Contents lists available at ScienceDirect
Transportation
Research
Part
B
journal homepage: www.elsevier.com/locate/trb
The
cost
of
congestion
and
the
benefits
of
congestion
pricing:
A
general
equilibrium
analysis
✩
Alex Anas
∗DepartmentofEconomics,415FronczakHall,NorthCampus,StateUniversityofNewYorkatBuffalo,Amherst,NewYork14260,USA
a
r
t
i
c
l
e
i
n
f
o
Articlehistory: Received9June2019 Revised27January2020 Accepted5March2020
JELcodes: D58 D62 H23 R13 R41
Keywords:
Pigouviantaxationofroadcongestion Doubledividend
Incometax Salestax Propertytax
a
b
s
t
r
a
c
t
Thebenefitsofcongestionpricingareexaminedwithaspatiallydetailedcomputable gen-eralequilibriummodeloftheGreaterLAregion.Themodeltreatschoicesofroadsona network,ofdrivingorpublictransit,ofresidenceandjoblocation,ofnon-worktrip pat-terns,ofhousingtypeandsize;vacanciesandnewconstruction,production,interindustry tradeand exports.The aggregatebenefitofpricingLACountyroadcongestionincreases 2.7foldwhenthetollrevenueisrecycledtocuttheincometaxofthepoorerLACounty workerswhilemaintainingregion-widetaxrevenueneutrality.Consumersget66%ofthe aggregatebenefitandlandlords41%,whileimporterssufferlossesequalto7%ofthe ag-gregatebenefit.Grossregionalproductincreasesby1.34%.Theaggregatebenefitchanges negligiblywhenthetollrevenueisrecycledtocutsalestaxesinsteadoftheincometax, butgrossproductstill risesby0.6%.Under thesales taxcut, consumers,importersand landlordsbenefitmoreevenlythanunder theincometaxcut,butwiththepropertytax cutnearlyallofthecutiscapitalizedintopropertyvalues.
© 2020TheAuthors.PublishedbyElsevierLtd. ThisisanopenaccessarticleundertheCCBY-NC-NDlicense. (http://creativecommons.org/licenses/by-nc-nd/4.0/ )
1. Introduction
Wepresentaspatially detailedempiricalgeneralequilibriumtreatmentofthemarket andwelfare effectsofroad con-gestionandof pricingthecongestionatits social marginalcost. Inthemodel,consumers andfirmsadjustto congestion
✩Earlierversionswerepresentedatthe11th UrbanEconomicsAssociationmeetingsinMinneapolis,November9-12,2016;attheKraksFondWorkshop
onCityStructure,January20,2017,Copenhagen;asakeynotepresentationatthe3rd UrbanicsWorkshoponUrbanDynamics,March13-16,2017inPucón, Chile;andatthe53rd annualconferenceoftheAmericanRealEstateandUrbanEconomicsAssociationinPhiladelphia,Pennsylvania,January5–7,2018. TheresearchwassupportedbytheUniversityofCaliforniaOfficeofthePresident’sMulti-campusResearchProjectInitiativecompetitionof2009,award 142934(January2010– July2016).AlexAnaswasthescientificdirectorandauthoroftheproposalthatwontheawardandfundedthedevelopmentof theRELU-TRANmodelfortheGreaterLARegion.ADavidE.LincolnFellowshipinLandTaxationfromtheLincolnInstituteofLandPolicy(January2014– December2015)forAnassupportedinpartthetaxanalysis.TheauthorisgratefultoHuibinChang,TomoruHiramatsu,DebarshiIndraandIevgeniiKudko fortheirdiligentresearchassistanceonvariousphasesoftheproject;toRichardChurch,MichaelGoodchildandWenwenLifortheirempiricalworkon thegeographicalzonalstructureofthemodelwhichtheyperformedwithintheUniversityofCaliforniaaward;totheSouthernCaliforniaAssociationof Governmentsforsharingdata;totheCenterforSustainableSuburbanDevelopmentattheUniversityofCaliforniaatRiversideforhostingandadministering theprojectand,especiallytoShaynaConaway;toRichardArnottforhisroleasprojectdirectoroftheCaliforniaaward,hisencouragementthroughoutthe projectandhishelpwiththeassistanceofMatthewFitzgerald,inthegatheringandevaluationofthetaxdata.Theauthorandnotthefundingagencies orothersmentionedhereareresponsiblefortheresultsandconclusions.Thecommentsofthreeanonymousreviewerswerehelpfulinmakingamore detailedexpositionofthesimulationresultspresentedinthepaper.
∗ Correspondingauthor.
E-mailaddress:alexanas@buffalo.edu
https://doi.org/10.1016/j.trb.2020.03.003
andtoits pricingin manymargins.The modelisstatic, not dynamic, andhenceit cannot determinewhich adjustments byconsumersandfirmswouldhappenquicklyandwhichwouldtakelonger. Therefore,themodelisusedtocomparethe effectsof congestionpricing afterthe markets havefully adjusted to such pricing.We comparemarket adjustments and welfaregainswhentherevenue frompricingisnotrecycled; andthen whenitis recycledbycuttingdistortionarytaxes, whilemaintainingaggregateregion-widerevenueneutralityfromalltaxsourcesincludingcongestiontolls.
Economistshave been interestedin congestion pricingever since Vickrey (1963) calledfor thepricing of roadtraffic accordingtothewelfareeconomicprincipleofPigou(1932).Sincethen,congestionpricingcontinuestobecomeincreasingly palatablepolitically because of the rising cost of congestion in the big cities of the world. London, U.K. (Leape, 2006); Stockholm,Sweden(Eliassonetal.,2009);andMilan,Italy(Rotarisetal,2010)haveintroducedcongestionpricingintheir centralareas,joiningSingaporeandsome smallercitiesinNorwaythathavehadcongestionpricingformanyyears.Fora detailedreviewoftheseschemes,seeAnasandLindsey(2011).1
Congestionandcongestionpricinghavebeenstudiedtheoreticallyinurbaneconomics,usingthemonocentriccitymodel. Inthismodel,itisassumedthatalljobsarelocateddowntown(thecenterofacircularcity)andcannot relocate,thereis no public transit,all trips are commutes to work, buildings are not durable and, often, all consumers are identical. The earliest theoreticaltreatment ofcongestion inthemonocentric city modelwasby Strotz (1965)andone ofthe most re-centby Wheaton(1998). Inthese and other analyses, the congestion externalityis internalized by tollingevery mileof theradial commute tothe downtown. In such a setting,consumers canblunt theimpact of thetoll, only by moving to a residencecloserto the downtownto reduce milestraveled andtollspaid. Consequently, populationdensities near the downtownincreaseandthemonocentriccitybecomesmorecompact.Asacounterpointtomonocentricanalysis,Anasand Kim (1996)demonstratedin atheoretical generalequilibriummodel thatcongestion andits pricingcancausethe emer-genceofjobcentersorthedecentralizationofjobsfromthedowntowncentertotheperiphery.Withhighcongestion,firms move closertotheir workers andcustomers,reducing their travel times,andbenefit by payinglower wages orcharging moreforoutput.
Ontheempiricalside,alikelymagnitudeofcongestionpricingwasfirstestimatedbyKeelerandSmall(1977),who cal-culatedoptimalusertollson asampleof suburbanandcentral cityhighways inthe SanFrancisco BayArea.They found thatthe socially optimalrush-hourspeedswere considerablyhigherthan theobserved speeds. Butlittleis knownabout howandhowmuchcongestionpricingwouldrealisticallymodifyurbanstructure; howmuchrents,wages,pricesandthe location of jobs andresidences would change; and how the benefits of congestion pricing would be distributed among consumergroupsandsectorsoftheeconomy.Modelingtheseeffectswithinareasonablycompletestructuralgeneral equi-libriummodelremainedachallenge.2 Someeconomistshaveexpressedcautionevenabouttherelativelylimitedchallenge
ofcalculatingcongestiontollsonanetworkofroads.Forexample:
“…Onthepracticalside, itmaybe computationallyinfeasibleto estimatemarginalcongestion costsonevery single link andintersectioninanurbanroadnetwork,particularlygiventhatpricingatonepointdivertstrafficelsewherewithinthe network,...” (Parryetal.2007,page393).3
Inourmodel,theequilibriumdeterminationofcongestiontollsonalllinksofaroadnetworkisrenderedfeasible,and labor,housingandother realestate,productionandtravelmarketsincludingtheroadnetworkaretreatedsimultaneously. Thespatialdistribution ofjobsand residencesare not predeterminedbutareinterdependent. Industries arealso directly interdependent because they exchange intermediate inputs. These interdependenciesgive rise to pecuniary savings from theproximal locationsofproducersandtheirlaborersandcustomers;andalsofromtheproximallocationofproducersin differentindustries.Thesepecuniarysavings,therefore,areaformofagglomerationeconomydespitetheassumptionof con-stantreturnstoscaleinproduction.Thecurrentversionofthemodelignorestechnological(non-pecuniary)agglomeration economies,alsoknownasMarshallianexternalities.Ifsuchexternalitieswereincluded,thenthetotalfactorproductivity co-efficient(scalefactor)oftheproductionfunctionwouldchangeaccordingtonon-pecuniaryexternalitiessuchasinformation flowsbetweenfirms.
Microeconomiclinksbetweenthetraveldecisionsofconsumersandtheirchoicesinthelabor,housingandgoods mar-ketsarecentral.Themodeltreatsthechoicesofroutesontheroadnetwork,thechoicebetweendrivingandpublictransit, commutingto workand thefrequency andlengthsofnon-work trips. Stocks ofbuildings inthe modelare durable,and increasebynewconstruction.Themicroeconomicrelationshipsofthemodelarequantifiedfromthedatawithcalibration andeconometrics.Themodelingassumptions,andthecalibrationstrategyareexplainedanddiscussedinSection2.
Section3presentsthewelfareanalysisofcongestionpricingbeforethedispositionofthecongestiontollrevenue,while Section4examinestheeffectsofrecyclingtherevenuebycuttingthedistortionarytaxes.InSection3,aggregatechangein socialwelfareismeasuredasthesumofthecompensatingvariationsofconsumersintheregionandofoutsideconsumers
1TheNorwegiantollringsareintendedtoraiserevenuenottoalleviatecongestion.Thesecitieshaveusedcordontollsonlyattheirentrancesandonly
someofthemchargefeesthatdifferbypeakandoff-peakperiodsoftheday.TheSingaporecongestionpricingschemeincludescordonchargesforthree restrictedzonesaroundtheCBD,tollsonexpressways,andtollsonarterialroads.
2 TheNBERurbansimulation modeldevelopedbyIngram,KainandGinn (1972)didnottreatroadcongestion.Thelinear programmingmodelof Mills(1972)didtreatcongestionbutignoredmanymarginsofadjustmenttocongestion,andcouldnotbeappliedempirically.
3Advancesineconomists’understandingofcomputationalmethods(Judd,1998)andthecomputationalversatilitymadepossiblebytheGeneral
importingfromtheregion,therealestatevaluechangesaccruingtolandlords,changesintaxrevenuesfromallsourcesand theunrecycledcongestionpricingrevenuethataccruestopublicfunds.Inonepolicy,alltheroadsinLACounty,thecenter oftheregion,arechargedPigouviancongestiontolls,andintheother alltheroadsinthesixCountiesare.Pricingcauses bothjobsandpopulationtodecentralizewhenonlyLACountyroadsaretolledandtocentralizeinLACountyunder region-wide tolling. Buttheseeffects turnout to bemodest underthe empiricallydetermined elasticitystructure ofthe model. Mostoftheadjustments tocongestionpricingaremadeintheroutingoftripsontheroadnetwork,andincurtailingthe frequencyandlengthofnon-worktrips,which,intheUS.are70%-80%ofallperson-trips(NelsonandNiles,2000).
Wefindthat,intheabsenceofcongestionpricingandkeepingregionalpopulationconstant,themonetizedcongestion delayexternalityis$550perconsumerperyearinthebaselineyear2000.4PricingonlyLACountyroads,thecenterofthe
region,theexternalityfallsto$462,andtheannualtollrevenueis$254perconsumerinternalizing55%oftheexternality. Underregion-widepricingallroadsarepriced,andtheannualcongestiondelayexternalityperconsumerfallsfrom$550to $398andisfullyinternalizedbypricingthecongestion.LACountypricingandregion-widepricingrevenuesare0.83%and 1.31% ofaverageyear-2000consumerincomesandamountto 0.34%and0.53%oftheregional grossproductrespectively. UnderLACounty(orregion-wide)pricing,carmilestraveleddecreaseby3.7%(or5.3%)andgasolineconsumptiondecreases by3%(or4.5%).Tollspaidarenomorethan34centspermileonanyroad.Perround-trip,thetollisashighas$4.20and aslowasalmostzeroandabout84centsonaverage.Thetotalannualeconomicbenefitis$236per consumerifonlyLA Countyistolledor$350iftheregionistolled.Theannualtollrevenueis$3.02billionwhenLACountyroadsaretolled,and $4.73 billion whenall roadsare; amounting to3.54% and5.55%respectivelyofthe totaltaxesonincome,sales,property andwages.Allourcalculationsareforavalue oftime whentraveling, setathalf theaveragewage.Witha valueoftime equaltotheaveragewage,welfareandtollrevenuesareabout80%higher.
Section 4examines the“secondwelfare dividend” from congestionpricingdependingonalternativelysubstituting the tollrevenuefrompricingroadsinLACounty,fortheexistingtaxesontheincomeoflowincomeworkersinLACounty,on salesandonproperty, sothat region-widerevenue fromall taxsources pluscongestiontollsremains thesameasinthe baseline casewhenthereisnopricing.Underarevenue-neutralincometax cutforthelowincomeconsumerswhowork inLACounty,thereisa second welfaredividendof$409per consumer,1.73timesthe$ 236welfarebenefitfromtolling itself.Recyclingthetollrevenuebythisincometaxcutraisesregionalproductby1.34%.
Inasimplifiednon-empiricalpartialequilibriummodelwithnospatialdetailandnoconsumerheterogeneity,Parryand Bento (2001) attributedthe second dividend of road pricingfrom an incometax cut, to a boost in the labor supply of theconsumer.Inourempiricalandspatiallydetailedmodelwhichtreatssystematicandidiosyncraticheterogeneityamong consumers,theseconddividend ofroadpricingfromtheincometax cutislargelyduetoariseinpurchasingpowerthat comes not onlyfromthepositive incomeeffect ofthetax cut, butalsofromtherelocation ofjobsand residencestoLA County.There is a consequent decrease inlabor supply inthe peripheral Counties,and wages risethere dueto the low labordemandelasticity.66.4%ofthetotalbenefitofpricingaccruestoconsumersand40.6%tolandlords;whilethosewho import fromtheregion suffera loss equalto 7% ofthe totalbenefit becauseproduct pricesincrease. Underthe revenue neutralsalestaxcutthebenefitsoftollingaresharedmoreevenlybytheregion’sconsumers(36.4%),importersfromother regions (38.4%)andthe region’srealestateowners(the remaining 25.2%). Theproperty taxcut isnearly neutralbecause almost all ofthe cut is capitalized into property values.This is because thedistortion from the substitution of land for structuralcapitalinbuildingsoccursonlyinasmallnewconstructionmargin,andnotfordurablestructuresinheritedfrom thepast.Recyclingthetollrevenuebymakingrevenue-neutralcutstosalesortopropertytaxesinsteadoftheincometax, yields a negligible second dividend.Under all revenue neutraltax cuts, we findthat there areonly smalltax interaction effectsbetweencongestionpricingandtheothertaxes.ConclusionsandpossibleextensionsarediscussedinSection5.
2. Modelstructure
TheLAversionofRELU-TRAN,theRegionalEconomy,LandUse(RELU)andTransportation(TRAN)modelisanextension oftheChicagoversion(AnasandLiu(2007)),andisfullydocumentedintheAppendixofthecurrentpaper.Readerschiefly interested in the details of the equation structure andthe solution algorithm of the LA model are directed to read the Appendixfirst.
2.1. Geographyandtheroadnetwork
Theregionispartitionedinto97compactzones(Fig.1).5EachzonelieswhollyinsideaCounty,andzonesaredistributed
as follows among the six Counties: Imperial (2), Los Angeles (46), Orange (17), Riverside(15), San Bernardino (14) and
4Theseandothernumbersallreportedfortheyear2000shouldbemultipliedby1.508,theCumulativeConsumerPriceIndexmultiplierforurban
consumers,togetasenseoftheirmagnitudesin2020dollars.
5 Li,ChurchandGoodchild(2014),solveda“p-compact-regionsproblem”,aggregatingthe4,109TrafficAnalysisZones(TAZs)ofthesixCountiesintothe
Fig.1. Countiesandmodelzones.
Ventura(3).Fig.2isamapofthe“arcs” (or“links”)andthe“nodes” definingtheroadnetworkthatconnectsthezones.An arcrepresentstwomodelroadsoneineachdirection,correspondingtoasegmentofafreeway,aStateroadoramajorroad oranaggregationofparallelsuchroads.Thereare696modelroads.Eachmodelzonehasacentroidthatisatriporiginor tripdestinationnodeonthenetwork,andthereare114othernodesthatarenottriporiginsordestinationsbutallowcar tripstoswitchfromonemodelroadtoanother.
2.2.Landandbuildings
Fig.2. Themodel’sarc-noderoadnetwork.
2.3. Laborandhousingmarkets
2.4.Productionandconsumption
Productioninthemodeloccursinnine primaryindustries6 andthereisalsoa constructionindustry foreachbuilding
type.Thesefourteenindustriesarecompetitiveandconstantreturnstoscale,andcanproduceinanyzoneutilizingasinput, laborthatcommutesthere,floorspaceincommercialbuildingsavailablethere,andintermediateinputsfromallthezones andindustries that are imperfect substitutes inproduction. Consumers who are in the labor force provide labor of four income-skilllevels. Non-workersare alsodivided intothefourtypesbutdonot earnwages.All consumers(workersand non-workers)chooseazoneofresidenceandahousingtypeinthatzone,andworkersalsochooseazonewheretheywork. Consumersacquireacompositeofthegoodsproducedbythenineprimaryindustries.Suchcompositegoodsfromdifferent zonesareimperfectsubstitutesinconsumerpreferences.Consumersmakenon-worktripsfromtheirhomezonestoallthe zonesoftheregiontoacquire thesecompositegoods,withquantity purchasedandtrips fromthehome zoneattenuating withthesumofthesalespriceandthemonetarycostoftravelwhich,inapricingregime,includestollspaid.
2.5.Travelmarketsandadjustmentstocongestionandtopricing
Personaltravelconsistsofworktrips(commutes)andnon-worktrips.Themodeltreatsaverage-over-the-dayconditions, focusingonthechoiceofrouteontheroadnetwork,choiceoftravelbycarorbypublictransit,ornon-motorizedmodesfor shorttripsincludingworkingathome.Workersareassumedtocommuteeachwork-day.Asmentionedabove,non-workers andworkersalso choose the numberof their non-work trips from their residencezones toall thezones to acquire the compositegoodsproducedthere,togetherwiththemodeofeachsuchtripanditsrouteonthenetwork.Alternativeroutes connectingatrip-origintoatrip-destinationareviewedbythetravelerasimperfectsubstitutes.
Acartripfromonezonetoanotherexperiencescongestionintra-zonallyintheoriginzonetoaccessthenetworkandin thedestinationzonetoegressfromthenetwork,andonthesequenceofmodelroads(alsoreferredtoasarcsorlinks) trav-eledtoreachthedestination.Roadcongestionismodeledbytheflowcongestionfunction(seeEq.(A.32)intheAppendix) suchthatthecongestedtraveltimeincreasesconvexlywiththeratioofthetrafficflowtothemodelroad’scapacity. Intra-zonalcongestionincreasesconvexlywiththeratioofaggregatetrafficflowtotheaggregatestreetcapacityofthezone.7The
flowoftrafficisobtainedbysummingoverconsumers’travelchoiceswhichareinturnderivedfromtheirlocationchoices andtheirconsumptiondemands.
The total monetized cost of a trip to work by car consists ofits cost of fuel, plus the travel time over the network multipliedbythe value oftimewhile traveling(vot).More congestionreducesspeedandincreasesfuel consumptionper mile,whereascongestionpricingincreasesmonetary cost,reducingtrips andcongestion, andinducing fastertravel.While sometripsswitchtothenon-carmodes,therestfavorthelesscongestedand,hence,thelesstollednetworkroutes.Given thesameresidenceandjoblocation,undermorecongestion,non-worktripsarereducedormadeclosertohome.Another response by workersisto reduce the commute distance by choosingjob andresidencelocationsthat are closer toeach other,orby choosingresidences withlower rents,orjobswithhigherwagestocompensate forthecostliertravel.Lower incomeconsumershavealow marginalrateofsubstitution betweentime lostincommutingandtheir disposableincome andare themostinclinedto switch to roadswith lower tollsandmakefewer andshorter non-work trips, whilehigher incomeconsumersaremoretolerantofhighertollsandbenefitmorefromthefastertraveltimesinducedbythetolls.Public transitistreatedasuncongestedwithexogenouszone-to-zone monetaryandtimecosts.The modeldealswiththechoice betweenpublictransitanddrivingbutnotwiththedetailsofroutingpublictransittripsoverthenetworkofstations;nor withtheinteractionofbus andcartrafficonroads.Thisisnecessitatedbydatalimitations,andshouldhaveminoreffects ontheresultsbecauseoftherelativescarcityofpublictransitintheLAregion.
Resultsaredriven bytheinteractionofvariouselasticities:inthetravelmarket thelowcrosselasticityofthedemand forpublic transit with respect to car travel times, andthe much higherelasticity of switching amongalternative travel routes;inthehousingmarket,therentelasticityofresidencelocation,therentelasticityofthedemandforhousingsize, theelasticityofvacancyreductionwithrespecttorentandthenewconstructionelasticitywithrespecttohousing;andin thelabormarket,thewageelasticityofjoblocationchoicebyconsumersandthewageelasticityoflabordemandbyfirms. Weshallseenext,thesourcesoftheestimatesoftheseelasticities.
2.6.Dataandcalibration
2.6.1. Data
Table1showssalientdataoftheregionby County,andTable 2showstheaggregationoftheworker andnon-worker populationsby incomeandthecalibratedelasticities. Theyear-2000USCensuswasused forhousing,populationand re-latedsocioeconomic datasuch as unearnedincomeby residenceat thecensus tractlevel, thenaggregated tothe model
6 Agriculture(includingforestry,fishingandmining),FinanceInsuranceandRealEstate,Manufacturing,PublicAdministration,RetailTrade,Services,
TransportationandWarehousing,Utilities,andWholesaleTrade.
7 Thisintra-zonalmodelingofcongestion,usingtrafficdensityoverthezonalroadarearatherthanadetailedstreetnetworkwithinthezoneis
Table1
SalientdataoftheregionbyCounty(baselineyear2000).
PercentShareofCounty
TOTAL Imperial LA Orange Riverside SanBern. Ventura
Landavailable(sq.ft) 31,143,832,948 57.4% 15.9% 1.0% 11.5% 13.5% 0.7% Residentialfloor(sq.ft) 6,941,415,080 0.4% 53.3% 14.0% 14.6% 12.4% 5.3% Commercialfloor(sq.ft) 7,918,442,323 1.0% 55.1% 19.8% 9.6% 10.6% 3.9%
Residents 11,872,702 0.5% 56.9% 19.3% 8.6% 9.7% 5.0%
Jobs 6,580,287 0.4% 59.6% 19.8% 7.4% 8.4% 4.4%
Dailyperson-tripsbyorigin 24,574,359 0.4% 58.1% 21.0% 8.4% 8.3% 3.8%
Commutes 6,580,287 0.4% 57.1% 19.3% 8.5% 9.7% 5.0%
Non-worktrips 17,994,072 0.4% 58.4% 21.7% 8.4% 7.7% 3.4%
Cartrips 17,936,023 0.4% 57.3% 17.7% 10.3% 9.8% 4.5%
Publictransittrips 425,960 0.04% 86.2% 8.1% 1.8% 2.9% 1.0%
Dailypersontripsbydestination 24,574,359 0.4% 58.7% 21.7% 8.0% 7.6% 3.6%
Commutes 6,580,287 0.4% 59.6% 19.8% 7.4% 8.4% 4.4%
Non-worktrips 17,994,072 0.4% 58.4% 22.4% 8.2% 7.3% 3.3%
Cartrips 17,936,023 0.4% 58.1% 18.6% 9.7% 8.9% 4.3%
Publictransittrips 425,960 0.1% 88.0% 8.0% 1.3% 1.8% 0.8%
Table2
Calibration.
Incomegroups
1 2 3 4 Total
Workersandnon-workers 7,520,192(63%) 2,262,441(19%) 1,160,358(10%) 929,710(8%) 11,872,702(100%)
Percentingroupwhowork 60% 48% 44% 51% 55%
Averageworkerincome($/yr) 20,327 46,715 67,426 137,157
Averagewage($/hr) 7.9 19.6 30.7 60.2
MRS(after-taxdisposableincome& commutetime)($/hr)
4.3 19.7 42.7 86.9
Expenditureshareofhousing 0.333 0.376 0.407 0.420
ELASTICITY
Locationdemandw.r.t.traveltime -0.05 -0.05 -0.04 -0.05
Locationdemandw.r.t.rent -0.62 -0.47 -0.39 -0.43
Laborsupplyw.r.t.wage +1.72 +1.17 +0.96 +1.01
Labordemandw.r.t.wage -0.02to-0.61byindustry(-0.11onaverage)
Carchoicew.r.tcartime -0.10
Routechoicew.r.t.generalizedcostof roadtravel
-1.53
Buildingtypes
ELASTICITY SingleFamily Multiplefamily Commercial ndustrial Public
Occupiedfloorsupplyw.r.t.rent +0.10 +0.10 +0.10 +0.10 +0.10
Constructionw.r.t.floorprice
Imperial +3.69 +1.57 +6.87 +15.40 +3.78
LosAngeles +2.89 +0.71 +1.15 +1.28 +1.09
Orange +5.39 +6.41 +5.34 +4.71 +9.01
Riverside +7.10 +3.61 +1.24 +0.53 +0.13
SanBernardino +2.72 +4.93 +2.59 +0.56 +0.21
Ventura +3.69 +6.15 +4.61 +2.87 +2.40
zones. Thethree partsofthe CensusTransportationPlanning Package(CTPP) for2000 providedinformationon commute tripsfromcensustracttocensustractandbyoriginanddestinationoftrips,andwasutilizedtoestimatethecommuting, job location andresidential location choicerelationshipsaswell asearnings andindustry composition by workplace.The Southern CaliforniaAssociationofGovernments(SCAG)homeinterviewtravelsurvey wasutilizedtocalibratethemodel’s non-work trippatterns;land usedatafromSCAGwasusedtodetermine developablelandandland usedbythe building typesineach zone.IMPLAN’s8 County-basedinter-industryflows,industry basedexports andother economic interactions
were usedtocalibratetheproductionfunctions, inputcost sharesandexportquantities.Propertyvaluesandfloorspaces bybuildingtypeandzonewereobtainedfromDataquick’spropertyrecords.9
8http://implan.com/.
2.6.2. Calibrationstrategy
The model’s calibrationfrom the above data sources blends several approaches: (i) setting certain parameters exoge-nously;(ii)settingthevaluesofalternative-specificconstantsintheutility andproductionfunctionstoperfectlyreplicate observedmarketsharesforconsumersandproducers;(iii)calibratingotherparameters toperfectlymatchtargetelasticity valuesextractedfromeconometricestimatesintheextantliterature;and(iv)findingaclosematchbetweentheobserved zone-to-zoneroadtraveltimesfromtheCTPPandthezone-to-zonetraveltimesproducedbytrialanderrorsimulationsof themodel.Asummarydescriptionofthiscalibrationstrategyfollows.
2.6.3. Elasticities
Setexogenouslyared=250workdaysperyear;H=8hoursofworkperday;
ρ
=0.02realinterestrate;1.2average passengersper car. Theconsumer’schoice probability isPijk| f (see Eq.(A.7)inAppendix) wherefisthe worker’s type by income-skilllevel,(i,j,k)isthediscretechoicebundle ofresidencelocation i,joblocationjandhousingtypek.Asiswell known,thechoiceprobabilitiescanbeperfectlymatchedtotherelativefrequenciesinthedatabyuniquelycalibratingthe alternativespecificconstantsofthemultinomiallogitmodel(Anas(1983)).Meanwhile,theelasticityofPijk| fwithrespectto anattribute xijkf intheindirectutilityU˜i jk|f,isη
x:i jk f ≡λ
f(
1−Pi jk|f)
xi jk f∂˜ Ui jk|f
∂xi jk f,where
λ
fis thedispersionoridiosyncratic heterogeneity coefficient of the multinomial logit model for worker typef. The probability-weighted average elasticity is thenη
x: f≡λ
fi jk
Pi jk|f
(
1−Pi jk|f)
xi jk f∂U˜ i jk|f∂xi jk f.Theprobability-weightedelasticityoflaborsupply(LS)withrespecttowage,the elasticityofresidentiallocationdemand(RLD)withrespecttorent,andwithrespecttocommutetime(CT)are:
η
LS: f≡λ
f
i jk
Pi jk|f
1−Pi jk|f
wj fHdMi j f
(
1−
τ
j f)
, (1)η
RLD: f≡λ
f(
1−α
f)
i jk
Pi jk|f
1−Pi jk|f
, (2)
η
CT: f≡λ
fγ
f
i jk
Gi jPi jk|f
1−Pi jk|f
, (3)
wherewjfisthewageforincome/skillgroupfemployedinzonej.1−
α
fistheshareofdisposableincomespentonhousing andγ
f (<0)isthemarginaldisutilityofcommutingtimeandequivalently,−γ
fisthemarginalutilityofleisureassuming commutingdecreasesleisure.GivenobservedvaluesofthebaselinePijk| fasrelativefrequencies,andofannual after-income-taxwageincomeswjfHd(1−τ
jf) 10,ofdisposableincomesMijfandcommutingtimesGij,parametersλ
f,γ
f,0<α
f<1are calibratedto matchelasticities (1), (2) and(3)for each income/skillgroup f. Maximumlikelihood estimation of a mode andresidentiallocationchoicemodelbyIndra(2014)for275U.S.MSAsyieldedthattheelasticityoflocationdemandwith respecttoresidentialrent,(2),acrossMSAsdeclinedwithincome,from-0.62to-0.43inLA,andtheelasticityofresidential locationdemand withrespecttocommutingtime, (3),inLAstayed inthe narrowrange-0.04to-0.05. Indra(2014) also foundthattheownelasticityofchoosingcarwithrespecttoits time-costforLAis-0.1,andthat thecrosselasticitywith respecttopublictransitisabouthalfthat.UsingU.S.national databetween1983and1986,KimmelandKniesner(1998)estimatedtheaveragelaborsupply elas-ticityas0.51,ifall wagesrosetogether.The choiceelasticity(1)atthe intra-urbanlevel, yieldsa highervalue becauseit measurestheincreaseinlaborsupplytoazoneonaverage,ifwagesinthatzoneoftheregionrose,keepingwagesinthe otherzonesconstant.There isnoeasywaytodecidehowmuchhigherthisintra-urbanelasticity(1)should berelativeto thenationalaggregatelaborsupplyelasticity.Aftersomeexperimentationwithsimulationresultsusingdifferentelasticities, wesettledonthevaluesinTable2whichdecreasewithworkerincomefrom1.72toabout1.0.
2.6.4. Thevalueoftime
Thevalueoftimeintravel,vot,issetathalftheaveragehourlywage(Small,1992):11
v
ot=1 2
f
⎛
⎝
NfPef
f
NfPef
i,j>0 ,k
Pi jk|fwj f
⎞
⎠
=$7.61, (4)whereNf is thenumberoftype-fconsumersandPef is thesharethat work. While(4) isusedfortravel on thenetwork, affectingroutechoicesdirectly,andmodechoicesindirectly,thevotalsoindirectlyaffectsthehigherleveljobandresidence
10 τjfisthecombinedFederalandStateincometaxrateapplicabletoincomegroupfcalculatedasaCountyaverageinthebasedata,attributedtothe
zonesjoftheCounty.TaxesarediscussedinSection4.
11 Perfectcongestionpricingrequiresknowingthevotofeachtravelerusingeachroadthatistolled.Ahighertollwouldbeleviedontripsthatdelay
locationchoicesbecausethecommutetimes,Gij,andmonetary costs,gij,weightedbythemodechoice probabilities,enter thechoiceprobabilitiesPi jk|f asexplainedintheAppendix.Fromtheindirectutilityfunction,theprobability-weightedMRS (marginalrateofsubstitution)betweenafter-taxdisposableincomeMijf andcommutingtimeis
MRS
Mi j f,Gi j=
γ
f
i,j>0 .k
Pi jk|fMi j f. (5)
Table2 includesarepresentative calculationofthe MRS.Thelowest incomegroupvaluesits leisuretime atbelowits wagerateonaverage, thesecond tolowest ataboutitswagerateandthehighergroupsatincreasingly morethan their wagerates.
2.6.5. Theelasticityofsubstitutionamongretaillocations
Fortheconsumer’selasticityofsubstitution 1 −1 σ
f amongretaillocations,wehaveexperimentedwithvaluesintherange
0.17and 2,settling on 2 (
σ
f = 0.5 forall f), whichhelps matchthe data ratioof 2.73 non-work trips per-work tripon aggregate.Tocompletetheconsumer-relatedcalibration,fixedeffects(constants)Eijk| f,ι
z| ijkf,Izr| i,sizr(seeAppendixfortheir appearanceintheequations)aresettomatchchoicefrequencies,non-worktrippatternsandtherelativequantitydemands ofpurchasedgoods.Bytrialanderror,theratioofthetravelcosttopurchasecompositegoodstothe(mill)pricesatretail locationsisabout10%,appearingreasonable.2.6.6. Productionandlabordemand
Inproduction,costsharesofinput groupsvarybyCounty andindustry basedoninput-output dataobtainedfrom IM-PLAN. The elasticity of substitution among floor space inputs in production is set as 1 −1 ζ
r =3, among zone-specific
in-termediate inputsfrom each of the other industries as 1 −1 ε
r =2.5, and amonglabor typesas 1
1 −θr =1.81. Maiti and
In-dra(2016) estimatedthelabor demandelasticitiesfordifferentindustriesacross allU.S.Counties butnot byskilllevelof workers.Labordemandelasticityintheirfourindustriesaredistributedlognormalwithmeansandstandarddeviations(in parentheses)asfollows:-0.15(0.13)forconstruction,-0.005(0.05)forfinance-insurance-real-estateandservices,-0.21(0.38) formanufacturing,and-0.09(0.25) forretailtrade.Thescaleparameterofthelognormalwassignificant forallfour indus-tries,confirmingthespatialheterogeneityinthelabordemandelasticity.Inan internationalstudy,Lichteretal.(2015) re-portedameanof-0.50 (0.77)fortheoverall labor demandelasticity.Theyreportedmeanvaluesbytime period,dataset, workforce, industry,andcountry.Intheshortrun,their meanvalue is-0.21(0.40) whileinthelongrun itis-0.34 (0.47). Attheindustrylevelitis-0.53(0.49),-0.04(0.20),and-0.05(0.23)forthemanufacturing,service,andconstructionsectors, respectively.Theydidnotreportestimatesforretailtrade.TheestimatesbyMaitiandIndra(2016)foralltheindustriesfall intherangesreportedinLichteretal.(2015)albeitontheirlower sides.ForoursixCountiesandourindustries(treating constructionbybuildingtypeasasingleindustry),MaitiandIndra’slabordemandelasticityvariedbetween-0.02and-0.61. MatchingtheirestimatesbyindustryandCounty,theacross-industriesaveragelabordemandelasticityinTable2is-0.11.
2.6.7. Exportdemands
Exportdemandelasticitiesfortheprimaryindustriesr=1,...9weresetasfollows:First,wecalculatedthenationaland regionalemploymentshare,respectively,foreachindustry.Wesetabaselineexportelasticityto-0.5.Thisvalueisarbitrary aswedonothaveanyreferenceinformationforit.Then,wesettheexportdemandelasticityforindustryrbymultiplying -0.5by theratio ofthenational job shareofindustry rto itsLA regionjob share.Ifthe industry’s shareislower inthe LAregionthanintheUS,itfacesamoreprice-elasticexportdemand,becausetheimporterswouldhavemoresubstitutes available to them from other regions. Set in this manner, the export demand elasticity varies froma low of -0.093 for agriculturetoahighof-0.806forretailtrade.Theconstants
κ
f|rz,χ
k|rz,andυ
sn|rzintheproductionfunction(seeEq.(A.10) intheAppendix),aresettoobtainlaborquantities,floorspacesandinter-industryflowsconsistentwithIMPLANdata.Total factorproductivityconstants,Arz,matchnominaloutputbyindustryandCountyintheIMPLANdata.2.6.8. Housingandbuildingsupplies
InthebottompartofTable2,thecalibrationofhousingsupplyisreported.Intheshortrun,thelandlord’svacancy reduc-ingsupplymodelrules.Ourestimateshereyieldanelasticityof+0.10forhousing.Weassumethesameelasticityforother buildingtypesbecauseofthelackofdataonnon-residentialvacancyrates.Inthecaseofconstruction,theannual construc-tionelasticitieswithrespecttothepriceoffloorspaceinTable2wereestimatedfromaggregatedataforeachCountyofthe LAregion.Indra(2014) obtaineda similaraverageelasticityasthatshowninTable2forsinglefamilyconstruction,using microdataonconstructionintheperiod1998-2012.Weusedtheratioofthesinglefamilyhousingmicro-dataconstruction elasticitytotheaggregate-dataconstructionelasticityforsingle-familyhousingtoadjusttheaggregate-dataelasticitiesfor non-singlefamilybuildings.
2.6.9. Congestion
Forthemodeltoproducea reasonablelevelofcongestionandtomatchwell theobservedtravel times,the Bureauof PublicRoads(BPR)typeofflowcongestionfunctiongivenbyEq.(A.32)intheAppendixiscalibrated.Weusethestandard
notexclusively.12 Meanwhile,the capacityofeach arca (capacity
a inEq.(A.32)) isnot a physicaldesigncapacity,butan effectivecapacityimplied by roadwidths andconditions,slope, curvature,densityof accessandegress ramps andother factors.We fittheequilibriumzone-to-zoneexpectedtraveltimes,tij,fromtheTRANsub-modeltothezone-to-zonedaily averagedatatraveltimes,ti j0 ,usingtheAveragePercentageError(APE)asourmeasure,whereNo
i j areobservedtrips:APE=
i j[
(
Noi j/kmNkmo
)(
|
ti j−ti jo|
/ti jo)
]. Given any value of c > 1, the exponent in the congestion function Eq. (A.32), APE is optimizedwith respectto the capacities by trialand error,yielding an APE of about10%. To pindown a value of c, we alsorequirethattheregion-widecongestionpermilebeconsistentwiththeTexasA&MTransportationInstitute(TTI), year-2000 congestionindexforthe LAregion (Schrank etal., 2015).The TTI indexisdefined astheratioof actual triptimes duringpeak hours,reported by asampleof drivers,to theuncongested (or free-flow)travel timesthe sametrips would experience.Hence, it isan average per-milemeasure over arandom sampleoftrips. In theyear2000,TTI reported1.41 fortheLosAngeles-LongBeach-AnaheimMSAand1.26fortheRiverside-SanBernardinoMSA.WeightingthesebytheMSA populationshares of roughly 75% and25%, ourTTI indexfor theregion is 1.3725. We adjustthisdownward because,as notedearlier,our modelgenerates travel timesonly foraverage-over-the-day, not peak conditions.Assuming an 18 hour traveldayand6peak hours,weassignaweightofa1/3topeak.SinceTTIdoesnotreportanoff-peakindex,we assume avalue 20%overfree-flow.Thisproduces anaverageTTIindexof1.262. Acongestionfunction exponentofc= 1.2inEq. (A.32)matchesthisTTIindex.2.7.Algorithmforsolvingthemodel
As explained indetail in Anasand Liu(2007), RELU-TRANis converged to equilibriumby cycling betweenthe TRAN modelandthe RELUmodel.Thatpaperincluded descriptionsof thealgorithm inthe formofflow charts,equationsand convergencecriteria.ThealgorithmhasbeenimprovedbyutilizingtheGeneralAlgebraicModelingSystem(GAMS).13
Thecyclicalprocedure isdescribedindetailinsection FoftheAppendix,wheretheequationsthat aresolved ineach oftheblock-recursivesteps arepresentedandcounted.Here,an intuitivesummaryisgiven.First,theTRANmodelobtains azone-to-zone person-trip matrixfromRELUandperforms the stochasticnetwork equilibriumassignment ofthesetrips on the roadnetwork, obtaining congested travel times andmonetary travel costs. Theseare then passedto RELU. RELU exploitsthe block recursivenatureof theeconomicandland userelationshipsandperformsa loopofthesecalculations. Giveninitialestimatesofwages andrents,thepricesofproducts are calculatedusingthezero-profit equationsofall the producersinallthezones,andgiventheseprices,outputsofallproductsinallthezonesarecalculated.Nextthebuilding andlabormarkets inallthezonesareequilibratedandwagesandrentsarethusupdated.From therents,buildingvalues arecalculatedforeachzoneandbuildingtype usinginvestors’zero-profitequations.Then, thebuildingvaluesareusedto calculatehowmuchconstructionofeachbuildingtype’sfloorspaceoccursineachzone.Finally,usingtherelevantresults fromtheforegoingcalculations,RELUcalculatesazone-to-zonepersontripmatrixthat isagainpassedtoTRAN,startinga newcycle.RELUandTRANarecycledinthiswayuntilequilibriumisreached.
3. Thebenefitsofcongestionpricing
3.1. Welfareanalysis
Weturn tothe costofthe congestionexternality, andthebenefits ofpricingit.The modeldoesincludethe effectof congestionon gasoline asexplained in the Appendix. But when calculating congestiontolls, we price only forthe time delaywhichcauses the bulk ofthe externality,ignoring the partstemming fromthe excessive useofgasoline. From Eq. (A.32),ourPigouviantollonamodelroadaisthegapbetweenthemarginalsocialandtheaverageprivatecostsoftravel:
tolla≡MSCa−APCa=
v
ot·c·ta0 baf lowa
capacitya
c, (6a)
Theabovetollfunction isaddedtothemonetarycostoftravelonroada,andequilibriumflowsarecalculatedso that theequilibriumtollssatisfy(6a).Thetotalexternalityontheroadaistolla × f lowaandthetotalregion-wideexternality,
a
tolla×f lowa,is:
Total externality=
v
ot·c·a
ta0 ba
(
f lowa
)
c+1
(
capacitya)
c. (6b)
Togetthewelfarechange,
W,fromcongestionpricing,withoutrecyclingthetollrevenue,wesum:(i)thecompensating variationsoftheregion’sworkersandnon-workers;(ii)thecompensatingvariationofanoutsiderepresentativeconsumer
12 Forexample,theGreaterParisregion(IledeFrance)hasadetailednetworkofabout30,000arcsandeacharcisassignedadistinctvalueofthe
exponentc,varyingfromaslowasonetoashighastwelve.
who imports fromthe region; (iii) the annualizedchange in real estatevalues in the region; (iv) the revenue fromthe congestiontoll;(v)thechangeinrevenuefromthetaxesonincomes,sales,propertyandwages.14 (i)-(v)areallexpressed
perconsumerintheregion,andthenumberofconsumersiskeptconstant:
W =CVcons,+CVimp.+
ρ
N(
RealEstateValues
)
+N1(
TollRev
enue+TaxRe
v
enues)
, (7)N=
f
Nf istotalconsumers,thesumofworkersandnon-workersbyskill/incomegroupf.TheCViswhataconsumer
wouldpayinmonetary unitsfortheincrease,orrequireascompensationforthedecrease,inexpectedutilityarisingfrom thecongestiontollingpolicy.From themultinomiallogitcalculus,theexpectedutilityofa representativeworker(e),EUe
f, ofskill/incomegroupfinthebaseequilibriumbeforetollingis:
EUef,Base= 1
λ
fln
i jk
exp
λ
fU˜i jke,Base|f. (8)
Post-pricing,indirectutilityisU˜i jke,Post|f =u˜ei jk,Post|f +lnMi j fe,Post,whereMi j fe,Postisthedisposableincomeatequilibriumandu˜ei jk,Post|f istherestoftheindirectutility.TheCVe
f oftheworkerofskill/incometypefisthensolvedfrom:15
EUef,Base=
λ
1f
ln
i jk
exp
λ
f˜
uei jk,Post|f +ln
Mei j,|Postf −CVfe
, (9)
andsimilarlywesolvefortheCVu
f,ofanon-worker(u).Theweighted-averagepercapitaCVin(7)isthen:
CVcons.= 1
f
Nf
f
Nf·Pef·CVfe+
f
Nf·
1−Pe f
·CVu f
, (10)
wherePe
f istheexogenousshareofthetotalworkersNfingroupf.TheCVimp. ofimporters,perconsumerintheregion,are
calculatedusingarepresentativeconsumermodelforimporters(seeEq.(A.9)intheAppendix).
3.2. Pricingcongestionwithoutrecyclingthetollrevenue
Wenowreportontwosimulationsinwhichthecongestiontollrevenueisincludedaspartofwelfareasshownin(7), not recycledback into theeconomy. Recycling willbe examined inSection 4. Inthe firstsimulation, acongestion tollis leviedonall tripsthatusethe285arcsoriginatingand/orendinginLACountyandonallintra-zonaltripsintheCounty. LACountyisthemostcongestedandcentralpartoftheregion,containing56.9%ofthepopulation and59.6%ofthejobs. Inthesecondsimulation,all696arcsandallintra-zonaltripsintheentiresix-Countyregionaretolled.Resultsofthetwo simulationsarejuxtaposedinTables3aand3b.Table3adisplaysthecongestiondelayexternality,changesintotalwelfare anditscomponents,distributionofCVbyincomegroupforworkersandnon-workers,travel-relatedchangesandthechange intheregionalgrossproduct.Table3bshowschangesinthedistributionofpopulationandjobs,andinrentsandwagesby County.
3.2.1. Externality,tollrevenueandwelfareincrease
From Table 3a,when only LACounty is tolled,congestionis reducedandthe delayexternality dropsfrom $ 550per consumerper yearinthe baselineto $462.The flow-weightedaverage congestiontoll is$ 254andinternalizes46.2% of thebaselinedelayexternality.Totalannualwelfareincreasesby$236.Whentollingthewholeregion,thedelayexternality drops to$ 398andis fullyinternalized bythe tollsandwelfareincreasesto $ 350. Inboth simulations,there isa small negativeeffectongrossregionalproduct.
3.2.2. Travel-relatedvariables
AsshowninTable3a,congestion,measuredastheflow-weightedaverageoftheflow-to-capacityratiooverallarcsand intra-zonalroads,decreasesby10.39%.Theaverageautotimedecreasesby1.21%(theaverageflow-weightedspeedincreases by1.41%).Cartripsandnon-worktripsdecreaseby2.91%and2.88%respectively,(carshareoftripsdecreasesbyonly0.60% duetothelowavailability ofpublictransit).VMT(VehicleMilesTraveled)decreaseby3.74%andgasolineconsumptionby 3.04%, more thantrips decrease,because consumers andfirms react to thecongestion pricingby moving closerto each other toreduce commutingdistance,andby consumersmaking shorternon-work trips.When thewhole regionistolled
14 Duetodatalimitations,weignoretherevenuechangeinthepublictransitsector,butinthecaseoftheLAregion,includingitwouldhaveonlyslight
effectsonourresults.
15HerrigesandKling(1999)comparedthecalculationofCVbythismethodtoitscalculationbymicrosimulation,anapproachthatwaslaterformalized
Table3a
Benefitsofcongestionpricingperconsumer,travelrelatedchanges,grossproduct(NOTES:Allchangesare rel-ativetoBase;$values>$100roundedtothenearestinteger).
TollinginLACounty Tollingentireregion
WELFARECHANGE$/year/consumer=a+b+c+d+e 236 350
a.ConsumerCV -12.39 -26.72
b.Importer’sCV 0.65 5.00
c.Realestatevalue(annualized) -3.01 -14.09 d.Taxrevenue
Income -0.61 -2.94
Sales -2.29 -6.70
Property -0.29 -2.20
Wage -0.17 0.19
e.Congestiontoll 254 398
f.Congestiondelayexternality(Base=550) 462 398 Un-internalizedcongestiondelayexternality=f-e 208 0 DISTRIBUTIONOFCV($/year/consumer)
Workersingroup1(lowestincome) -37.55 -65.68
2 -5.83 -33.58
3 21.31 1.14
4(highestincome)) 69.17 51.66
Non-workersingroup1(lowestincome) -2.11 -1.56
2 -3.49 -2.30
3 -2.66 -1.51
4(highestincome) -4.57 -1.91
TRAVELRELATED Percentchanges
Trips -2.11 -2.87
Non-worktrips -2.88 -3.92
Cartrips -2.91 -3.97
Carmodeshare -0.60 -0.83
Caraveragetraveltime -1.21 -1.44
Averagetraveltime -1.12 -1.35
Congestion(flow/capacity) -10.39 -14.29
Gasoline -3.04 -4.45
VMT -3.74 -5.30
\
GROSSREGIONALPRODUCT -0.03 -0.10
Table3b
Changesinjobs,population,wagesandrentsduetocongestiontolling.
TollinginLACounty Tollingentireregion
POPULATION
Imperial -18 -13
LosAngeles -2,465 +1,103
Orange +2,092 -1,127
Riverside -1,003 -1,324
SanBernardino +1,023 +843
Ventura +371 +518
JOBS
Imperial +18 +36
LosAngeles -2,299 +216
Orange +950 -1,546
Riverside +711 +535
SanBernardino +428 +388
Ventura +192 +371
WAGES(Averageoverallskill/incomegroups) Percentchanges
Imperial -0.01 -0.11
LosAngeles +0.03 -0.12
Orange -0.13 -0.04
Riverside +0.12 +0.00
SanBernardino -0.13 -0.22
Ventura -0.11 -0.20
RENTS(Averageoverbuildingtypes) Residential Commercial Residential Commercial
Imperial +0.46 -0.01 +0.43 -0.04
LosAngeles -0.14 -0.02 -0.18 -0.08
Orange -0.01 -0.04 -0.26 -0.07
Riverside +0.41 -0.01 +0.29 -0.10
SanBernardino +0.03 +0.04 -0.09 -0.02
Fig.3. The25mostcongestedroadlinksinthebaseline.
changesshowninTable3aarelarger, butlessthantwice thechangesoftollingLACounty alone,eventhoughmorethan twiceasmanyroadsaretolled,becauseLACounty istheoriginanddestinationofmosttrips andhasthemostcongested roadswhiletherestoftheregionhaslesscongestedandevensomeuncongestedroads.The25mostcongestedmodelroads areall inLACountyandshowninFig.3.Table4juxtaposesthepervehicle-miletollsonthese25mostcongested model roadsunderregion-wideandLACountypricing,rankedbythetolltheyreceiveunderregion-widepricing.Thetollsdonot exceed33.7centspermile.
A.
Anas
/
Tr
ansport
a
tion
R
esear
ch
Pa
rt
B
13
6
(2020)
11
0
–
13
7
12
3
Tollspermileonthemostcongestedmodelroadsofthenetwork.
ArcNo. Freewayormajorroadofarc Freewayorroadnumber Arc’slocationfrom… Arc’slocationto… LACountyTolling Region-wideTolling $/mile $/mile
89 SanGabrielRiverFr. 605 91,ArtesiaFr. DelAmoBlvd. 0.174 0.337 43 SanDiego 405 WilshireBlvd. OhioAve. 0.307 0.307 19 Century 105 405,SanDiegoFr. HawthorneBlvd. 0.299 0.300 68 HawthorneBlvd. 7 W.CarsonStr. PacificCoastHwy. 0.295
34 SanDiego 405 2 10 0.285 0.286
24 SanDiego 405 187 90 0.284 0.284
32 SanDiego 405 10 187 0.262 0.262
145 Riverside 91 5,SantaAnaFr. N.EuclidStr. 0.217 29 SanDiego 405 Westchester 105,Century 0.211 0.211 140 Artesia 91 ValleyViewStr. 39,BeachBlvd. 0.210 25 SanDiego 405 90 W.ManchesterAve. 0.188 0.188 229 SanBernard. 215 66,FoothillBlvd. W.9thStr. 0.182 73 Harbor 110 405,SanDiegoFr. 91,W.ArtesiaBlvd. 0.295 0.176
208 0.174
Fig.4. Sixmodelzoneswiththemostjobs.
(NOTE:Modelzonenamesidentifythemajorcityorareacontainedwithinthezone).
transit inresponse to tollingcartrips. Where switchingto publictransit is feasible,trips are onaverage shorterin such places.Hence,theshareoftheshortercartripsisreducedwhichkeepsaveragecartraveltimesfromdecreasingalot.Third, thetollpromptssomeconsumerstoeconomizebyusinglesscongested(andlesstolled)butsometimeslongerroutes.Those lesscongestedroutes,towhichsomedriversarediverted,becomemorecongested,butstillremainlesscongestedthanthe roadsthattravelersswitchedfrom.Thehighlyelasticrouteswitchingbehaviormoderatestheshorteningoftraveltimesin commutesandshoppingtrips.
3.2.3. Locationofjobsandresidences
modelzonebyjobs,containedonly6.4%.PricingroadsinjustthecentrallylocatedLACountymakesdrivingontheoutskirts lessexpensiverelativetodrivingtoandfromthecenter,inducingsomepopulationandjobstorelocateoutward.Butwhen congestioninallCountiesispriced,therelativeattractivenessofresiding,shopping,andworkingintheoutskirtsdiminishes comparedtowhenonlyLACountyroadsarepriced.
Theseresults are in Table 3b.Pricing LA County roads only slightly reduces LACounty population andjobs andalso Riversidepopulation,asmanyLACountyworkersresideinRiverside.JobsandconsumersspreadoutslightlyfromLACounty to theperiphery. Orange County is themain recipient of the outflowof population becauseit is nearLA andRiverside, servingastheclosestsubstitutefortoll-avoidance.Whenthewholeregionistolled,theeffectsaregenerallyreversedinthe sensethat someconsumerswho commute intoLACountyfromOrangeandRiversiderelocatetoLACounty orelsewhere toshortentheircommutingandblunttheeffectofthetolls.SomejobsrelocateoutofOrangeCounty.Inbothsimulations, wagesandrentsdecreasebutonlyslightly.
Itis interesting tothink abouthowthe populationandjob changes are relatedtothe wageandrentchanges. Taking OrangeCountyasanexample,wecanseefromTable3bthat,whenonlyLACountyroadsaretolled,Orangeresidentsand jobsincreaseby2,092and950respectively.TollavoidanceinducestheOrangeresidentstoprefertoworkinOrangewhich increaseslaborsupplytoOrange,causingwagestofallby0.13%becausetheincreaseinlabordemandisnotasstrongasthe increaseinlaborsupply.Meanwhileresidential rentsdecreaseveryslightlyby0.01%becausethehousingdemandincrease bythepopulationincreaseisoffsetbytheincomeeffectofthewagedecreaseonrents.InthesecondsimulationofTable3b, whenalltheroadsoftheregionaretolled,Orangeresidentsandjobsdecreaseby1,127and1,546respectively.Asfirmsand workersmoveout, thedemandforlabor andthesupplyoflabor decreasecausingwagesto fallslightlyby 0.04%.Inthis case,thedropinthedemandforhousingworkstogetherwiththeslightdropinwagestocauseresidentialrentstofallby 0.26%.
3.2.4. Consumerutility
LevyingcongestiontollsonLACountyroadsinducesbroadnegativeincomeeffectswhichsetoff changesindrivingand inshoppingdemandpatterns.ThetollsreducethedisposableincomeofworkerswhocommutetoorfromLACounty,which reducesconsumerdemandandcausesthenumberofnon-worktripstofall.AconsumerwhoresidesorshopsinLACounty isdiscouraged by the tollsandshops more atcloser-in locations, making fewer long-distance shopping trips. Weakened consumerdemand,inturn,causesfactordemandstofallandcommercialrentsandwagestobelowered.Asloweredwages reducedisposableincomes,housingrentsfalltoo.
3.2.5. Compositionofwelfare
Onaverage,thenegativeincomeeffectofcongestiontollingdominatesthebeneficialeffectonutilityfromthe external-ityreduction.Consequently,in Table3a,the averageconsumerCV isnegativebutonlyslightlyso.Real estatevaluesand revenuesfromothertaxesalsofall,butalltheseeffectsaresmall,indicatingnoimportantinteraction betweencongestion tollingandtheothercomponentsofwelfare.Revenuefromcongestiontollingexceedsthedecreaseofotherpartsofwelfare, andoverallwelfareincreasesastollingisextendedtotheentireregion.
Table3a includesthedistribution ofthebenefits oftollingby incomegroup andforworkers andnon-workers. When tollingLACounty,theCVisnegativeforlowerincomeandpositiveforhigherincomeworkers.Thisisbecausewhilethetoll hasa negativeincomeeffectonallworkersasexplainedabove,thericherhavea higherMRSbetweendisposableincome andtravel time (Table 2), and so their benefits from time savings outweigh thenegative income effectof the tolls. For thepoorer,thenegativeincomeeffectdominates.Whentollingtheentireregion,thenegativeincomeeffectissomewhat stronger.Fornon-workerstheCVisnegativeandsmall.Thereasonisthatnon-workersdonotbenefitfromcommutetime reductionsbutpaytollswhen making other trips. Nor are they exposed towage changes sincethey do not earnwages. The indirecteffects ofcongestion tollingon them islargely due torent changes. From Table 3b,where residential rents increase (decrease) slightly due to tolling, they increase less (decrease more) under region-wide than under LA County tolling.
Whyiscongestiontollingrevenueashighasitis?Ifallresponses tocongestiontollingwereperfectlyinelastic,theper capitatollwouldequalthebaselinepercapitacongestionexternalityof$550.Whentheentireregionistolledthepercapita tollis$398.Hencetollavoidancebehaviorreducesthepercapitacongestionexternalityby28%.Recallthat,inthemodel, aggregate labor supply andaggregate commute trips are fixed. Yet avoidance behavior occursin several other margins: choosing less congested routes, relocating residence or jobs, shopping at less congested locations, making shorter/fewer non-worktrips,andusingslightlymorepublictransit.Moreover,fromTable3a,thedecreaseinnon-worktripsislimited. Relativelyinelastic(elastic)responses acrossseveralmarginsmeanthat thetollshavetoberelativelyhigh(low)toinduce optimalbehavior.Meanwhile, themagnitudeofthecongestiontollsdependscruciallyonthevalue oftime (vot)in travel usedinthemodel.Toexplorethisaspectweconductedsensitivityanalysis.
3.3.SensitivitytotheValueofTime(vot)
Fig.5. SensitivityofthewelfarecomponentstotheValueofTime(vot).
to usingan alternativevot.Undercongestionpricing, the trafficflow to roadcapacityratio,markedlydecreases withvot, butthenumberoftrips,averagedrivingtime,averagetraveltimebyall modes,VMT(vehiclemilestraveled),andgasoline consumptionalldecreaserelativelyless.Itmayseemparadoxicalthatthereductionincongestion(flowtocapacityratio)is significantlygreaterthanthechangesintheseothervariables.Thishappensbecauseconsumersrespondelasticallyspreading from the morecongested roads to the lesscongested ones, butsince the average estimated elasticityof location choice withrespecttoaveragetraveltime issmall(-0.04to-0.05),not manyconsumersrelocateaslongasthechangesintravel timescausedbytollingaremoderate.Consequently,drivingdistancesbetweensomeorigin-destinationpairsbecomelonger, bluntinganyinitialdecreaseinVMTduetotripreductions.Sincepublictransitisslowerthandrivinginmostcases,asvot
increasessodoesthedisutilityofridingpublictransit.Thereforethehighercongestiontollonroadsduetothehighervot
causessomeswitchingtopublictransit,yet— atthesametime — thehighervotreducespublictransitdemand.Ahigher
votpushesmorejobsandpopulationoutofLACountyasdelaysassociatedwithdrivinginthecentralregionbecomemore costly.Still,becauseofthelowlocationelasticity,changesinjobsandpopulationinallCountiesremainrelativelysmallin the faceof quadrupling thevot from onequarter ofwageto the full wage.Strikingly, changesin theCV are both stable andsmallasvotincreases.Thisisbecause,ontheonehand,tollingunderahighervotshortensaveragetraveltimesmore, whichimprovesutility;but,ontheotherhand,thetraveltimedisutilitybecomeshigherasvotincreases.Theseeffectsoffset eachother,leavingutilitylittlechanged.Meanwhile,thetollrevenueincreaseswithvotanddrivesupthewelfaregainfrom tolling(Fig.5).
Asequeltothispaperwillexplorethesensitivityoftheresultstootherkey calibratedparameters, inparticulartothe threeelasticitiesshowninEq.(1)–(3),usingMonteCarlosimulation.However,adhocsensitivitytestssuggestthatthetolls aremuchmoresensitivetothechangesinvot,thantotheseotherparameters.
4. Substitutingthecongestiontollrevenuefordistortionarytaxes
We nowturnto recyclingthe revenuefromcongestionpricingto reduce othertaxes. Ourtarget taxesarethe income tax forthelowest incomegroup who workin LACounty,andthe salesandthe propertytax inLACounty. Ineach case congestionis tolledonly in LACounty in themanner explained inthe previous section, butnow theaggregate toll rev-enue replaces part ofthe revenue fromanother tax in LA County so that the aggregate tax revenue from all taxesplus the tollsin the entireregion remains unchangedfromthe baseline level. This is done by endogenouslyadjusting, in an outer loop, the targetedLACounty tax rateso that when themodel convergesto equilibrium, revenue neutrality isalso satisfied.
Howtheincome,salesandpropertytaxesenterthemodel’sequationsisexplainedintheAppendix.Theincometaxis leviedonthetotalwageandnon-wageincomeofthemodel’sconsumers,ataratethatincreaseswithincome,andconsists ofthecombinedFederalandStatetaxrates.Recall,however,that– asexplainedearlier– onlytheincometaxofthelowest quartileworkersemployedinLACountywillbecut.Therationaleforthisistorecyclethecongestiontollrevenueinfavor ofthose workerswho havethehighestmarginal utilityofincome.Thesalestax islevied ad-valoremonthe pricesofall consumptiongoodsattheirplaceofsalebutnotonhousing.Itispaidby theconsumersintheregionaswellasthosein therestoftheworldimportingfromtheregion,anditvariesbyCountyintheregion.Thepropertytaxisleviedannually ad-valoremonresidentialandnon-residentialbuildingandlandprices.
Table5
Compositionofbaseregionaltaxrevenue(mill.$/year)andshares(%).
Sales Income Property Wage TOTAL
Revenue Share Revenue Share Revenue Share Revenue Share Revenue Share
Imperial 132 0.5 102 0.2 348 2.4 37 0.7 619 0.7
LA 15,592 63.0 24,348 59.2 7,440 51.6 3,944 77.2 51,324 60.2 Orange 4,481 18.2 9,778 23.8 2,365 16.4 640 12.6 17,264 20.2 Riverside 1,508 6.1 2,455 6.0 2,133 14.8 131 2.6 6,227 7.3 SanBernardino 1,858 7.6 2,438 5.9 1,297 9.0 201 3.9 5,794 6.8 Ventura 1,068 4.3 2,027 4.9 834 5.8 155 3.0 4,083 4.8
Total 24,638 100.0 41,149 100.0 14,416 100.0 5,107 100.0 85,311 100.0
Share 28.9 48.2 16.9 6.0 100.0
Table6
PercentageaveragetaxratesbyCounty.
Imperial LA Orange Riverside SanBernardino Ventura
Incometaxraterange 10.5-19.2 12.1-23.4 11.9-22.6 11.5-21.03 11.7-21.1 11.2-23.8
Salestaxrate 8.11 9.04 8.01 8.04 8.03 8.14
Prop.taxrate 0.81 0.66 0.61 0.94 0.74 0.67
theaggregateregionaltaxrevenue,followedbythesalestax(28.9%),thepropertytax(16.9%)andthe“wagetax” (6%).16LA
Countygenerates60.2%oftheaggregatetaxrevenueintheregion.Thebaselineaggregaterevenuefromcongestiontolling LACountyisnearly $3.02billionannuallywhichamountsto3.54% ofthe totalrevenueof$85.31billion fromtheother taxesinthebaseline(Table5).
4.1. Incometaxcut
Whenthecongestiontollrevenueisusedtocuttheincometaxofthelowestskill/incomegroupwhoworkinLACounty, thisconfers apositiveincomeeffectmorethancompensatingforthenegativeincomeeffectofthecongestiontoll onthe samegroup.Theaggregatewelfarechangeis$645perconsumer2.73timesthe$236welfarechangethatoccurredbefore thetollrevenuewasrecycled.Aswecan seefromTable7a,thetoll-for-income-taxsubstitution yieldsa 1.34%increasein grossproduct, highestamongall the tax-substitutionpolicies. The higherdisposable income drives up housingdemands (Table7b), whichcauses3.6% higherhousingrentsinLACounty.Demandsforgoods andservicesacquiredvianon-work tripsincreaseenoughtoalmostcompletelyoffsetthedecreaseinsuchtripscausedbythecongestionpricing.
Althoughwe saw that tollingLACounty without recyclingdrove asmall numberofjobsandpopulation to theother Counties(Table3b),the incometaxcut morethan offsetsthis. The consumerpopulationresiding inLACounty increases by48,851andjobsincreaseby78,535whileallother Countieslosepopulationandjobs.Thisdrivesup laborsupplyinLA CountyandreducesitinotherCountiesaslowerincomeworkerschooseLACountyjobstobenefitfromtheincometaxcut. ThelaborsupplyincreasecauseswagesinLACountytofallby0.81%andwagesintheotherCountiestorisebyalmost5to 6%.Thiswageincreaseintheother Countiesoccursbecauselabordemandisquiteinelasticaswesaw.Thewageincreases inturncause peripheralrents to risecausingmore newconstruction than inthe caseoftollingwithout theincome tax cut:singlefamilyhousingconstructionis15%higherinLACounty and6%higherinRiversideCounty,thaninthecaseof norecycling.EventhoughtheLACountyincometaxcutcausesjobsandresidencestoshifttoLACounty,thelabormarket effectintheperipheralCountiescausesmorehousingconstruction,hencemoreurbansprawl,measuredasdevelopedland area.
4.2.Salestaxcut
Thereare importantdifferencesbetweenthe incometaxcut forthe lowestincomegroup workinginLA County only, andthesalestaxcutforall,andincludingforimportersfromtheregion.Thesetwotaxesworkindifferentmarginscausing differentdistortions. While the incometax-cut directly increasesdisposable income (butin our caseonly forthe lowest incomegroupworkersinLACounty),thesalestax-cutreducestheafter-taxpricesofconsumptiongoodsinthemodel ex-ceptbuildings,andindirectlyincreasesafter-taxdisposableincometoo.Importersfromtheregionaswellastheconsumers inthe region benefit directlyfrom the salestax cut. Importers are hurt by the incometax cut, because prices of goods theypurchaseareincreased.Whiletheincometaxcutworksprimarilybyinducingapositiveincomeeffectforthosewho receiveit,thesalestaxcutworksprimarilybyinducingasubstitutioneffectforall.
16 The“wagetax” includessocialsecuritytaxestogetherwithemployeecompensation,benefitsandothercharges.Wejustmodelalloftheseinaggregate
Table7a
Congestiontollingreplacingothertaxes:welfare,travelaggregates,grossproduct(NOTES:ChangesarerelativetoBaseline;$values>$100roundedto nearestinteger;taxrevenueneutrality:d+e=0incolumns2,3and4).
TollinginLACounty (withouttaxsubstitution)
TollsreplacingtaxinLACounty
Incometax(lowincome) SalesTax Property Tax
WELFARECHANGE($/year/cons.)=a+b+c+d+e 236 645 244 235
a.ConsumerCV -12.39 428 88.87 -9.22
b.Importer’sCV 0.65 -44.85 93.62 0.22
c.Realestatevalue -3.01 262 61.41 246
d.Taxrevenue
Income -0.61 -394 16.16 -0.17
Sales -2.29 78.45 -283 -2.23
Property -0.20 44.29 10.37 -253
Wage -0.17 0.56 -0.52 -0.32
e.Congestiontoll 254 271 257 256
f.Congestionexternality(Base=550) 462 479 465 462
Un-internalizedcongestionextern.=f-e 208 208 208 208
DISTRIBUTIONOFCV
Workersingroup1(lowestincome) -37.55 1,121 65.18 -35.11
2 -5.83 91.73 229 1.61
3 21.31 195 350 34.51
4(highestincome) 69.17 367 686 93.05
Nonworkersingroup1(lowestincome) -2.11 -51.13 2.65 -2.07
2 -3.49 -83.90 2.59 -3.41
3 -2.66 -65.41 1.17 -2.51
4(highestincome) -4.57 -136 1.61 -3.90
TRAVELRELATED Percentchanges
Trips -2.11 -0.01 -1.87 -2.11
Non-worktrips -2.88 -0.01 -2.56 -2.88
Cartrips -2.91 -1.09 -2.7 -2.92
Carmodeshare -0.6 -0.79 -0.61 -0.6
Caraveragetraveltime -1.21 1.44 -0.83 -1.21
Averagetraveltime -1.12 1 -0.81 -1.12
Congestion(flow/capacity) -10.39 -8.76 -10.16 -10.4
Gasoline -3.04 -2.03 -2.91 -3.04
Carvehiclemilestraveled(VMT) -3.74 -2.53 -3.6 -3.74
GROSSREGIONALPRODUCT -0.03 1.34 0.6 -0.02
The salestax cut raises consumer demandforgoods and servicessold inLA County by reducing the after-taxprices. From Table 7a (column1), tollingwithout recyclingthe revenue reduces non-work trips by 2.88%and mitigates conges-tion. Cuttingthe salestax stimulates demands andthe ensuing consumption leadsto morenon-work trips compared to congestionpricingwithoutrecycling,butthisstimulusislimitedandtrips arestilllowerby2.56%relativetothebaseline. VMTandaggregategasolineconsumptionarealsolowerthanthebaselinebutremainhigherthaninthecaseofcongestion pricingwithoutrecycling.
The increase in consumerdemand forproduct, increasesfirms’ demands forthe factors ofproductiontoo, leadingto higherwagesandhighercommercialrents(Table7b).Thewageincreasesdriveupconsumerdemandsfurther,compensating forthenegativeincomeeffectofthetoll.Thepositiveincomeeffectofthesalestaxcutoffsetsthenegativeincomeeffect ofthetoll,resultingina netincrease ingrossproduct,rentsandwages.Higherwagesincreasedisposable income,raising housingrents.The higherwages causejobsandpopulationto increaseinLAandRiversideCounties,morethanreversing thedispersiveeffectoftollingwithouttaxsubstitution.AsseeninTable7a,underthesalestaxcut,consumersintheregion, outsideconsumersimportingfromtheregionandrealestateownerssharethebenefitsofthetollmoreevenlythaninthe caseoftheincometaxcut.Buttheeffectsofthesalestaxcutonwagesandrentsaremuchsmaller.
4.3. Propertytaxcut
Table7b
Population,jobs,wageandrentchangeswhencongestiontollsreplaceothertaxes.
TollinginLACounty(withouttaxsubstitution) TollsreplacingtaxinLACounty
Incometax (lowincome)
Sales tax
Property tax
POPULATION
Imperial -18 -670 -222 -23
LosAngeles -2,465 +48,851 +5,667 -2,148
Orange +2,092 -14,532 -278 +1,933
Riverside -1,003 -17,144 -3,919 -1,068
SanBernardino +1,023 -12,791 -1,608 +965
Ventura +371 -3,714 +360 +341
JOBS
Imperial +18 -742 -62 +13
LosAngeles -2,299 +78,535 +3,256 -1,931
Orange +950 -35,088 -1,294 +765
Riverside +711 -16,549 -828 +639
SanBernardino +428 -17,813 -966 +368
Ventura +192 -8,343 -106 +146
WAGES(Averageoverskill/incometypes) Percent changes
Imperial -0.01 +5.65 +0.63 -0.01
LosAngeles +0.03 -0.81 +0.69 +0.05
Orange -0.13 +4.89 +0.38 -0.12
Riverside +0.12 +6.15 +0.74 +0.12
SanBernardino -0.13 +5.68 +0.48 -0.12
Ventura -0.11 +5.22 +0.38 -0.10
RENTS(Averagedoverbuildingtypes)
Res. Com. Res. Com. Res. Com. Res. Com.
Imperial +0.46 -0.01 +1.75 +1.52 +0.58 +0.12 +0.44 -0.02
LosAngeles -0.14 -0.02 +3.60 +0.79 +0.34 +0.60 -0.14 -0.01
Orange -0.01 -0.04 +1.69 +1.14 +0.21 +0.17 -0.01 -0.04
Riverside +0.41 -0.01 +1.42 +1.42 +0.49 +0.21 +0.41 -0.02
SanBernardino +0.03 +0.04 +1.75 +1.62 +0.22 +0.30 +0.03 +0.04
Ventura +0.06 -0.05 +1.74 +1.24 +0.33 +0.18 +0.06 -0.05
densityincreases.Theneteffectonpopulation densitythen dependsontheelasticityofsubstitution betweencapitaland landinnewconstructionandtherentelasticityofthedemandforhousingsize.Ourmodeltreatsbothstructuraldensity and the demand for apartment size, butbecause we treat buildings as durable the stockof buildings changes only by newconstruction. Thepropertytaxcut thatfallsonadurablebuildingisalmost entirelycapitalizedintothevalue ofthe buildingand its underlying land. The substitution ofcapital for land applies to new construction only but thiseffect is limitedbecauselittlenewconstructionisgeneratedbycongestionpricing.
AsshowninTables7aand7b,marketadjustmentsfollowingthetoll-for-property-taxsubstitutionremainnearly negli-gible.Forrealestateinvestors,thetaxcut reducescapitalcosts andincreasestheexpectedinvestmentreturns,butnearly allof thisiscapitalizedasgainsin realestatevalues.Thisconfirmstheallocative near-neutrality oftheproperty taxcut relativetotheincomeandsalestaxcuts.Furtherconfirmationofthisnear-neutralityisthatthejobandpopulationchanges inTable7b,andthetransportation-relatedchangesinTable7athatoccurunderthepropertytaxcut,arealmostthesame asthecorrespondingchangesundercongestionpricingwithoutrecyclingofthetollrevenue.
4.4.Welfare-rankingofthetaxsubstitutionsandthedoubledividendhypothesis
Ourresultsshedempirical lighton thetheoretical “doubledividend hypothesis” ofPigouviantaxation (Bovenberg and deMooij,1994;BovenbergandGoulder, 1996).Thehypothesis statesthatbytherevenue-neutralsubstitutionofPigouvian tax revenuefor therevenue froma distortionarytax, total welfareincreases.We found that the Pigouvianrevenue from congestionpricinginLA Countyreplaces 3.54%oftheaggregate revenuefromtheother taxes,andthesubstitution being revenue-neutral,theoverall welfareimproves.Moreimportantly,themagnitudeoftheseconddividend variesenormously accordingtowhichtax isreplaced.InthecaseoftheincometaxcutforthepoorestLAworkerstheseconddividend is$ 409perconsumer,inthecaseofthesalestaxitis$7andinthecaseofthepropertytaxitiszero.
caseofcongestiontollingwithoutrecycling.Landlordsenjoyabout104%oftheaggregatewelfarechangewhileconsumers sufferalossequaltoalmost4%andimportersenjoyatrivialbenefit.
4.5. Distributionofconsumerbenefitsbyincome
Undertheincometaxcut theCV of+$1,121perworkerinthelowestincomegroupisthegreatestbecauseitistheir taxthat hasbeencut,butallothergroupsbenefitaswell becauseofthewageincreases.Non-workers donot earnwages butbearthehigheroutputpricesand