道路拥堵基本法则The Fundamental Law of Road Congestion

THEFUNDAMENTALLAWOFROADCONGESTION:EVIDENCEFROMUSCITIESByGillesDurantonandMatthewA.TurnerHOWTORELIEVEROADCONGESTION?Buildmoreroads?Publictransportation?Congestionpricing?OUTLINEModelSpecificationReviewoftheliteratureI.RoadsandTraffic:ASimpleFrameworkII.DataandEstimationIII.ImplicationsoftheFundamentalLawofRoadCongestionIV.WhereDoesAlltheVKTComeFrom?V.Conclusion现实的启示——以北京为例FutureResearchMODELSPECIFICATIONlanekilometersofroadsvehicle-kilometerstraveled(VKT)Background：“TheFundamentalLawofHighwayCongestion”(AnthonyDowns,1962,1992):VKTincreasesoneforonewithinterstatehighways.Extension:“TheFundamentalLawofRoadCongestion”——Increasedprovisionofinterstatehighwaysandmajorurbanroadsisunlikelytorelievecongestionoftheseroads.THEREASONSOFTHISRESEARCH(1)In2001anaverageAmericanhouseholdspent161person-minutesperdayinapassengervehicle.Thecostsofcongestionarelarge.(2)ThereiscurrentlylittleempiricalbasisforacceptingorrejectingtheclaimsbytheARTBAthat“addinghighwaycapacityiskeytohelpingtoreducetrafficcongestion”.(3)Withtheincreasingcertaintyofglobalwarmingcomestheneedtomanagecarbonemissions.TheroadtransportationsectoraccountsforaboutathirdofUScarbonemissionsfromenergyuse.REVIEWOFTHELITERATURERoyE.Jorgensen(1947);PhilB.Goodwin(1996);RobertCervero(2002):newtrafficforparticularfacilitiesaftertheiropeningorafteracapacityexpansion.ThesestudiesgenerallyfindapositiveelasticityofVKTtothesupplyofroads,althoughtheirestimatesofthiselasticityvarywidely.Fewstudiestakeanapproachsimilartooursandassesstheeffectofroadprovisionontrafficoverentireareas.IMPROVEMENTSINOURSTUDY(1)Weusemoreandmorecomprehensivedata.Averageannualdailytraffic(AADT)andadescriptionoftheroadnetworkfromHPMSfor1983,1993,and2003.Weaddadescriptionofindividualandhouseholdtravelbehaviortakenfromthe1995NPTSand2001NHTS.ThesedataareapowerfultoolwithwhichtoinvestigatethewaythatVKTrespondstochangesinthestockofroadsandtransitinUSmetropolitanareas.Extantresearch:onespecificstate(usuallyCalifornia)orasmallsubgroupofadjacentstates(usuallyontheEastCoast)takingcountiesorsmalleradministrativeunitsastheunitofobservation.(2)Existingliteratureeitherdoesnotrecognizethatroadsandtrafficmaybesimultaneouslydetermined,orfailstosolvethisidentificationproblem.Toidentifythecausaleffectofroadsontraffic,weexaminebothtimeseriesandcross-sectionalvariationinourdataandexploitthreeinstrumentalvariablestopredicttheincidenceofroadsinmetropolitanstatisticalareas(MSAs).Ourresultsstronglysupportthehypothesisthatroadscausetraffic.(3)Weextendtheconclusionsoftheexistingliteratureinthreeways:a.WefindnoevidencethattheprovisionofpublictransportationaffectsVKT;b.Metropolitanareaswithlesstrafficexperiencealargerincreaseintravel;c.wedescribethefoundationsunderlyingthefundamentallawofhighwaycongestion:Peopledrivemorewhenthestockofroadsintheircityincreases;commercialdrivingandtruckingincreasewithacity’sstockofroads;andpeoplemigratetocitiesthatarerelativelywellprovidedwithroads.Surprisingly,ourdataalsosuggestthatanewlanekilometerofroadwaydivertslittletrafficfromotherroads.I.ROADSANDTRAFFIC:ASIMPLEFRAMEWORKEquilibriumVKT,Q*(R)ischaracterizedby:C (R,Q):thetotalvariablecostofVKT;R:lanekilometers—constant;Q:VKT;P(Q):theinversedemandforVKT（1）（2）（3）（4）X:avectorofobservedcitycharacteristics;ϵ:unobservedcontributorstodrivingWeareinterestedinthecoefficientofR,theroadelasticityofVKT,≡∂lnQ/∂lnR.Cross-Sectional:TimeSeries:OUTCOMESOFEXTANTLITERATURETheextantliteraturehasestimatedvariantsofequations(2),(3),and(4)onasmallsamplesofcountiesormetropolitanareas.AllfindapositiveassociationbetweenVKTandlanekilometersofroadway,withestimatedelasticitiesgenerallyrangingbetween0.3and0.7.(FrankS.Koppelman1972;MarkHansenetal.1993;HansenandYuanlinHuang,1997;Noland,2001)（5）ispredictedlanekilometersofroadwayasestimatedinthefirststage.cov(Z,R|X)≠0andcov(Z,ϵ|X)=0.ThepossiblesimultaneousdeterminationofVKTandlanekilometersisrecognizedbyseveralauthors.Shouldallthreemethodsarriveatthesameestimateof,thenallarecorrect,orallareincorrect,andanimprobablerelationshipexistsbetweenthevariouserrorsandinstrumentalvariables.II.DATAANDESTIMATIONTomeasureeachMSA’sstockofinterstatehighwaysandtraffic,weusetheUSHPMS“universe”and“sample”datafor1983,1993,and2003.WeuseacountyidentifiertomatcheverysegmentofinterstatehighwaytoanMSA.Wethencalculatelanekilometers,VKT,andAADTperlanekmforinterstatehighwayswithineachMSA.Fromthesampledatawecalculateroadlength,location,AADT,andshareoftrucktrafficforallmajorroadsintheurbanizedarea.SIXFUNCTIONALCLASSESDESCRIBEDINDOT(1989)1.interstatehighway2.collector3.minorarterial4.principalarterial5.otherhighwaymajorurbanroadA.CROSS-SECTIONALESTIMATESOFTHEROADWAYELASTICITYOFVKTCategoriesofroadsandtravel:1.AllMSAinterstates(IH):国道2.UrbanizedMSAinterstates(IHU):三环、四环3.Majorurbanroads(MRU):长安街4.NonurbanMSAinterstates(IHNU):偏远地区洲际公路，七环？平谷；MSAs:metropolitanstatisticalareas---统计的大都会区域，整个北京；Urbanizedareas：海淀，朝阳，不包括平谷、密云。B.FIXEDEFFECTSANDTIME-SERIESESTIMATESOFTHEROADWAYELASTICITYOFVKT人口会迁移至增长缓慢的MSA系数不随估计方法的改变而改变C.IVESTIMATESOFT