was to determine which commuters in the Greater Vancouver Area wouldbe most impacted by the transport pricing policy included in Vancouver’s Climate Emergency ActionPlan. We tested the hypothesis that congestion pricing combats climate change at the expense ofdisadvantaged groups by estimating the relationship between relative income and toll exposure. However,our results are quite in line with the existing literature which finds that the congestion pricing policiesimplemented so far in Beijing, Stockholm, and Seattle are progressive. In Vancouver, we find that tollincidence is greater among higher-income commuters and that the proposed road pricing policy is thusprogressive. We also estimated the toll impact across commuters with varying margins of substitutionbetween driving and taking public transit, discovering that the toll is progressive along transportationsubstitution dimensions as well. Toll exposure increases as commuters’ margins of substitution increase.Based on our results, the congestion pricing policy is designed in such a way that those more affected bythe toll are higher-income and coming from places with greater access to transit and thus more options fordifferent modes of transportation should they want to start driving less.Our study tested the progressivity of Vancouver’s proposed toll along relative income and transitsubstitution dimensions prior to any redistribution plans for revenues generated from the policy. Based onthe existing literature, a significant determinant of political reception to such road charges are theiraccompanying redistribution policies (Schweitzer, 2009). Similarly, Eliasson and Mattsson (2006) findthat net equity impacts of Stockholm’s congestion charge were determined by how the toll revenues werespent. In a different context, Toronto’s proposed congestion charge is expected to be progressive afterconsidering revenue redistribution efforts, but without such efforts, toll incidence is predicted to behighest among low-income communities. This is not the case in Vancouver, as our study finds theproposed toll is overall progressive before considering any redistribution plans for toll revenues.Nevertheless, we expect that as with any proposed tax, the policy may still be met with some hostility.19Combining our research with the City’s redistribution plan for the revenues generated from the toll wouldlikely strengthen efforts to implement the road charge by 2025. The summaries of toll impact presented inTables 1 and 2 show that Vancouver is the city with the highest toll incidence as 60% of their commutersare expected to pass through the toll area, while Burnaby, the city with the next highest toll incidence, canexpect 7.8% of their driving commuters to pass the toll area. The information presented in these tablesshow that it is commuters within Vancouver who face the highest toll exposure, addressing equityconcerns about toll incidence disproportionately impacting commuters coming from non-Vancouver CTs.Situating our research in the existing literature, the magnitudes of our results and their implications are inline with the studies described in our literature review. As found in Beijing, Stockholm, and Seattle,higher-income and city centre inhabitants face the highest toll incidence, suggesting the congestionpricing policies implemented in those cities are progressive (Linn, Wang, and Xie, 2016; Eliasson andMattsson, 2006; Kitchen, 2019). Studies on Stockholm and London’s respective congestion charges showthat they reduced traffic by 19-25% (Eliasson and Mattsson, 2006; Lee, 2018). When a road pricingpolicy is implemented in Vancouver, it would be curious to see the toll’s efficacy in reducing congestionlevels and compare these results to other cities with similar policies. There are a number of areas forfuture research on Vancouver’s proposed transport pricing policy that we were unable to explore due tothe time constraint on our study. Some unanswered questions that remain ask about other substitutioneffects of the proposed policy. We touched briefly on commuters’ margins of substitution between drivingand taking public transit, but it would be curious to explore other modes of transportation and the toll’sprogressivity along these dimensions. Additionally, we were unable to estimate toll impact acrossdifferent sociodemographic variables such as ethnicity, education level, and employment status.Analyzing the distribution of toll exposure across these sociodemographic indicators would be aninteresting area for future research that would address further equity concerns that may remainsurrounding congestion pricing.20ReferencesCity of Vancouver. (n.d.). Transport pricing. Retrieved April 14, 2021, from, J., & Mattsson, L. (2006). Equity effects of congestion pricing. Transportation Research Part A:Policy and Practice, 40(7), 602-620. doi:10.1016/j.tra.2005.11.002Eliasson, J. (2014, July). The Stockholm congestion charges: an overview (Rep.). Retrieved, C. et al. (2019, September). Congestion Pricing in NYC: Getting it right (Rep.). Retrieved, M.. (2019, July). Fair and Efficient Congestion Pricing for Downtown Seattle(Rep.). Retrieved, J., Wang, Z. & Xie, J. (2015).  Who will be affected by a congestion pricing scheme in Beijing?Transport Policy, 47, 37-40. of Citizens’ Services. (2018, March 13). Issue 18–51: Canada Income Survey, 2016 - Provinceof British Columbia. Issue 18–51: Canada Income Survey, 2016.,growth%20of%200.9%25%20in%202015.Ragan, C. et al. (2015, November). We can't get there from here: Why traffic congestion is critical tobeating it (Rep.). Retrieved, Katherine. 2019. “Results from the 2016 Census: Commuting within Canada’s largest cities”.Insights on Canadian Society. May. Statistics Canada Catalogue no. 75-006-X.Schweitzer, L. (2009). The Empirical Research on the Social Equity of Gas Taxes,Emissions Fees, andCongestion Charges (Rep.) Retrieved, V. et al. (2010, April). International scan: Reducing congestion and funding transportation usingroad pricing (Rep.) Retrieved, L.. (2010, April). Exploring how road pricing on the DVP andGardiner would impact income groups in the GTHA (Rep.) Retrieved, Tetyana; Gilmore, Jason and Sébastien LaRochelle-Côté. 2019. “Results from the 2016 Census:Long commutes to work by car”. Insights on Canadian Society. February. Statistics Canada Catalogue no.75-006-X.Zeidler, M. (2020, November 07). As Vancouver ponders downtown toll for drivers, experts warn it couldharm low-income commuters. Retrieved April 14, 2021, from


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items