道路拥堵基本法则TheFundamentalLawofRoadCongestion

1、T THEHE F FUNDAMENTALUNDAMENTAL L LAWAW OFOF R ROADOAD C CONGESTIONONGESTION: E: EVIDENCEVIDENCE FROMFROM US C US CITIESITIESBy Gilles Duranton and Matthew A. TurnerBy Gilles Duranton and Matthew A. TurnerHOW TO RELIEVE ROAD CONGESTION?Build more roads?Public transportation?Congestion pricing?OUTLIN

2、EModel SpecificationModel SpecificationReview of the literatureReview of the literatureI. Roads and Traffic: A Simple FrameworkI. Roads and Traffic: A Simple FrameworkII. Data and EstimationII. Data and EstimationIII. Implications of the Fundamental Law of Road III. Implications of the Fundamental L

3、aw of Road CongestionCongestionIV. Where Does All the VKT Come From?IV. Where Does All the VKT Come From?V. ConclusionV. Conclusion现实的启示以北京为例Future ResearchMODEL SPECIFICATIONlane kilometers of roadsvehicle-kilometers traveled (VKT)BackgroundBackground：“The Fundamental Law of HighwayHighway Congesti

4、on ”(Anthony Downs ,1962, 1992): VKT increases one for one with interstate highways.Extension:Extension:“The Fundamental Law of RoadRoad Congestion”Increased provision of interstate highways and major Increased provision of interstate highways and major urban roads is unlikely to relieve congestion

5、of these urban roads is unlikely to relieve congestion of these roads.roads.THE REASONS OF THIS RESEARCH(1) In 2001 an average American household spent 161 person-minutes per day in a passenger vehicle. The costs of congestion are large.(2) There is currently little empirical basis for accepting or

6、rejecting the claims by the ARTBA that “adding highway capacity is key to helping to reduce traffic congestion”.(3) With the increasing certainty of global warming comes the need to manage carbon emissions. The road transportation sector accounts for about a third of US carbon emissions from energy

7、use. REVIEW OF THE LITERATURERoy E. Jorgensen (1947); Phil B. Goodwin (1996) ;Robert Cervero (2002): new traffic for particular facilities after their opening or after a capacity expansion. These studies generally find a positive elasticitypositive elasticity of VKT to the supply of roads, although

8、their estimates of this elasticity vary widely.Few studies take an approach similar to ours and assess the effect of road provision on traffic over entire areas.IMPROVEMENTS IN OUR STUDY(1) We use more and more comprehensive data.Average annual daily traffic (AADT) and a description of the road netw

9、ork from HPMS for 1983, 1993, and 2003. We add a description of individual and household travel behavior taken from the 1995 NPTS and 2001 NHTS.These data are a powerful tool with which to investigate the way that VKT responds to changes in the stock of roads and transit in US metropolitan areas.Ext

10、ant research: one specific state (usually California) or a small subgroup of adjacent states (usually on the East Coast) taking counties or smaller administrative units as the unit of observation.(2) Existing literature either does not recognize that roads and traffic may be simultaneously determine

11、d, or fails to solve this identification problem.To identify the causal effect of roads on traffic, we examine both time series and cross-sectional variation in our data and exploit three instrumental variables to predict the incidence of roads in metropolitan statistical areas (MSAs).Our results st

12、rongly support the hypothesis that roads cause traffic.(3) We extend the conclusions of the existing literature in three ways: a. We find no evidence that the provision of public transportation affects VKT; b. Metropolitan areas with less traffic experience a larger increase in travel; c. we describ

13、e the foundations underlying the fundamental law of highway congestion: People drive more when the stock of roads in People drive more when the stock of roads in their city increases; commercial driving and their city increases; commercial driving and trucking increase with a citys stock of roads; a

14、nd trucking increase with a citys stock of roads; and people migrate to cities that are relatively well people migrate to cities that are relatively well provided with roads. Surprisingly, our data provided with roads. Surprisingly, our data also suggest that a new lane kilometer of also suggest tha

15、t a new lane kilometer of roadway diverts little traffic from other roads.roadway diverts little traffic from other roads.I. RI. ROADSOADS ANDAND T TRAFFICRAFFIC: A S: A SIMPLEIMPLE F FRAMEWORKRAMEWORKEquilibrium VKT, Q * (R) is characterized by:C C(R, Q): (R, Q): the total variable cost of VKT;R: R

16、: lane kilometersconstant;Q: Q: VKT;P (Q): P (Q): the inverse demand for VKT（1）（2）（3）（4）X: a vector of observed city characteristics; ?: unobserved contributors to drivingWe are interested in the coefficient of R, the road elasticity of VKT, ln Q/ ln R.Cross-Sectional:Time Series:OUTCOMES OF EXTANT

17、LITERATUREThe extant literature has estimated variants of equations (2), (3), and (4) on a small samples of counties or metropolitan areas.All find a positive association between VKT and lane kilometers of roadway, with estimated elasticities generally ranging between 0.3 and 0.7.(Frank S. Koppelman

18、 1972; Mark Hansen et al. 1993; Hansen and Yuanlin Huang ,1997; Noland, 2001)（5） is predicted lane kilometers of roadway as estimated in the first stage.cov (Z, R | X) 0 and cov (Z, ? | X) = 0.The possible simultaneous determination of VKT and lane kilometers is recognized by several authors.Should

19、all three methods arrive at the same estimate of , then all are correct, or all are incorrect, and an improbable relationship exists between the various errors and instrumental variables.II. DII. DATAATA ANDAND E ESTIMATIONSTIMATIONTo measure each MSAs stock of interstate highways and traffic, we us

20、e the US HPMS “universe” and “sample” data for 1983, 1993, and 2003.We use a county identifier to match every segment of interstate highway to an MSA. We then calculate lane kilometers, VKT, and AADT per lane km for interstate highways within each MSA.From the sample data we calculate road length, l

21、ocation, AADT, and share of truck traffic for all major roads in the urbanized area.SIX FUNCTIONAL CLASSES DESCRIBED IN DOT (1989)1. interstate highway2. collector3. minor arterial4. principal arterial5. other highwaymajor urban roadA. CROSS-SECTIONAL ESTIMATES OF THE ROADWAY ELASTICITY OF VKTCatego

22、ries of roads and travel:1. All MSA interstates (IH): 国道2. Urbanized MSA interstates (IHU):三环、四环3. Major urban roads(MRU):长安街4. Nonurban MSA interstates (IHNU):偏远地区洲际公路，七环？平谷；MSAs: metropolitan statistical areas-统计的大都会区域，整个北京；Urbanized areas：海淀，朝阳，不包括平谷、密云 。B. FIXED EFFECTS AND TIME-SERIES ESTIMATES

23、 OF THE ROADWAY ELASTICITY OF VKT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. IV ESTIMATES OF THE ROADWAY ELASTICITY OF VKTCov (X,U)=0Three instruments: planned interstate highway kilometers from the 1947 highway plan; 1898 railroad route kilometers; the incidence of major expeditions of exploration between 1835

24、and 1850.We generally report the results of conventional TSLS estimations, but in the few cases where our instruments are weak, we also report the corresponding LIML estimates.A qualifier:Our instruments are good predictors of MSA-level stocks of interstate highways and urban interstate highways. Th

25、ey are not good predictors of MSA level stocks of major roads or of nonurban interstate highways. For this reason, we conduct IV estimations only for interstate highways and urban interstate highways.THE CONDITIONAL EXOGENEITY OF OUR THREE INSTRUMENTSPopulation growth around 1947 is uncorrelated wit

26、h planned highway kilometers.Thus, the 1947 plan was drawn to fulfill its mandate and connect major population centers of the mid-1940s, not to anticipate future population or traffic demand.Cities that receive more roads in the 1947 plan tend to be larger than cities that receive fewer. large citie

27、s have higher levels of VKT, 1947 planned interstate highway kilometers predicts VKT by directly predicting population and indirectly by predicting 1980 road kilometers.Thus the exogeneity of this instrument hinges on having an appropriate set of controls, population in particular.This network was b

28、uilt, for the most part, during and immediately after the civil war, and during the industrial revolution.The rail network was developed by private companies with the intention to make a profit from railroad operations in the not too distant future.Instrument validity: Instrument validity: condition

29、al on control variables, rail routes be correlated with the dependent variable only through contemporaneous interstate highways.With this said, after controlling for historical populations and physical geography, it is difficult to imagine how a rail network built for profit could anticipate the dem

30、and for vehicle travel in cities 100 years later, save through its effect on roads.Some of these expeditions were explicitly charged with finding an easy way from one place to another, and it is hard to imagine that this objective was not also important to the others. While we expect that these earl

31、y explorers were drawn to attractive places, after controlling for historical populations and physical geography it is difficult to imagine how these explorers could select routes that anticipate the demand for vehicle travel in cities 150 years later, save through their effect on roads.198319932003

32、弹性减小道路被分配到旅行需求较少的MSAs；就业受到冲击的城市，道路较多；越早建的公路越有效，吸收交通越多；美国洲际公路的生产力效应随时间下降。III. IMPLICATIONS OF THE FUNDAMENTAL LAW OF ROAD CONGESTIONTwo logical implications:Two logical implications:A. Traffic and TransitB. convergence of AADT Levels（平均年度月交通水平的收敛性）A. TRAFFIC AND TRANSITNew road capacity be met with a

33、 proportional increase in drivingRemove a subset of a citys drivers from a citys roads, then others would take their placePublic transit serves to free up road capacity by taking drivers off the roads and putting them in buses or trainsThe fundamental law implies that the provision of public transit

34、 should not affect the overall level of VKT in a cityTo measure an MSAan MSAs stock of public transits stock of public transitDATAMSA-level dataSection 15 annual reportsDaily average peak service of large buses in 1984, 1994, and 2004Not available:light rail, independently of busesIV ESTIMATIONEndog

35、enous:Endogenous:public transit & MSA interstate VKTInstrument:Instrument:the MSA share of democratic vote in the 1972 presidential election,a good predictor of the 1984 MSA provision of buses,log 1984 buses and 1972 democratic vote is 0.34Description:Description:“stickiness”,the provision of buses

36、at this time did not seek to accommodate traffic congestion during the 19832003 periodBACKGROUND:1972 US PRESIDENTIAL ELECTIONRichard Nixon vs George McGovernFocus:the Vietnam War and McGoverns very progressive social agendaTime:after the 1970 Urban Mass Transportation Act,predates the first oil sho

37、ck and the 1974 National Mass Transportation Act that followedPersistent:1960,$5 billion;1970, $15 billion;1974,$44 billionResult:during the 1970s public transit expanded and evolved from a private fare-based industry to a quasi-public sector activity sustained by significant subsidies1small,趋近于0fai

38、l to support the hypothesis that increased fail to support the hypothesis that increased provision of public transit affects VKTprovision of public transit affects VKTB. CONVERGENCE OF AADT LEVELSWhy examine the evolution of AADT?each MSA have an intrinsic natural level of traffic conditional on lan

39、e kilometers of roadway,a deviation from this natural level ought to be followed by a return to ite.g.:1980-2000,the cross-MSA standard deviation of all interstate AADT decreases from 1.40 to 1.28 Pool first differences in interstate AADT for 1990 and 2000 and regress them on initial interstate AADT

40、 levels1improvements to highways cause large increases in the improvements to highways cause large increases in the use of these routes by long-haul truckers, while use of these routes by long-haul truckers, while improvements to the local road network cause smaller improvements to the local road ne

41、twork cause smaller increases in local commercial trafficincreases in local commercial trafficB. INDIVIDUAL DRIVING BEHAVIOR AND HIGHWAYS1995 and 2001 NPTS(Nationwide Personal Transportation Survey):(1)household survey: age, race, education, and income,etc.(2)vehicle survey:total VKT for the househo

42、ld(3)person survey:commuting behavior(4)trip survey:all household daily VKTDifference between NPTS and HPMS: NPTS HPMSa per household measure of VKT on all roadsaggregate VKT on interstates and major urban roads within MSAsall household driving on any roads, but ignores commercial or through traffic

43、 and changes in populationall traffic on a subset of roads TWO SERIES OF ESTIMATIONSFirst stage:First stage:Adjusted city level cross-section estimating equation(2)Observation unit:a person or household in a particular city and year denotes VKT on all roads for household (or individual) j, and i ind

44、exes MSAsControl variables: MSA-level characteristics and individual demographic characteristics Second stage: Second stage: individual- or household-level analog instrumental variables estimating equation未控制人口未控制人口10%10%州际道路的增加导致所有道路上个人驾驶里程州际道路的增加导致所有道路上个人驾驶里程1%1%的的增加。但不知是哪种道路类型带来的增加。增加。但不知是哪种道路类型带

45、来的增加。C. POPULATION GROWTHCost of transportation,attractive cityHigh mobility of the US population,a citys road network should be met with changes to a citys populationMichaels (2008)and Amitabh Chandra and Eric Thompson (2000):interstate highway lead to increases in retail earnings,firm earnings,the

46、 level of local economic activity population increasesDuranton and Turner (2008) consider US MSAs between 1980 and 2000 and investigate the way that population growth responds to changes in the road network.A10 percent increase in the extent of the road network causes a 1.3 percent increase in MSA p

47、opulation over 10 years, and a 2 percent increase over 20 yearsD. DIVERSION FROM OTHER ROADSWhether changes to one class of roads affects VKT on the others: urbanized area interstates, nonurbanized area interstates, and major urbanized area roads:E. AN ACCOUNTING EXERCISEConsider whether above four

48、sources are sufficient to explain the fundamental law and assess their relative importance：1.changes in trucking and commercial driving1.changes in trucking and commercial drivingTable 6,10%-10.3%;Table 1,trucks on VKT,13%; Table 9,2.3% interstate highways,18% total VKTtrucks account for between 19

49、and 29 19 and 29 percent of the total increase in interstate VKT2.changes in population(migration)2.changes in population(migration)Duranton and Turner (2008),2.1%,Table 6,0.30.6% interstate VKT,6% total increasePanel B of Table 5,controls for the endogeneity of population in first-difference estima

50、tes,higher estimates,1.02Migration accounts for between 5 and 21 5 and 21 percent of the total increase in interstate VKTE. AN ACCOUNTING EXERCISE3.substitution across roads3.substitution across roadsneed only be concerned with diversion of traffic from major urbanized-area roads. panel C of Table 1

51、1,diverts 0.52 percent of traffic from major urban roadsTable 1,1 percent increase in interstate VKT, or about 10 percent of the total effectTable 11,no substitution from major urban roads toward interstates,cannot rule out the absence of a substitution effectE. AN ACCOUNTING EXERCISE4.changes to ho

52、usehold behavior4.changes to household behaviorTable 10,a 10 percent increase in interstate lane kilometers causes a 1.1 percent increase in household annual VKTUnfortunately, our data do not allow us to apportion household driving to different road networks：(1)A first possibility is to assume that

53、this 1.1 percent increase in driving is proportional to current driving across all road networks,an unrealistic lower bound(2)Alternately, suppose that the 1.1 percent increase in household driving takes place only on interstates,the increase in interstate VKT would account for 4.1 percent of the to

54、tal change in VKT, or 39 percent of the effect increases in household driving account for between 9 and 39 between 9 and 39 percentpercent of the total increase in interstate VKTSum upSum up, of four possible sources for the new traffic following an increase in lane kilometers of interstates, change

55、s to individual individual behavior behavior and changes in commercial drivingcommercial driving are the most important. Migration and traffic diversion are significantly less importantV. CONCLUSIONOur estimates of the elasticity of MSA interstate highway VKT with respect to lane kilometers are 0.86

56、 in OLS, 1.00 in first difference, and 1.03 1.03 with IVConfirm two implications two implications of the fundamental law of road congestion: no evidence that public transit affects VKT, and there is convergence of traffic levels;roads are assigned to MSAs with little or no regard for the prevailing

57、level of traffic.Sources of new trafficSources of new traffic:changes to individual driving behavior and increases in trucking most important;Migration somewhat less important;diversion of traffic from other road networks does not appear to play a large roleMain candidate tool:Main candidate tool:Ro

58、ad capacity expansions and extensions to public transit Congestion pricing Congestion pricing Why?Why?交通拥挤收费的经济学分析交通拥挤收费的经济学分析【北大林肯研究简报2011】 宋顺锋教授 长期以来，中国主要依靠供给政策来缓解交通拥挤问题 城市现有布局、建设用地、交通出行的“三重集中”，“交通拥挤扩建道路私人交通增多公交需求减少交通更加拥挤”恶性循环。以“交通供给”为主转变为以“需求管理”为主，通过价格杠杆克服市场失灵。从经济学角度来看，出行是一种派生需求。出行者只关心自身的成本，不在乎外部

59、成本。没有交通拥挤收费的情况下，市场是失灵的，出行者的均衡数量大于社会最优状态，使得城市路网过于拥挤、资源浪费和环境污染。 从整个路网和社会的角度来看，达到最优的条件是社会边际收益等于社会边际成本，而非边际收益与个人边际成本交叉的G点。“谁造成拥挤、谁付费”：只有拥堵收费机制最能直接影响出行的时间、线路和方式，而且与出行的距离和拥挤程度直接关联，更能分流高峰时段的需求，缓解交通拥堵。从公平角度讲，中国目前乘私车出行绝大多数还属于高收入阶层，一般市民上下班基本上靠其他方式(如自行车和公交车)。因此，拥堵收费在很大程度上可以帮助社会的公平分配，让富裕的人群多付出一些金钱，用来改善公交服务水平，帮助

60、大多数老百姓的出行。矛盾点：高收入阶层不在乎那点收费，与本文另一结论公交水平提升不起作用矛盾，拥堵问题仍然存在？9月2日 “北京拥堵日”早高峰收费前还得做点啥？以北京交通状况为例首都-“首堵”现有政策：限行、2011年初摇号限购、大幅增加停车收费结果：限车令丝毫没有影响市民的买车热情，道路状况稍有好转，不断新增的私家车随即又拉长了缓慢蠕动的车行队列；交通拥堵费引来一片反对之声。征收拥堵费难点在哪里？征收拥堵费难点在哪里？公共交通体系人均利用率低 截至2012年7月底北京机动车达到511.6万辆大关，私家车的增多大大削弱了公交系统的利用率。根据北京交通发展研究中心的统计数字显示，2010年北京市