The Stockholm congestion charges: an overview · The Stockholm congestion charges: an overview Jonas Eliasson, KTH Royal Institute of Technology CTS Working Paper 2014:7 Abstract

TheStockholmcongestioncharges:anoverview

JonasEliasson,KTHRoyalInstituteofTechnology

CTS Working Paper 2014:7

AbstractCongestion chargeswere introduced in Stockholm in2006as a seven‐month trial,followedbyareferendumwhereamajorityvotedinfavourofthecharges.Thisledto the reintroduction of congestion charges in August 2007, and they have beenoperational since then. The system has attractedworldwide attentionworldwide,bothbecauseitachievedsubstantialcongestionreductions,andbecausethesystemovercame fierce initial hostility, surviving a heated and complicated political andlegal process, and eventually gaining support bymore than 2/3 of the populationand all political parties. This report summarises the story of the Stockholmcongestioncharges,pointingoutexperiencesandlessonslearnt.Note: Parts of the material in this report have been published before in otherpublications by the author and various coauthors.The intention of this report is tomakethematerialaccessibleinonesingle,coherentsource.Keywords:Congestioncharges,Stockholm,sustainabletransport.JELCodes:H23,H54,R41,R48.

Centre for Transport Studies SE-100 44 Stockholm Sweden www.cts.kth.se

TheStockholmcongestioncharges

2

1 INTRODUCTION

Congestion charges1 were introduced in Stockholm in 2006 as a seven‐month trial,followedbyareferendumwhereamajorityvotedinfavourofthecharges.Thisledtothe reintroduction of congestion charges in August 2007, and they have beenoperationalsincethen.Thechargingsystemconsistsofacordonaroundtheinnercity,withatime‐differentiatedtollbeingchargedineachdirection.Trafficacrossthecordonwas reduced by around 20%, leading to substantial congestion reductions in andaroundthecity.Congestionpricinghasbeenadvocatedbytransporteconomistsandtrafficplannersforalongtimeasanefficientmeanstoreduceroadcongestion.Despitegrowingproblemswithurbancongestionandurbanairquality,anddespiteaconsensusthatinvestmentsin roads or public transit will not be sufficient to tackle these problems, cities havebeenreluctanttointroducecongestionpricing.Inrecentyears,however,itseemsthatthis is changing. London (2003), Stockholm (2006), Durham (2002), Milano (2008),Rome (2001), Gothenburg (2013) and Valletta (2007) have all introduced differentforms of charging or permit systems to combat congestion and/or environmentalproblems,andmanyothercitiesareconsideringit.NewYork,Manchester,CopenhagenandEdinburghhaveallrecentlytriedtointroducecongestioncharges,andeveniftheseattemptshavebeenunsuccessful,itisasignthatcongestionchargesarebeingseriouslyconsideredtoagreaterextentthanadecadeago.Thesoonubiquitous“valuepricing”roadsintheUSarealsoexamplesofhowcongestionproblemsarenowbeingtackledthroughpricingmeasures.The congestion charges in Stockholm have attracted enormous attentionworldwide.Obviously, the opportunity to gauge the effects of congestion charges on traffic,congestion levelsand travelbehaviourhasattractedgreat interest.Butperhapsevenmore interesting is that the congestion charges overcame fierce initial hostility,survivedaheatedandcomplicatedpoliticalandlegalprocess,includingareferenduminitiallyforcedthroughbyopponentstothecharges,andhaseventuallygainedsupportby more than 2/3 of the population. The Stockholm charges went from “the mostexpensive way ever devised to commit political suicide” (to quote the then‐secretfeelingsexpressedby theHeadof theCongestionChargingOffice2) to something thatthe initially hostile media eventually declared to be a “success story” (e.g. DagensNyheter,June22,2006).This report3 tries to summarise the lessons that can be learnt from the Stockholmexperiences.

1 Legally, the congestion charge is a tax, according to Swedish law, and the official Swedish term is hence “congestion tax”. We have chosen to use the international standard term “congestion charge”. 2 Quote Gunnar Söderholm, social-democratic head of the Congestion Charging Office during the trial, when (after the trial) describing the local Social Democrats’ feelings when the national Social Democratic government more or less forced the congestion charges onto the local Stockholm party district. 3 The report draws heavily on previously published papers by the author and coauthors. Some parts are taken verbatim, in particular from (Börjesson, Eliasson, Hugosson, & Brundell-Freij, 2012; Eliasson, 2008, 2010, 2014). Further references are given in the report. The intention is not to present new material, but merely to collect previously published works in a single easily accessible place.


3

1.1 Background: some basic insights about congestion pricing

It is well established, both theoretically and empirically, that infrastructureinvestmentsarenotsufficienttoeliminateroadcongestioninthecoresoflargecities.There are several reasons for this: two of the most important are the eventuallyinevitablescarcityofurbanlandandpublicresources.Congestion charging will of course not solve everything. Introducing congestionchargeswillusuallyreducetheneedfortransportinvestments,butgenerallyspeakingnoteliminateit.Normally,agrowingurbanregionwillneedbothcongestionchargingandtransport investments,perhapsbothroadsandpublictransport.Obviously,citiesaredifferentas towhat investmentsare themostcost‐efficientandthemostneeded.Generallyspeakingagain,asustainableurbantransportsystemmustincorporatefourstrategies: attractive public transport, walkability, compact spatial planning, andrestraintsoncar traffic.All these fourwill strengtheneachother,andwithoutoneofthem, the remaining three will lose effectiveness. The purpose of this paper is topresentevidenceandlessonsaboutcongestioncharging–whichisonewaytorestraincartraffic;parkingpricingisanotherexample–buttheotherthreeneedtobeincludedinacomprehensivesustainableurbantransportstrategy.It isprobablyworthtopointoutthatit isalsowellestablishedthatcardriversareinfact sensitive to costs. It is a commonmisconception among laymen, and sometimesamongdecisionmakers,thatdrivers“haveto”drive,anddonotreacttochangesinthedrivingcosts.Thisnotionhasbeenrefutednumeroustimesinmanykindsofcontexts.Increasingthecosttodriveatcertainplacesatcertaintimeswilldecreasethenumberofdriverschoosingtodrivethereandthen.Howlargethedecreasebecomesdependsontheeaseofadaptation,amongotherthings–inotherwords,howgoodalternativesthereare.Alternativesmaybeothertimeperiods,modes,routes,destinationsetc.Itisimperativetokeepasmanyoptionsopenaspossibletoachievegoodtrafficreductioneffects–butitisuptothedriversthemselvestochoosehowtoadapt.

2 THE STOCKHOLM STORY IN BRIEF

2.1 Basic facts about Stockholm

TheCityofStockholmhasaround0.9millioninhabitants,andisthecentralpartoftheStockholm county, with a total of 2 million inhabitants. Around 2/3 of the Cityinhabitantsliveintheinnercity–thatis,withinthetollcordon‐andtherestoutside.The area4 of the toll zone is around 35 km2. The zone has around 330 000 inhabitants, of which approximately 60 000 commute to workplaces outside of the zone. The zone has close to 23 000 workplaces, employing approximately 318 000 persons, of which more than two thirds are commuting from outside the zone. Thepopulationof Stockholmhasbeen growing rapidly formanyyears, and togetherwithincreasingtriplengthsandcarownership,thishasledtosteadilyincreasingtrafficvolumes.Traffic volumesacross the cordonused to increase at the samepaceas thetraffic in the county as awhole from the early 1970’s (when regularmeasurementsstarted) up until the early 1990’s, when traffic across the cordon stopped growing.Trafficintherestofthecounty,however,continuedgrowingatthesamepace,asdidthe number of transit trips across the cordon. The most likely explanation of thissuddenendtotrafficgrowthissimplyalackofroadcapacity.Trafficacrossthecordon4 As a comparison, the congestion charge scheme of London introduced in 2003 encompassed a 21 km2 zone, this was almost doubled by the western extension in 2007.


4

then remained surprisingly stable for the next 15 years or so, despite significantchangesinemploymentlevels,fuelpricesetc.,withtheonlyappreciableeffectbeingatrafficdecreaseofaround5%whentheSouthernBypassopenedin2004.Because of its topology, with lots of water and well‐preserved green wedges, roadcongestion levels inStockholmarehighcompared to the city’smoderate size.Beforethe introduction of the congestion charges, themain roads arterials leading to, fromandwithinthecitycentrehadcongestionindicestypicallyaveragingaround200%,i.e.threetimesthefree‐flowtraveltime.Partlybecauseofthis,andpartlybecauseofgoodpublictransportsupply,thetransitshareishigh:60‐65%ofallmotorizedpersontripstoandfromthecitycentrearemadebytransit.Duringrushhours,theshareincreasesto80%.ThepublictransportsysteminthecountyofStockholmconsistsofasubwaynetworkwith100stationsandoveramilliontripsperday,acommuterrailnetworkwith51stationsandnearlyaquarterofamilliontripsperday,fivelightraillineswith98stationswithabitmorethan100000tripsperday,andanextensivebusnetworkwithnearly amillion trips per day. Public transport fares are subsidized at a rate ofaround50%ofactualcosts.

2.2 The story of the charges – an overview

Just like inmany other cities, transport planners and economists had suggested thatStockholmshould introducecongestionpricingfora longtime,withoutgettingeitherpublicorpolitical support. In theearly1990’s, road tollswereproposedas away topartiallyfinancealargeinfrastructurepackageforStockholm.Thisignitedtheinterestfromenvironmentalists,whoappreciatedthetrafficmanagementpotentialofthetolls,eveniftheydidn’tapproveofthattherevenueswereplannedtobepartiallyusedfornewmotorways.Theinfrastructurepackageagreementbrokedowninthelate1990’s,and the tolls were never introduced – but the ball had been set rolling. Severalstakeholderscarriedoutanalysesof congestionchargingschemes,andperhapsmoreimportant, the issue had entered the agenda of the environmental movement, inparticulartheGreenparty.In2002,thesocial‐democraticnationalgovernmentsetupacommissiontonegotiateanewinfrastructureagreementforStockholm.Theideawasfloatedtouseroadpricingas a funding source. When the Conservative party accused the social‐democrats ofhaving secret plans to introduce ”road tolls” after the election 2002, the social‐democraticmayorinStockholmpromisedveryclearlyandpubliclythattherewouldbeno road tolls in Stockholm during the next election cycle (although she wanted toprepare a suggestion in time for the next election). The social‐democratswent on towin both the national and the Stockholm election, provided that they could ensuresupport from the Green party. In return for support for a social‐democratic nationalgovernment, the Green party demanded that a ”several‐year, full‐scale congestionchargingtrial”shouldbecarriedoutinStockholm.Thesocial‐democratsobliged.Thisledtoanextremelyheateddebate.Congestionpricingwasanunpopularmeasurefromtheoutset,andthebrokenelectionpromisemademattersworse.Theoppositionraged,whilesilentlycelebratingwhattheyanticipatedtobea landslidevictoryinthenextelection.Bothproponentsandopponentsofthechargesuseddramaticrhetorictodescribewhatwould happenwith orwithout congestion charges, respectively. Evenmany of those in favour of congestion chargeswere sceptical: theway the trialwasintroduced,andtheshorttimeavailableforpreparation,madethemfearthatafailedattempt at introducing congestion chargeswould block the question formany yearsahead.Themediapicturewasoverwhelminglynegative:39%ofallnewspaperarticlesonthetopicwerenegative,comparedto3%positive(therestwereneutral)(Winslott‐


5

Hiselius, Brundell‐Freij, Vagland, & Byström, 2009). Opponents to the chargessuggestedareferendumaboutthecharges,confidentthattheywouldwin.Theideawassilentlywelcomedbythesocial‐democrats,whosawitasawaytoputsomedistancebetween them and the charges:with a separate referendum, itwould be possible tovoteforthesocial‐democratsandstillvotenotothecharges.However,itwasdecidedthatthereferendumshouldnotbehelduntilafterthetrial,inconjunctionwiththenextregularelectioninSeptember2006.Thisturnedouttobeofcrucialimportance.Thetrialhadthepurposeto“testwhethertheefficiencyofthetrafficsystemcouldbeenhancedbycongestioncharges”.Thetollwasexpectedto“reducecongestion,increaseaccessibilityandimprovetheenvironment”,bothintermsofemissionsfromcartrafficandtheperceivedurbanenvironment.Thetollratewassetsoastoreachthetargetofreducing car traffic across the cordonwith 10‐15 percent, a target loosely based onpreviouslysuggestedroadpricingschemesforStockholm.The congestion charging trial started in January 2006, when a time‐differentiatedcordon toll around the inner citywas introduced. Traffic across the cordon droppedimmediately, leading to dramatic congestion reductions all over the city. After a fewweeks, the decrease in traffic volumes across the cordon during the charged periodstabilized around 22% compared to 2005 levels, resulting in congestion reductionsaround 30‐50% (Eliasson, Hultkrantz, Nerhagen, & Rosqvist, 2009; Eliasson, 2008).Public attitudes gradually became more positive, while the media picture changedcompletely:theshareofpositivenewspaperarticlesincreasedfrom3%to42%whiletheshareofnegativearticlesfell from39%to22%(Winslott‐Hiseliusetal.,2009).Inthe referendum in September, 53% of valid votes were in favour of keeping thecharges5.Representatives for all political parties in Stockholm had promised to follow theoutcome of the Stockholm referendum. The election ended up withliberal/conservativemajoritiesbothnationallyandinStockholm,andthenewmajorityinStockholmobliginglyaskedthenewnationalmajoritytoreintroducethecongestioncharges, which had been turned off before the referendum. The crux was thenegotiationabouttherevenues.Legally,thechargewasanationaltaxthatendedupinthe national government’s coffers, but the Stockholm region understandably arguedthat it was really their money. Eventually, the regional and national politiciansbrokeredahuge,ten‐yearinfrastructurepackagewortharound10billioneuros,whereonepartof thedealwas that thechargerevenueswereearmarked foranewbypassaroundStockholm.Thetrialhenceturnedouttobeamilestoneinthedevelopmentofurbanroadpricing.First,tothesurpriseofallbutafewhard‐headedroad‐tollenthusiasts,itfinallytippedthe balance of a forty‐year political consideration of road‐tolls in Stockholm byinvokingmoreorlessaland‐slidechangeoftheopinionofthegeneralpublicinfavouroftolls.Second,itwasthethirdfull‐scaledemonstrationofanurbancongestioncharge,

5 Some surrounding municipalities also arranged referenda, although they had no legal influence, since the congestion charges were entirely within the border of the city of Stockholm. (Around 2/3 of the city’s population live inside the cordon. It should be noted, by the way, that residents within the cordon pay on average much more in charges than residents outside the cordon, while they get less of the travel time benefits, since congestion problems mainly exist into the city in the morning and out from the city in the afternoon). Counting the votes of all referenda that were held, there was a majority against the charges, but the selection was heavily skewed: referenda were only held in municipalities with liberal/conservative majorities and where opinion polls showed that there was a majority against the charges.


6

after Singapore and London, and the second to be based on a time‐differentiatedscheme,afterSingapore.As time went on, all political parties accepted and, eventually, even embraced thecongestioncharges.Thereasonsforthisincludedthecongestionreduction,themeansto finance infrastructure, thepossibility toget leveraged funds from thegovernment,andthesteadily increasingpublicsupport for thecharges.Themedia interest for thecharges faded, after having been in the headlines almost daily for four years. Ratherthan discussing the existence of the charges, the political parties and otherstakeholders graduallymovedon todiscussinghow the charges couldbe redesignedandhowtherevenuesshouldbeused.Thetrafficreductionhasremainedremarkablystableovertime(Börjesson,Eliasson,Hugosson,&Brundell‐Freij,2012).Atthetimeofwriting (February 2014), the liberal/conservative majorities in the nationalgovernment,thecityofStockholmandthecountyofStockholm–theformeropponentsof the charges – have agreed to substantially increase the level of the charges andintroduceanewtollonthewesternbypass,withthedualpurposetofinanceametroextensionandreducecongestionevenfurther.Theonlyobjectionfromthe left/greenoppositionisthatitistoolittle,tooslowandtoolate.

3 THE CHARGING SYSTEM AND ITS EFFECTS

TheeffectsoftheStockholmsystemwerestudiedinanextensiveevaluationprogram.Thisevaluationprogramwasparticularlyimportantsincethefateofthechargeswouldbedecidedbyapopularreferendum.Theevaluationcoverednotonlyeffectsontrafficvolumesandtraveltimes,butalsoeffectsonemissions,perceivedurbanenvironment,trafficsafety,deliverytraffic,publictransport,taxisetc.,generatingalmost30differentsub‐reports.

3.1 Description of the system

The system consists of 18 charging points located at the main bottlenecks on thearterialsleadingintoandoutfromtheinnercity.These18pointsformacordonaroundthe inner city. The cordon shape is just a consequence of the topology of the city,though,notadesignconstraint;itisacommonmisconceptionthatcongestionchargingsystemshavetobeshapedascordons(systemdesignisdiscussedinsection5).Vehiclesareregisteredautomaticallybycameras thatphotograph thenumberplates;thereisnoopportunitytopayatthecontrolpoints.Theownerofthecaristhensentamonthlyinvoiceforthetotalchargeincurredduringamonth.Duringthetrial,themainmeansofidentificationwastransponders(“tag‐and‐beacons”orDSRC,dedicatedshort‐range communication). When the charges were reintroduced, the automatic cameraidentification,originally intendedonlyasa secondarymeansofvehicle identification,workedsowellthatitwasdecidedtoabolishthetransponders.


7

Figure1.The charged area. The dashed line is the charging cordon, the dots are charging points and the solid line is the non-charged Essinge bypass. Right: Charges in different time intervals (weekdays only).

The cost for passing a control point in any direction is 1‐2€ (using 10 SEK = 1€)dependingonthetimeofday(seeFigure1),withamaximumamountpervehicleandday of 6€. The cost is the same in both directions, and each passage is charged. Nocongestion charge is levied during nights, weekends, holidays on in July. Variousexemptions (e.g.buses, foreigncarsand for trafficbetween the islandofLidingöandtherestofthecounty)meanthatabout15%ofthepassagesarefreeofcharge.Originally, therewasanexemption foralternative‐fuelcars, intendedtostimulatethemarketintroductiononsuchcars.Thisprovedtobeaneffectivemeasure:theshareofalternative‐fuel cars increased from 3% in 2006 to 15% in 2009. The sales ofalternative‐fuel cars was also stimulated in other ways, but several studies haveconcluded that the congestion charge exemption played an important role. Theexemptionwasabolishedforvehiclessoldfrom2009andlater,withtheargumentthatithadfilleditsroleasafacilitatorformarketintroduction.In2012,theexemptionwasabolished for all vehicles. The intention of the exemption was to stimulate moreenvironmentally friendly vehicles, but the decision to equate “more environmentallyfriendly”with “alternative fuels” received criticismafter awhile, especiallywhen thepositiveenvironmentaleffectsofethanolcarswerequestioned.Theexemptionappliedto any car that could be propelled by other means than gasoline or diesel, such asethanol, biogas or hybrids, regardless of which fuel was actually used (biogas andethanolcanalternativelybedrivenomordinarygasoline)andregardlessofhowmuchemissions they generated. Subsequent Swedish policies intended to stimulate “greencars” have used other definitions, usually using constraints on fuel consumption perkilometer.Theimportantpolicyconclusionfromthe“greencar”exemptionusedinthecongestionchargingsystemisnottheparticulardefinitionofwhatconstituteda“greencar”–manywouldarguethatthisdefinitioninhindsightprovednotwellsuitedtotheintention – but that it showed that this kind of exemptions can be a very effectivestimulusonthevehiclemarket.There is no congestion tax levied on vehicles driving on the Essinge bypass pastStockholm.Thisistheonlyfree‐of‐chargepassagebetweenthenorthandsouthpartofthecounty.TheEssingebypasswasheavilycongestedevenbeforethecharges,sofromapuretrafficperspective,therewasastrongargumentforalsochargingvehiclesonthe

6:30‐7:00 10SEK7:00‐7:30 15SEK7:30‐8:30 20SEK8:30‐9:00 15SEK9:00‐15:30 10SEK15:30‐16:00 15SEK16:00‐17:30 20SEK17:30‐18:00 15SEK18:00‐18:30 10SEK


8

bypass. The opposition from the surroundingmunicipalitieswas so strong, however,thatthepoliticiansof theCityofStockholmdecidedthatthebypassshouldbefreeofcharge.

3.2 Traffic effects

Traffic across the Stockholm cordon had remained largely constant since the early1990’s,despitegrowingpopulation,andcarownership.Roadtraffichadgrownintherest of the region, so themost likely explanation for the lackof traffic growth in theinner citywas simply lack of capacity. Figure 2 shows average daily traffic volumesacross the cordon (weekdays 6:00‐19:00) since 2000. A slight drop is seen in 2005whentheSouthernBypasswasopened.WhenthechargeswereintroducedinJanuary2006,theyhadasubstantialeffectoncartraffic from day one. After a fewweeks with an even larger initial effect, the trafficreduction stabilized around 22% across the cordon during the charged period(comparingmonth‐by‐monthtoaccountforseasonalvariation).Beforethestartofthetrial, there was some doubt as to whether the traffic reduction would actually takeplace,especiallysincethetrialwasonlysevenmonthslong,buteffectsturnedouttobeimmediateandpersistent.When the trial ended July 31 2006 and the charges were abolished, traffic volumesimmediatelyreboundedtoalmostthesamelevelasbeforethecharges–butnotquite.A residual effect remained even after the charges had been abolished. From August2006 to August 2007, i.e. between the end of the trial and the reintroduction of thecharges, traffic volumes remained 5‐10% lower than in 20056. The most likelyhypothesis is that some car users developed new travel habits during the trial thatpersistedevenafterthechargeswereabolished.

Figure2.Averagetrafficvolumesacrossthecordon,weekdays6:00‐19:00excl.July.Blue:nocharges.Red:charges. “2006a” is the trialperiod January‐July2006,and “2006b” is theremainderof2006.

6Theexactsizeof theresidualeffect isuncertain,sincedatafromthisperiodare lessreliableduetoroadworksandtechnicalproblemswiththemeasurementequipment.


9

“2007a”istheperiodJanuary‐August152007,beforechargeswerereintroduced,and“2007b”istherestof2007.

WiththereintroductioninAugust2007,trafficlevelswereagainreducedtothesamelevel as during the trial period in 2006. Since then, they have remained roughlyconstant,despiteinflation,economicgrowth,growingpopulationandanincreasingcarfleet.Ratherthanattenuatingovertime,theeffectofthechargesseemstohavegrownover time – otherwise, traffic would have been expected to increase due to variousexternal factors (populationgrowth, inflationetc.)– seeTable1.Controlling for suchexternalfactors,Börjessonetal(2012)showthattraffic’selasticity7withrespecttothechargeshasapparentlyincreasedfrom‐0.70in2006to‐0.85in2009andonwards.Itmaybetooearlytotellifthisisastable,long‐termvalue,althoughitseemslikely.Butthemostimportantconclusionisthattherearenosignsthattheeffectofthechargesiswearing off, but instead increasing somewhat over time. This is consistent with theobservationthattherearemoreadaptationmechanismsavailableinthelongtermthanintheshortterm.ThisresultisincorrespondencewiththatofGoodwinetal.(2004),who note that price impacts tend to increase over time as consumers have moreoptions.InSeptember2012,theexemptionforalternative‐fuelvehicleswasabolished.Thisledtoafurtherreductionoftraffic;beforethat,6‐8%oftrafficacrossthecordonhad been subject to that exemption. The observed reduction indicates that theexempted vehicles had the same price sensitivity as the rest of the traffic (a 20%reduction of these 6‐8 percentage points translates to a 1.2‐1.6% reduction of totaltraffic,whichisveryclosetowhatisobserved).Table 1.Traffic reduction across cordon compared to 2005 traffic levels (chargedweekdays 6:00‐19:00).Secondrow:reductioncompared toa theoreticalcounterfactualwhereexternal factorsarekeptconstant.(nosuchcalculationsareavailablefor2012‐2013).

2006a 2007b 2008 2009 2010 2011 2012 2013

Traffic reduction from charges, compared to 2005

‐21.0% ‐18.7% ‐18.1% ‐18.2% ‐18.7% ‐20.5% ‐21.4% ‐22.1%

Traffic reduction adjusted for changes in external factors8

‐21.4% ‐20.9% ‐20.7% ‐21.9% ‐21.7% ‐22.3% ‐ ‐

Detailedstudiesofthecharges’effectsgetlessandlessmeaningfulovertime,sincethecounterfactualbecomesincreasinglydifficulttopredict.Mostofthedetailedanalysesofeffects presented below are hence based on comparing 2005with 2006, i.e. the lastyearwithoutchargeswiththefirstyearwiththecharges.Sincetrafficlevelshadbeensostableforseveralyearsleadingupto2005,andhaveremainedsostablefrom2006onwards (withagap for theperiodAugust2006‐August2007), theessential insightsgainedfromthesecomparisonsstillseemrelevantandapplicable. In relative terms, thedecreasewas largest in theafternoonpeakperiod (‐23percentbetween16:00‐18:00),andsomewhat lowerinthemorningpeakperiod(‐18percentbetween 7:00‐9:00). This indicates that a larger share of discretionary trips ismadeduringtheafternoonpeakthaninthemorningand/orthatdeparturetimesfromwork

7 Note that these elasticities are neither comparable with the usual cost elasticity of car traffic, nor with the fuel price elasticity of car traffic. Since fuel costs make up around half the marginal cost of driving, the cost elasticity of car traffic is around twice the fuel price elasticity. The elasticity of traffic across the cordon with respect to the charge is higher, since there are more adaptation mechanisms available, such as changing route, destination or time of travel. 8 Fuel price, total employment, car ownership, inflation, exempted share of traffic.


10

arelessfixedthanarrivaltimestowork.Trafficdeclinedineveningsaswell.Hence,thereductionofoutboundtrafficduringeveningbecauseoffewerincomingvehiclesinthemorning outnumbered the increase of evening traffic because of within‐modesubstitution fromtravelduringday‐time to free‐of‐chargeevening.Despite the lowercharge during mid‐day (9:00‐15:30), traffic decreased almost as much during thisperiod (‐22 percent). The seemingly high cost sensitivity during this time period ispartiallyexplainedbythefact thatmanyof thetripscrossingthecordonduringmid‐daypaythehigherchargewhengoingintheotherdirection.Inaddition,effectsontrafficwereseenfurtheroutfromthetollzonethaninitiallywasexpected. The number of vehicle kilometers driven in the inner city decreased byaround16percent.Outsidetheinnercity,ontheoutlyingapproachroadsandoutlyingstreets, traffic volumes fell by just over 5per cent. These effectshave also remainedroughlyconstant;trafficvolumeshaveeitherremainedconstantinthefollowingyears,orfollowedtheirlong‐runtrend.Queueswerereducedalsofarfromthecordonduetoreduced spillback congestion.Hence, the chargesdonot seem tohave generated anyseveresecond‐bestproblemsneitherintheshortnorinthelongtermperspective.Consequently,unwantedsideeffectsthatwereanticipatedoutsidethecordon,suchasanincreaseoftrafficoncircumferentialroadsatthecity’soutskirts,werenotfound.

3.3 Travel times

Even more dramatic than the reduction in the number of vehicles was the reducedcongestion.ThenewspaperheadlinesinFigure3provideanillustration.Improvementsin travel timeswere tangible and easily perceivedby the general public. Travel timeimprovements also occurred far from the inner city, when spillback queues weresubstantiallyreduced.

Figure3.ThefrontpagesofMetroandDagensNyheterafterthefirstdayofcongestioncharging.(Theheadlineinbothnewspapersreads:“Everyfourthcardisappeared”.)

Figure4 showaverage congestion indices for various typesof roads, comparingMay2005 andMay 2006. Congestion ismeasured by a congestion indexwhere 0 percent


11

corresponds to free-flow travel time, while 100 percent corresponds to twice the free-flow travel time.Ascanbeseen,traveltimesforvehicletrafficdeclinedsubstantiallyinsideandclosetotheinnercity.Particularlylargedeclineswereseenonarterials,onwhichdelay times (time in excess of free‐flow travel time) fell by one‐third during themorningpeakperiodandbyone‐halfduringtheafternoon/eveningpeakperiod.Thisconsiderably improvedreliabilityof travel times, i.e., travellerscouldbemorecertainabout the duration of a car trip. The figure also shows the range of travel timevariability,measuredasthedistancebetweenthehighestandlowestdecilesoftraveltimes (during a given 15‐minute period). The highest decile of the travel‐timedistributionfelltoathirdorlesscomparedtothepre‐trialstateforsomecategoriesofroads(suchasarterialsduringPMpeak).

Figure4.Relative increaseof travel times forvariouscategoriesof links.0percentcorresponds tofree‐flowtraveltime.Thecolouredbarsshowaveragetraveltimeswhilethe“errorbars”indicatetheworstdecileand thebestdecileof the travel timedistribution.Measurementswere taken fromallweekdaysforsixweeksinApril‐May.”AMpeak”refersto7.30‐9.00,”PMpeak”refersto16.00‐18.00.

Figure5showsasimilarpicture,butshowingmoreyearsandanothercategorizationofstreetsandroads.Thefigureshowsthreeperiodswithoutcharges(April2005,October2005andOctober2006)andtwoperiodswithcharges(April2005andOctober2007).Generally, congestion is worse in April than in October, due to slightly larger trafficvolumes and more pedestrians and bicyclists interacting with the traffic flow. It isapparent that the drop in traffic volumes across the cordon translated to congestionreductioninlargeareabothinsideandoutsidethecordon.Inparticular,thecongestionreductionontheouterarterialsshouldbenoted,sincethesemeasurementpointsaresituated rather far away from the actual cordon. Moreover, it can be seen that thegeneralcongestionlevelwasaroundthesameduringthetrial(April2006)asafterthereintroduction(Oct.2007).Congestionlevelsduringthenon‐chargedOctoberperiods2005and2006wereabout equal,while the congestion levelduring thenon‐chargedApril2005wasabithigher.

0%

50%

100%

150%

200%

250%

300%

Arterialsinwards

Arterialsoutwards

Streets inside Roads inside

AM 2005

AM 2006

0%

50%

100%

150%

200%

250%

300%

350%

Arterialsinwards

Arterialsoutwards

Streets inside Roads inside

PM 2005

PM 2006


12

Figure5.Averagetraveltimeincrease over free-flow travel times for various categories of links, April and October 2005-2007.

Thereducedcongestionalsomeantthattraveltimereliabilityincreased.Forexample,theworsttraveltimedecilewasreducedbya factorof3ormoreforarterialsduringthePMpeak.Thescattergrambelow(Figure6) illustrates the increase in travel timereliability,showingstandarddeviationforlinktraveltimesfor2005(onthex‐axis)and2006(onthey‐axis).Nexttoitisasimilarscattergramshowingthecongestionindicesforseveral links. Inspectionofthediagramsrevealsthatmostpoints liebelowthe45degreeline,i.e.thesituationhasimproved(ifithaschangedatall).

Figure6.Scattergramsof link congestion indices (left)and standarddeviationof link travel times(right), April 2005 and 2006.

3.4 Environmental effects

The reduction in vehicle kilometers travelledmeant that emissions from trafficwerereduced.Thereductionwaslargestintheinnercity,between10and15percent(thereduction differed across different types of emissions). Since this is a most denselypopulatedarea,thisisanimportanteffectfromahealthpointofview.In the inner city, of air-borne pollutants were reduced between 10 and 14 percent. For nitrogen oxides (NOx) the reduction was smaller (8.5 percent), since the extended bus traffic used older buses with higher emission factors. Overall, the results reveal that air quality was improved in many

0%

50%

100%

150%

200%

250%

inner city, AM inner city, PM arterials, AM arterials, PM outer arterials, AM outer arterials, PM

Apr -05 (no charges)

Oct 2005 (no charges)

Apr -06 (w. charges)

Oct 2006 (no charges)

Oct -07 (w. charges)

0.00

0.50

1.00

1.50

2.00

2.50

3.00

3.50

4.00

0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00

Congestion index 2005

Co

ng

es

tio

n in

de

x 2

00

6 (

w C

C)

AM

(45 degrees)

PM

0

50

100

150

200

250

300

350

400

450

500

0 100 200 300 400 500 600

Standard dev. 2005

Sta

nd

ard

dev

. 200

6 (w

ch

arg

es)

AM

(45 degrees)

PM


13

streets in the inner city. Carbon dioxide emissions from traffic in the whole metropolitan area (the county of Stockholm) decreased by 2-3 per cent. Theuseofairpollutionmodellingallowedforadetailedassessmentofthechangesinpopulation exposure due to the change in traffic. Since the main effects were seeninside the cordon where the daytime population density is high, this also results inimportant changesof averagepopulation exposure.The estimated reductionsofNOxand exhaust particles emissions are refer to the change in concentration at rooftoplevelintheinnercity(so‐calledurbanbackground).Thisprovidesanindicationoftheaverage load of the population in this part of Stockholm. International researchascribesreducedmortalitydue to forexample fewercasesofcardiovasculardiseasesand lungcancerasthemost importanthealthbenefit.Forsbergetal. (2006)estimatethattherewillbe20‐25fewerprematuredeathsperyearinStockholm'sinnercityandatotalof25‐30lessprematuredeathsannuallyintheStockholmmetropolitanarea.9Theseareapproximatelythreetimeslargereffectsthanwhatwouldbefoundifamoregeneralpolicymeasure, suchas a fuel tax increase,wasused toobtainadecreaseofemissionsofanequalmagnitude,sincethesereductionswereconcentratedtothemostdenselypopulatedareas.Oneofthegoalsforthetrialwasto“improvetheperceivedurbanenvironment”. It isdifficult to draw definitive conclusions about this, not least because of confoundingwith weather effects. However, studies indicate certain improvements for traffic‐related indicators such as perceived traffic congestion, air quality and accessibility –not only for car drivers but also for cyclists and pedestrians. The result points toperceived improvementsof exactly those factors forwhichmeasuredchanges canbedemonstrated,i.e.thoseconnectedtotrafficreductions.Inthecityenvironmentstudy,citizens feel there is an improvement in traffic tempo, air quality and vehicleaccessibility.Thesametendencyisseenininterviewswithcyclistsintheinnercityandchildrenlivingintheinnercity.Inner‐citychildren’sperceptionofthecityenvironmenthasveryclearlyimprovedandmanycycliststhinktherearefewercarsintheinnercityandthatthetrafficenvironmenthasgotbetter.

3.5 Retail

There had been fears that retail inside the cordon would be adversely affected, butstudies of the retail markets were not able to show any effects of the congestioncharges (Daunfeldt, Rudholm,&Rämme, 2009). For example, the durables survey inshopping centres, malls and department stores during the Stockholm Trial periodshowedthatthesedevelopedatthesamerateastherestofthecountry.Thesameheldforotherretailsectors.Fear for adverse impact on retail inside a cordon is common in many cities. LargeeffortsweremadeinStockholmtotracksucheffects,onlytoconcludethattheywereverysmallornon‐existent.Similarconclusionshavebeenreachedinothercitieswithcongestioncharges.Theremaybeeffectsonparticularstores,especiallyiftheylieclosea cordon, but the average effect in an urban centre is usually small. This should beevident, especially in the long term: if the retail market inside the cordon gets lessattractive, then floorspace rents will, in equilibrium, decrease to counteract this,makingtheeffectonthenumberofstoresevensmaller.

9 The standard cost-benefit model used for infrastructure planning in Sweden has a dose-response relationship based on older studies. This indicates that the reduction in traffic due to the trial saved about five life years on an annual basis. This is the number used in the cost-benefit analysis by Eliasson (this issue).


14

3.6 What did the disappearing drivers do instead?

Trafficmeasurementswill only give changes in aggregate terms, anddoesnot revealhow car drivers adapted – how many switch to other modes, departure times ordestinations,forexample.ApaneltravelsurveywithtwowaveswasusedbyFranklin,KarlströmandEliasson(2010)to investigatehowprivatetripswereaffected.Results(in 1000’s car trips across the cordon during charged hours) are shown in Figure 7.These results shouldbe treatedwith caution, since it isdifficult to separate seasonalandtrendeffectsfromtheeffectsofthecharges.Approximately25%persontripsacrossthecordondisappeared.Outofthese,around10percentagepoints outwerework trips switching to transit,while onepercentagepointwasworktripsswitchingtotheEssingeBypass(hence,routeswitchingwasonlyaminoradaptationstrategy).Sixpercentagepointswerediscretionarytripsswitchingtootherdestinationsorreducingtripfrequency,possiblybytripchainingorcombiningtrip purposes, and under one percentage point switching to the Essinge Bypass. Theremainingfivepercentagepointsaredisappearingprofessionaltraffic—deliveries,taxi,craftsmenetc. Sincewedonothave travel surveys for this typeof traffic,we cannotdecompose it further. That professional traffic is affected at all by the relatively lowcharges may come as a surprise, but there is significant evidence in the interviewstudiescarriedoutwithprofessionaldriversthattheyinfacttriedtoplantheirroutesandtripchainsinordernottocrossthecordonunnecessarilyoften,andmoreovertodecreasethenumberoftripsaltogether.

Figure 7. Estimated changes in car trips across the cordon during charged hours (1000’s trips).(When comparing tomeasurements of vehicles across the cordon, note that “through” trips arecountedastwocordoncrossings.)

It isworthemphasizingthat therearemanywaystoadapt.Routeandmodechangesarefarfromtheonlyadaptationstrategies.Trips,especiallydiscretionarytrips,arenot“replaced”inasimpleone‐to‐onefashion.Manypeople,trafficexpertsnotleast,seemto be unconsciously stuck with the assumption that there is a more or less fixednumberoftripstobemade,andthattheeffectofthechargesshouldbepossibletosortneatly into categories like “mode change”, “destination change” and “departure time

Trips

Work - change dep. time; 2

Work - remaining; 119

Work trips - to Ess.; 2Discretionary - to Ess.; 1

Discretionary - "disappeared"; 22

Discretionary - remaining; 82

Professional traff ic - remaining; 115

Professional traff ic - "disappeared"; 20

Work - to transit; 36


15

change”. In reality, adaptations are much more multi‐faceted. This means thatcommonlyencounteredstatementssuchas“congestionchargingwon’tworkinourcitybecauseourtransitsystemistoobad”or“…becausewehavenoringroad”aremissanessential point: there are many more ways to adapt than changing mode or route.Whichadaptationstrategywilldominatedependsonthecharacteristicsofthecityandthedesignofthesystem,andwhichtravelalternativesitleavesopen.Anotherimportantinsightisthattrafficisn’tjustworktrips.Worktripsonlymakeupafraction of car traffic – a typical figure could be 30‐40%, with the rest beingdiscretionary trips andprofessional traffic (wherea typical figure couldbe15‐25%).Discretionary trips are easier to affect, because there aremoreways to adapt in theshort run, and represent a significant fraction of traffic, especially during afternoonpeak hours, when congestion is often just as severe as during the morning peak.Professional trips are very heterogeneous: some types are very difficult to change,while some are not. Typically, values of time are very high, which means that timesavings for professional traffic will constitute a significant part of the travel timebenefits.Despiteall this, it’scommonthat thediscussion focusesexclusivelyonworktrips,bothamongplanners,policymakersandthegeneralpublic.Thisisamistakethatsurprisinglyoftenconfusesthediscussionaboutwhatcongestionchargingcandoandhowtheymaywork.Overtime,thequestionofhowdriversadaptedbecomesincreasinglypointless.Thisisbecause travel patterns aremuch less repetitive and stable thanmany people think.Manyoftheaffecteddriversareoccasionalcardrivers,whodriveonthechargedroadperhapsacoupleof timeseachmonth.Around two thirdsof thedrivers crossing theStockholmcongestionchargingcordononanygivendayare”occasional”drivers,whodriveacross thecordon threedaysaweekor less.Otherdays, theyuseothermodes,timesorroutes.Thesedriverswillchangeonthemargin,anditwilloftenbeimpossibleto tell if andhow they changed. In fact,many cardriverswill not evenknow if or inwhat way they adapted. A study in the spring of 2006 showed most drivers wereunawarethattheyhadreducedtheirtripsacrossthecordon.Acomparisonofdrivers’statedchangeinbehaviourandobjectivetrafficmeasurementsshowedthataround¾ofthedecreaseintripshadapparentlygoneunnoticedbydrivers.Moreover,therearemanyotherchangingprocessesgoingon.Peoplemoveandchangejobs,forexample:betweenanytwoyears,20‐25%oftheworkforcewillhavechangedjobs(orstartedworking),and15‐20%ofthepopulationwillhavemoved.Afterjustafewyears,itispointlesstoaskhowagivenpersonhas“adapted”–becausetheentiresituationwheretravelchoicesaremadehaschanged.

3.7 Public transport

TheStockholmtrialconsistednotonlybycongestionchargesbutalsoofanextensionof public transit services. The extended services were motivated partly to meetincreaseddemandforpublictransport,andpartlybyapoliticalwilltoshow“carrots”andnotjust“sticks”(Kottenhoff&BrundellFreij,2009).Driversswitchingfromcartopublictransportmeantthatthenumberofpassengersinthetransitsystemincreasedbyaround4‐5%.Crowdinginthepublictransportsystem,measuredbythenumberofstandingpassengers,increasedsomewhatinthemetrobutdecreased on the commuter trains, most likely thanks to expanded public transportcapacity.


16

Reduced road congestion in and around the inner city led to increased speeds andpunctuality forbusservices.Duringthetrialperiod in2006,bus timetableswerenotadjusted so the improved accessibility did not significantly shorten travel times forinner‐citybuses, but therewere signsof improvedpunctuality.Bus traffic across thechargecordon–whichdonothavefixedtimetablesoncetheyhavepassedthecordon‐experienced considerably shorter travel times. After the trial period, no dedicatedstudies of this are available, since the counterfactual becomes more and morehypothetical.Travelsurveysshowedthatfewcardriverswereenticedtoswitchtopublictransportbytheserviceextensioninitself,evenifthedataareinconclusive.Thesecondpurposewas to provide additional capacity in the transit system. The added capacity wasrelativelyminor: the increase in passenger capacity by the busesmeant that around14000 trips were made each day by the buses, compared to well over one millionpublictransporttripsacrossthecordoneachday.Butinthespecificcorridorsservedby the buses, theymost likely contributed to keeping crowding on commuter trainsfromincreasing.Thethirdpurposeofthetransitextensionwastoincreasetheeffectofthe congestion charge by making the switch from car to public transport easier.However,onboardsurveysonthenewbusesfoundveryfewformercardriversonthebuses.Ofthevehicle‐trafficreductionof22%overthechargecordon,atmost0.1%canbeascribedtotheextendedbusservices(Eliassonetal.,2009).

4 PUBLIC AND POLITICAL ATTITUDES

The main obstacle for congestion charging is often the lack of public acceptability.Hence, the most remarkable and interesting development in Stockholm may be thechange in attitudes, from fiercely hostile to overwhelmingly positive. This sectionpresentssomeevidenceaboutthisdevelopment,anddiscusseswhatfactorsinfluencedit10.

4.1 The change in attitude – “familiarity breeds acceptability”

Whenthedecisionwasmadetocarryoutacongestionchargingtrial inStockholm, itwasmetwithvocalresistance,althoughpollsshowedamoremixedpicture.Figure8showshowthesupportforcongestionchargeshasevolvedovertime.Inthespringof2004, 43% of Stockholm citizens stated that theywould “probably” or “most likely”voteyestopermanentcongestioncharges.Supportfell,however,oncethestartofthetrialapproached.Rightbeforethestartofthetrial,supporthadfallento34%,withthe“most likely yes” group falling the most. Once the trial started, however, supportincreased to 53%. The media image also changed once charges were in place, fromintensely critical to, in many cases, very positive. The percentage of trial‐relatednewspaperarticleswithapositiveangleincreasedfrom3%intheautumnof2005to42%inthespringof2006,whiletheshareofnegativenewspaperarticleswasalmosthalvedfrom39%to22%(Winslott‐Hiseliusetal.,2009).Thetrialendedon31July,2006,andwasfollowedbyareferenduminSeptemberatthesametimeasgeneralandlocal‐government elections were held. Excluding blank votes, 53% of Stockholmcitizensvotedtokeepthecharges.Aftertheelection,thecentre/rightcoalitiongainedpowerbothatthenationallevelandinthecityofStockholm.Thecentre/rightcoalitionin Stockholm had opposed the congestion charges, but had promised to follow theoutcomeofthereferendum,sotheyhadtoaskthenationalGovernmenttoreintroducethe charges permanently. After a few weeks of consideration, the new centre/rightGovernment said it would do so, but as part of a broader package of transport10 This section draws heavily from Eliasson (2014).


17

investmentsinStockholm,tobenegotiated.Therevenuesfromthecongestionchargeswereearmarkedforroadinvestments.Ontheotherhand,theinvestmentpackagealsocontained major rail investments, but these were claimed to be financed by othersourcesoffunding.Afterthedecisiontoincludethechargesinaninvestmentpackage,nopoliticalpartiesproposedabolishingthemanymore.ThechargeswerereintroducedpermanentlyinAugust2007,althoughthenegotiationoverrevenueusewasnotsettleduntil late2007.Apoll inDecember2007showeda65%support forthecharges,andseveralpollssincethenhaveshownsimilarorhighersupport.Themostrecent,inlate2013(notshowninthediagram),showedasupportof72%(excluding“don’tknow”).

Figure8.Wouldvote”yes”inreferendumaboutcongestionpricing(excl.”Don’tknow”).

Figure 9 shows support for the charges in four groups: people without car in thehousehold,carownerswhoneverorveryseldomcrossthechargecordon,carownerswhosometimespaythecharge,andcarownerswhooftenpaythecharge.(2007and2010 data are missing, since these surveys did not ask about driving across thecordon.)ThesupportshowsthesameU‐shapeineachgroup.Infact,thedipfrom2004to2005ismostpronouncedfortheunaffectedgroups,i.e.peoplenotexpectingtopaythecharge.


18

Figure9.Supportforcongestionchargesdependingoncarownershipandpaidcharges.

Evidently,theamountoftollspaidmakesalargedifferenceforthesupport,lookingateachcross‐section.Butthefigurealsoindicatesthatthechange inattitudesisat leastpartlydrivenbyother factors than self‐interest variables such as tolls paid and timegains.Thechangesinattitudeovertimeforeachgrouplookverysimilar,regardlessofhowmuch theyareaffectedby thechanges in travelcostsand travel times.Even thegroupwhodonotownacarshowsthesameU‐shapedattitudepattern.Andeveninthemostaffectedgroup,supportforthechargeshasmorethantripledfromalowpointof15%to53%.Inotherwords,thereisamajorityinfavourofthechargesinallgroupsby2011.Justasrespondentsseemtobeunawareoftheirbehaviouralchanges,theyseemtoberemarkablyunabletoremembertheirpastattitudes.Thisisactuallywellknownfromvoting research: people tend to forget that they have ever had another opinion thantheircurrentone,orhaveevervoteddifferentlythantheywouldvotenow.Surveysin2006 and 2007asked respondents whether they had changed their attitude to thecharges. From 2005 and 2006, voters intending to vote ”yes” in the referendumincreasedby19percentagepointsfrom30%to49%(includingundecidedvotersinthebase).29percentagepointsofthe49%yes‐votersstatedthattheyhad”becomemorepositive” during 2006 – that is, some of the positive votersmust have become evenmore positive. Butwhen the same question is repeated in 2007, only 13%of votersstatethattheyhad”becomemorepositive”during2006.Notonlyisthislessthanhalfthe 29% from2006, it is lower than the 19 percentage points that became somuchmorepositivethattheychangedfrom”no”to”yes”.Thedevelopmentofattitudes is remarkablyclose to thegeneralpatterndescribed inGoodwin(2006),reproducedinFigure10.


19

Figure10.”Thegestationprocessforroadpricingschemes”–reproducedfrom(Goodwin,2006).

Once the idea has been introduced and explained, a fairly large fraction of thepopulationisgenerallywillingtosupporttheideaofcongestionpricing.Howlargethisfraction is depends on how the question is formulated and framed – for example,revenueuse, thepurposeof the charges andwhatpolicy alternatives it is contrastedagainst all matter. But once a detailed proposal is worked out, support generallydecreases.Theremaybeseveralreasonsforthis–forexample,thatthedisadvantagessuddenly become more evident than the potential advantages, or fears that thetechnical system will not work or become very expensive. This is sometimessummarisedintheformula“acceptabilitydecreaseswithdetail”.Butoncethesystemisin place, support will generally increase, which is often summarised as “familiaritybreedsacceptability”.Thereareprobablyseveralreasonsforacceptabilitytoincreaseonceasystemisinplace.

1. “Better than you thought”: the benefits may turn out to be larger thananticipated. A major reason for public resistance to congestion charges is abeliefthattheysimplywillnotwork(Bartley,1995;Jones,2003).InStockholm,the positive effects on road congestion and urban environment were muchlargerthanmostpeopleexpected.

2. “Notasbadasyouthought”:thedownsidesofcharges–increasedtravelcostsand/orchangesintravelbehaviour–mayprovetobenotasbadasexpected.Once the chargesare inplace,manypeoplemaydiscover that the chargesdonot in fact affect them negatively as much as they had feared. There is aevidence that this phenomenon was important in Stockholm (Henriksson,2009).Transitcrowdingdidn’t increasesignificantly,congestiondidn’tappearin new places, the technical systemworked smoothly, and it may have beeneasier to adapt to the charges than anticipated (for example, drivers havereducedtheircartripsacrossthecordonaroundfourtimesmorethantheyareawareof,accordingtoastudywhichaskeddrivershowtheyhadchangedtheirbehaviourandcomparedthistoactualtrafficmeasurements).


20

3. Familiaritywithroaduserchargingmayreducethegeneralreluctancetowardspricingapreviouslyunpricedgood.Thereisevidencethatpeopleinmanycasesdo not like prices as an allocationmechanism (Frey, 2003; Jones, 2003). Butoncefamiliarwiththeconceptthatroadspaceisinprincipleascarcegoodthatcanbepriced–much likeparkingspaceor telecommunicationcapacity– thisreluctancemaytendtodecrease.

4. Finally, it is a general psychological phenomenon to resist change. In thepsychological literature, different variants of this general phenomenon arereferred to as status quo bas, loss aversion, endowment effect, or cognitivedissonance.Thismeansthatoncepeoplehaveadaptedtothechargesandtheshort travel times, they value the time gains higher and the increased travelcosts lower than they did before the change. Some people may react with“accepttheunavoidable”:oncethechargesareinplace,itislessworthwhiletospendenergyonopposingthem.

Further analysis of Stockholm attitudes show that even if self‐interest and belief ineffects strongly affect attitudes, these are not the sole determining factor, especiallywhen looking at how attitudes change over time. All groups, regardless of travelpatterns, car ownership andbelief in the charges’ effectiveness (exante andexpost)show the same U‐shaped change in attitudes – more negative attitudes before theintroduction of the charges, and increasingly more positive attitudes after theintroduction.Infact,thispatternismorepronouncedforunaffectedgroups.Thiscanbeinterpretedusingresultsfromsocialpsychologyexplaininghowattitudesareformed.Firstofall,attitudesaremoreorlesswelldeveloped.Theytendtobemoredevelopedinissueswhereanindividualforexamplehasalotofdirectexperience,hasencounteredtheissuemanytimes,knowalotabouttheissueandtowardswhichtheyhavestrongemotions.Whenpeoplearefacedwithanewissuewhereattitudesarenotwelldeveloped,newattitudesareoften formedbyassociating thenew issue tosomesimilarbutfamiliarissue,wheretheindividualalreadyhasawell‐developedattitude.Therehavebeensuggestionsthattheintroductionofcongestionpricingestablishedanewsocialnorm,wheredrivinginrushhourswaslesssociallyaccepted.Thereisscantevidenceofthis,however.Infact,itiseasiertointerprettheintroduction,polarizationandsubsequent lackofcontroversyintheoppositeway–thatthechargescouldgainbroad acceptance once itwas shown that it could in fact be reconciledwith existingsocialnorms:inparticular,thatthechargeswerenotanattackonmobilityorcaruseassuch.Thisinterpretationsimultaneouslyexplainswhyitwaspoliticallyrationaltofirstignore congestion charges, then to advocate them in spite of public resistance fromlargegroups,andwhythisresistancethendieddown.Italsoillustratestheimportanceoflegitimacy.Thestoryhasfourphases.

1. Fora longtime,plannersandeconomists inStockholmhad limitedsuccess inadvocating congestion pricing. Their argument was that it would increasesefficiency in the transport system. But very few people have a pre‐existingattitude to this concept, and even fewer have an emotional engagement in it.Instead,whenfacedwiththequestion,peopleassociatedtosuperficiallysimilarissues such as mobility restrictions or taxation, where they have an existingattitude,mostlikelyanegativeone.Foranissuetobepoliticallyinteresting,itmustgenerateenthusiasmamongasufficientlylargegroupofvoters.Butsincetransport efficiency is simply not an issue thatmany people get enthusiasticabout,theissuehadvirtuallynopoliticalupside.

2. Thischangedwhencongestionpricingwasreinterpretedasanenvironmentalpolicy,whichhappened inStockholmduring themid1990’s.While allocation


21

efficiency in the transportsectorcouldnotarouseenthusiasmorengagementamong the general public, environmental concerns definitely could. Thiswaswhatwasneededtogetcongestionpricingonthepoliticalagenda–alinktoanareawithwherestrongandemotionalattitudesexisted.

3. Whenthedecisiontocarryoutthecongestionchargingtrialwasmadeaftertheelectionin2002,afiercedebatebrokeout.Consistentwithwhatwassaidaboveabout the necessity of emotions in politics, the arguments soon turnedprincipal,moralandemotional,leavinglittleroomforcompromise.Thismighthave been an inevitable development: if congestion pricing had not beenelevatedtoamoral‐emotionalquestion, ithadn’tenteredthepoliticalstageinthe firstplace.But justas inevitable, themorally superchargedarguments forcongestionpricing implied(orcouldbeperceived to imply) thatall car trafficwasevilandunnecessary,andshouldbebanished.Thismightbeonereasonforthe decreasing support also among car drivers thatwere actually unaffected:theymightsimplyhavebeenalienatedbytheanti‐carrhetoric.Therearealsoother reasons that may have caused non‐affected groups to develop morenegativeattitudes:thechargeswereclaimedtohaveadverseequityeffects,beunfair,andawasteoftaxpayers’money.Themostrecurringargumentwasthelack of democratic legitimacy. The social democrats had made a very clearpromise not to introduce road pricing during the election period – and heretheyweredoingitanyway.

4. When the referendum ended in a yes to the charges, the new governmentdecided to earmark the revenues for a motorway tunnel west of Stockholm.Fromanattitudepointofview,thiswasprobablyimportantforseveralreasons.First,thechargesnowhaddemocraticlegitimacy.Inadditiontothereferendumresult,therewasnowapoliticalagreementaboutthechargesandtherevenues– made by the liberal/conservative alliance no less, which meant that allpolitical parties had now sanctioned the charges in some way. Second, therevenues were earmarked for roads. As it was really part of a multimodalpackage, therevenuescould justaswellhavebeenearmarkedfortherailwayinvestments thatwere alsopartof thepackage.But earmarking the revenuesforroadsnotonlyspoketomotorists’self‐interest.Itsentamoralsignal:it’sOKto be a car driver. It indicated a reinterpretation or reassociation of thecongestion pricing issue from a morally charged anti‐car measure to atechnical‐rationalmeasurethatwaseffective– it“worked”inthesensethat itgenerated revenues and reduced congestion. And technical measures arousemuchlessemotions:peopleusuallydonotlovethem,buttheydonothatethemeither.Themostimportantfunctionoftheearmarkingmayhencehavebeentodischargesomeofthesentimentsaroundthecharges,movingthedebatefromthemoraldomaintothetechnical‐rationaldomain.Thirdly,itcalmedthefearsof Stockholmpoliticians (fromall parties) that revenueswould endup in thenationalcoffers,eitherdirectlyorindirectly,bysubtractingtherevenuesfromStockholm’s“fairshare”ofnationalinfrastructuregrants.

Theintroductionofacertainmeasurealwayshasthepotentialofbecomingpoliticized.Congestion pricing is an evident example: pricing as suchmaymeet resistance evenamong people who agree in principle that road traffic should be reduced. In fact,regulation (banson car driving in certain areas, odd/evennumberplate restrictions,slowingdowntrafficthroughphysicalmeasures)areoftenmorepopularthanpricingmeasures,despitebeinginefficientinthestandardeconomicsense.Muchofthedebatebeforethechargeswereintroducedcentredaroundwhetherpricingwas,inprinciple,afair, legitimateanddemocraticallocationmechanism.Afterthechargeswereinplace,this debate faded into the background – pricing seemed to become a reasonable


22

acceptedway toallocateroadspace, justaswithmanyotherscarcegoods(includingtransport‐relatedthingsastraintickets,airportslotsorcarownership).Onewaytoviewthecongestionchargingdebateisthattheissueseemstohavemovedfrom the technical‐rational domain to the moral domain and back again. Whenpresented as a purely technical‐rational suggestion, it failed to gain political interestbecausethetypeofbenefits itcouldpotentiallybring(increasedtransportefficiency)could not generate sufficient enthusiasm. It seems that too few people had strongattitudes regarding this type of benefits, meaning that it could not generate thenecessaryemotiontogainpoliticaltraction.By reinterpreting it as an environmentalmeasure, the issuemoved gradually to themoral domain. This connected to strong attitudes regarding local and globalenvironment, andmaybealsogeneral anti‐car sentiments in somegroups, andhencethe necessary political engagement emerged. But the other side of the coin is that itmade the issue divisive – even unaffected groups became more negative. Thesenegativesentimentswereprobablybolsteredbyothermoralarguments,e.g.aboutlackofdemocraticlegitimacy,wasteofpublicfunds,over‐taxationandfreedomofmobility.Theseattitudeswerealsostrongandwell‐developed,makingthedebateveryheated.But after the referendum, the infrastructure agreement and the establishment ofdemocratic legitimacymoved the issue back into the technical‐rational domain. Thisdischargedsomeof thenegativemoral‐basedattitudes.Moreover, this connected theissue to concepts such as “rationality” and “efficiency”, which in Sweden have verypositiveassociations.

4.2 Other factors influencing public attitudes

Several factors influencing attitudes to congestion charges have already beenmentionedabove,but it isworth summarizingwhatkindsof factorshave repeatedlybeenfoundtoinfluenceacceptabilityinseveralcountries.First, self‐interest variables are obviously important. All else equal, individuals getmorepositivethelesschargestheypay(orexpecttopay),themoretimegainstheyget,thehigher theyvaluetravel timesavings,andthemoresatisfiedtheyarewithpublictransport. Individuals also becomemore positive if revenues are used in away theyappreciate,whichcanbeviewedasa formofself‐interest (Eliasson& Jonsson,2011;Hamilton&Eliasson,2012;Hårsman&Quigley,2010;Schade&Schlag,2003a).Amongotherthings,thismeansthatthesystemhasgottodeliverbenefits. InStockholm,theperceivedeffectof the chargeswas themost important factor explaining attitudes tothecharges.Evenifoneshouldnotconfuse“perceived”effectswith“objective”effects–since attitudes influences what effects are actually perceived – it seems clear thatachieving objective effects is necessary to reach acceptance. This underscores theimportanceofdesigningthesystemcarefully,andonlyusingcongestionchargeswhencongestion really is a problem.Moreover, it seems likely thatmeasuring effects andcommunicating the results through, for example, the kind of scientific evaluationcarriedoutinStockholmwill increasetheawarenessofpositiveeffects–provided,ofcourse,thatthereareinfactpositiveeffects.Second, positive attitudes to congestion charges are strongly influenced by concernsaboutandengagementinenvironmentalissues(Eliasson&Jonsson,2011;Hamilton&Eliasson, 2012). In Stockholm, the charges were to a certain extent marketed as“environmental”charges,andvoters’environmentalconcernwasan important factorexplainingtheacceptabilityofthecharges.Thisisinlinewithfindingsintheliterature


23

that social norms of this type influence acceptability in general, and that supportdependsnotonlyonthe“objective”characteristicsofthemeasureitself,butalsoonthedefinedobjectiveofcongestioncharges.Moreover,severalauthorshavefoundthatitisnot justperceived individualbenefits thatdetermineacceptability:perceivedsocialorcollectivecostsandbenefitscanalsoaffectacceptabilitystrongly.Hence,the“branding”of the charges matters – how they are marketed, explained and perceived. InStockholm, re‐labelling congestion charges to “environmental charges” andemphasizingtheirpositiveeffectsonairqualitymayverywellhavehadanimpactonacceptability.Othercitiesmayemploydifferentstrategies,but thegeneralconclusionremains: it is important how the charges are “branded”. A condition for this to bepossibleisthatthesystemdesigniswellalignedwiththestatedpurposeofthecharges.Asystemmarketedas“congestioncharges”systemshouldforexamplenotlevychargeswhereorwhenthereisnocongestion.Aswasemphasizedabove, themost important factor seems tobeownexperienceofcongestionpricing.Thesamepatternhasbeenobservedine.g.OsloandLondon.Onestudy(Hamilton&Eliasson,2012)comparedattitudesinStockholm,HelsinkiandLyon,concluding that the only variable that could explain the much higher support forcongestion charges in Stockholm was simply that the Stockholm population hadexperiencedtheintroductionofcongestionpricing,whiletheothershadnot.Congestion is viewed as one of themost important urban problems (65‐80% of thethreepopulationsagreed).ButHamiltonetal.foundnosignificantcorrelationbetweenattitudestocongestionpricingandconcernsaboutroadcongestion.Ontheotherhand,theyfoundastrongcorrelationbetweenconcernsaboutroadcongestionandbeinginfavour of expanding road capacity. Apparently, it is not mainly concerns aboutcongestion that is driving support for congestion pricing, despite the fact that suchconcernsarewidespread11.This isconsistentwithearlierseveralstudies findingthatone of the most common arguments against congestion pricing is a distrust incongestionpricing’sabilitytoreducecongestion(Jones,2003;Schade&Schlag,2003b).Ontheotherhand,thisdistrustmaypartiallybeareflectionofself‐interest:SchadeandBaum (2007) find that respondents who expect congestion pricing to bedisadvantageous to themselves not only havemore negative attitudes to it, but alsoperceivethemaslesseffectiveandmoreunfairthanotherrespondents.Congestionpricingattitudesarerelatedtoattitudestopublicinterventionsingeneral.Hamilton et al. show that negative attitudes to congestion pricing are stronglycorrelatedwithnegativeattitudes to taxation ingeneral,speedenforcementcameras,andbeliefinapublicadministration’sabilitytodistributeascarceresourcefairly.Thisfindingmaypartlyexplaintheapparentparadoxthatleft‐wingpartiesareoftenmoreinfavourofcongestionpricingthanliberal/conservativeparties.Equityeffectsareoftencitedasoneof themainreasons foroppositiontocongestionpricing.Whether congestion pricing has progressive or regressive effects depend onthe design of the system and on initial travel patterns. As to Stockholm, there areseveralstudies(Eliasson&Levander,2006;Eliasson&Mattsson,2006;Franklinetal.,2010; Karlström & Franklin, 2009). These find no regressive effects; some indicateprogressive effects, while some indicate neutral effects. It should be noted that theequity argumentmaybe used for other reasons thanhonest equity concerns: itmaysimplybeperceivedasamorelegitimateargumentthanself‐interest(Schade&Baum,

11 Other studies have got mixed results on this issue. While Rienstra, Rietveld and Verhoef (1999) found that respondents with more concerns about congestion were more positive towards congestion charges, Hårsman et al. (2000) and Schade et al. (1999) found the opposite.


24

2007). This is supported by the finding that Hamilton et al. found weak or nocorrelationbetweentheattitudetocongestionpricingandagreeingwiththestatement“Moreshouldbedonetoreducethedifferencebetweenrichandpoorinsociety”.Equity and “fairness” can be interpreted in different ways. Initially, the dominatingperspective is often “before‐after” – how travel costs and travel times change fordifferentgroups,suchasrichvs.poor,menvs.women,innercityvs.suburbresidents.Atleastincitieswithdecenttransitshares,itisoftenthecasethat“rich”willpaymorethan“poor”,withmiddle‐incomegroups“suffering” themost, relativelyspeaking.Butonce the charges are in place, another perspective becomesmore important – “fair”pricing.Inotherwords,whatpriceis“fair”tocharge?Fromthisperspective,itis“fair”that one pays more to drive on a congested road or to cause emissions in denselypopulatedareas–irrespectiveofincomeorplaceofresidence,forexample.

4.3 Political acceptability

Political acceptability is different from public acceptability. Obviously, politicalacceptabilityisinfluencedbythelevelofpublicacceptability–butpublicacceptabilityis neither anecessarynor a sufficient condition forpolitical acceptability. Crucial fortheanalysis andunderstandingofpolitical acceptability arepower issues: thepowerover the design of the charging scheme, the power over the revenues, and how thecharges and their revenue stream will affect decisions and funding of transportinvestmentsingeneral.ThefactthatcongestionchargesarenowpoliticallyacceptedinSwedenisnotonly,orperhapsnotevenprimarily,duetothehigherpublicsupport.Itisalsobecause thechargeshavebeen integrated in thegeneral transport investmentplanning process, and this has – at least partly – solved the power and negotiationissuesabove.Tounderstandthepoliticalandinstitutionaldriversbehindthisdevelopment,onemuststartwiththelegalcontext.Swedishcongestionchargesarenot“charges”butnational“taxes” from a legal point of view. Existing infrastructure cannot be “charged”, only“taxed”, according to the constitution’s definition of a charge, and Swedishmunicipalities12cannotlevytaxesonotherthantheirowncitizens.Hence,althoughitwasthecityofStockholmthatwasresponsiblefordesigningthechargingsystemandcarryingout the congestioncharging trial, the responsibility for actually levyingandadministering the charges had to be assumed by the national government through aparliamentarydecision13.Moreimportant,thismeantthatitisthenationalgovernmentthat has the formal power over both scheme design and revenues. Although theGovernmentpromisedtorefundtherevenuestotheStockholmregion,disagreementsquicklyemerged regardinghowrevenues shouldbe calculated,howrevenues shouldbeusedandwhichvehiclesshouldbeexempt.Furtherdisagreements,suchaswhetherandhowchargelevelsshouldchangealongwithinflationandeconomicgrowth,canbeexpected.Many politicians have stated that theirmain argument against introducingthe congestion charge was the uncertainty about the political power over schemedesignandrevenues.Addingtotheseuncertaintieswastheuncertaintyabouthowtheexistenceofthenewrevenue stream would affect the complicated negotiation between national andregional levels about national infrastructure grants. Most of the major transport12 A ”municipality” (”kommun” in Swedish) is the smallest geographical administrative unit in Sweden, roughly corresponding to a city. Most of the spatial planning responsibility, including infrastructure planning, lies at the municipal level. 13 This task was given to the National Road Administration, and later moved to the National Transport Agency.


25

investments in Sweden are paid for by the national government, whereasmunicipalities and regions are responsible for local streets and transit operation. Asexpected, there is often disagreement on where the border between differentresponsibilitiesshould lie.Thepoliticians inStockholm,regardlessofpoliticalcolour,hadlongarguedthattheywerenotreceivingtheirfairshareofnationalinfrastructuregrants. Whether this claim was founded or not, it meant that the arrival of a newrevenue stream in the form of congestion charges was not necessarily welcomed.SeveralpoliticiansfearedthiswouldmeanthatStockholmwouldhavetopayanevenlarger share of transport investments with their ownmoney. The government, theyargued, would point to the revenues from the congestion charges and claim thatStockholmobviouslyneededevenlessnationalinfrastructuregrantsthanbefore.The solution to this dilemma was an agreement where the charge revenues werefundingpartsofamajortransportinvestmentpackage,wherethenationalgovernmentalsomadeamajorfundingcommitment–muchlargerthanhadbeenthecaseforalongtime.Thechargerevenueswereearmarkedfortheroadinvestmentsintheagreement,while the substantial rail investmentswere claimed to be paid forwithmoney fromother sources. With this, support for the charges had been secured from regionalpoliticiansofallparties.From a political perspective, it is often a decisive questionwho has the power overrevenues and charge levels. If it is the national level, then regions and cities willobviouslybemuchmore reluctant to introducecharges.Buteven if the regionkeepstherevenues,anotherissueisimportant:howtheexistenceofthisnewrevenuestreamaffectsthecomplicatednegotiationbetweennationalandregionallevelsaboutnationalinfrastructuregrants. InNorway, thisproblem ishandledbyadecision that thestate“matches”incomefromregionalchargeswithequalnationalfunding.ArecenttrendinSweden is that national funding is often leveragedwith regional funding, sometimesfrom congestion pricing or toll schemes. This has made congestion charges morepopular among politicians, which shows the importance of institutional context andincentives.Givingregionstheincentiveandopportunitytointroduceroaduserchargestoobtaintransport investments, where regional funds are leveraged by national funds, mayfundamentally change the transport investment planning process. There are severaladvantages:regionsaregivenanincentivetoprioritizebetweentransportinvestmentsandotherresponsibilities,astheyareforcedto“puttheirmoneywheretheirmouthis”.When there is congestion, regions are more likely to introduce congestion charges,whichisobviouslyapotentandefficientpolicymeasure.Ontheotherhand,thereareseveraldisadvantages:sinceregional funding is leveraged,regionswillbe temptedtooverinvest in transport infrastructure relative to other types of (non‐leveraged)spending.

5 THE SYSTEM DESIGN PROCESS

5.1 Designing the charges

Thegoalofthesystem,asformulatedbythepoliticians,wasto“reducecongestiononthemostcongestedroadsandimprovespeedthroughthebottlenecks”.Thisgoalwasthenquantified, somewhat illogically, asadecrease in trafficvolumes: thenumberofvehiclescrossingthecordonshouldbereducedby10‐15%,anumberlooselybasedonpreviousproposalsforcongestionchargingsystems.


26

Basedontransportmodelsimulationsmadebeforethetriala10‐15percentreductionofthenumberofvehiclescrossingthecordonwasexpectedduringcharginghours.Infact,thetrafficforecastspredictedalargerand,asitturnedout,morecorrectpredictedmagnitude (Eliasson et al. 2003a, 2003b, 2004) – but such a large decrease seemedunreasonableatthetime,eventothemodellersthemselves.Noforecastsweremadeofthe effects on accessibility, as the static network equilibrium models available forforecastingwereconsideredasnotbeingreliableforsuchpurposes.During the process of designing the charges using traffic models, it quickly becameapparentthatatargetwasneededthatwasdirectlyrelatedtotravelspeeds.Thetrafficvolumereductiontarget,althougheasytocommunicateandmeasure,simplyrevealedtoolittleinformationabouttheeffectontravelspeeds.Anobviouscandidateforsuchan alternative target was to try to maximize social surplus. However, this wasproblematicbecause the travel timesof thestaticnetworkequilibriummodel (whichwas the only available traffic model) were judged to be too unreliable at highcongestion levels. Instead, themost important target usedduring the designprocesswasthecongestionlevelsatmajorbottlenecks.Several variants of the charging scheme were tested, of which the most importantvariationswerehigherorlowerchargelevelsandchargesonthesomeofthebridgesinside the cordon. There is a potential conflict betweenmaking congestion charging“efficient” in the theoretical sense, andmaking it easily understood (Bonsall, Shires,Maule, Matthews, & Beale, 2007). In the design of the Stockholm system, ease ofunderstandingwasamajorconsideration.Thisexplainsthesymmetriesofthedesign:onesinglechargingcordonwiththesamechargeatallpointsofentry,thesamechargein both directions, and the same for both the morning and afternoon/evening peakperiods. The system could most likely have been made more efficient if thesesymmetrieshadbeenabandoned–butatthetime,easeofunderstandingwasjudgedtobemore important than improving efficiency, and it is possible that thismight havecontributedbothtoacceptabilityandeffectiveness.A lesson from trying out different designs in the forecastingmodelswas that itwaseasytodesignachargingsystemthatcreatedmoreproblemsthanitsolved,by“movingaround”congestion.Hence,spendingalotofeffortonthedesignofthechargingsystemisimperativeforanycitywantingtointroducecongestioncharges.TheStockholmTrialprovidesinterestinginsights intowhataroad‐tollsystemshouldlook like – something which is also useful for other cities. Transport planners andeconomistshave longdiscussed towhatextenta charge‐zone tollof thekindused inStockholmissufficient forcontrollingtraffic inanentirecity.Trafficrelationschangefromstreettostreetandfromminutetominute.Whenthechargezoneisaslargeasitis in Stockholm, therewas concern that even if it had a big effect on travel over thecharge cordon, streets inside the zonewould soonbe full ofmotorists already in thezone increasing travel as they realized the streets were less congested. Alternativesolutions were discussed for several years prior to the Stockholm Trial, involvingseveralsub‐zoneswithvaryingratesofthecongestiontax. Noneoftheexistingroad‐toll systems threwmuch lighton thisquestion. InLondon, it is aquestionof a smallarea in the city centre, in Singapore access to cars is also regulated and inOslo andBergenthesystemisdesignedtoaffecttrafficaslittleaspossible.TheStockholmTrialconfirmsthatasimplecharge‐zonetollcreatessignificanteffectswithinalargearea.


27

5.2 Can transport models be trusted?

Eliassonet al. (2013)provideadetailed comparisonbetween forecastandoutcome,concluding that the main predictions about behavioural responses were sufficientlyaccurate to draw correct conclusions. For example, traffic across the cordon waspredicted to decrease 17% during peak hours and 16% during the entire chargedperiod (6:30‐18:30); the actual figures were 19% and 20%. The transport modelpredictedthataroundhalfofthedisappearingtripswouldswitchtopublictransport,whichwouldleadtoa6%increaseinpassengervolumes;inreality,aroundhalfofthedrivers did in fact switch to public transport, leading to an increase of passengervolumesof4‐5%.Infact,themodelseemstobemuchbetteratpredictingchangesinbehaviourthanthetravellers themselves,bothexante andexpost. Surveys in the fallof2004, the fallof2005andthespringof2006askedrespondentsaboutchangesintheirtravelpatternsin response to the charges. Respondents gave reasonably consistent answers in thethree surveys. The answers can be transformed to an equivalent aggregate trafficreduction, yielding an equivalent aggregate traffic reduction of 5‐10%. This can becompared to an observed reduction of private trips14 around 30%. In other words,around 3/4 of the reduction in car trips across the cordon seems to have goneunnoticedbythetravellersthemselves.The onemajor shortcoming of themodel systemwas the prediction of travel times.Sometraveltimeswerepredictedfairlyaccurately,butsomeeffectswerewideofthemark. This was the case especially for links subject to spillback congestion (queuespropagating back from a bottleneck, blocking other junctions and links). For linkscrossing the cordon, model‐predicted travel time reductions were close toobservations.Forlinkswithinthecordon,modelpredictionswerelessaccurate:delayreductions were underpredicted by 34 per cent, since the network does not modelcongestionatjunctions,trafficlightsetc.Forlinksoutsidethecordon,predictionswerecompletelyoff,sinceastaticmodeldoesnotcapturespillbackcongestion.Hence,itwasimpossible for thenetworkmodel to foresee that congestion outside the cordonwilldecrease when the queues propagating upstream from bottlenecks located at thecordonarereduced.Thiswasaknownissueduringthedesignofthechargingsystem.Ratherthanusingtraveltime‐basedevaluationmeasures(suchasconsumersurplusorcongestionindices),designtargetswereformulatedintermsofvolume/capacityratiosinthemostimportantbottlenecks.Theresultsindicatethatbest‐practicetransportmodelsseemtobereliableenoughtobe used as decision support and design tools even for substantial changes of thetransportsystem–providedthattheanalystsareawareofinherentlimitationsofthemodelandinterpretresultsaccordingly.Itshouldbestressedthatthepredictedeffectsof thechargeswereso largethatseveralexpertsconsideredthe forecastsunrealistic.As it turned out, however, the model gave much more accurate predictions thanexperts’ judgments, in addition to providing more detail and consistency. Thescepticism towards the predictions was understandable: the introduction of thecongestion charges affected the whole Stockholm transport system in a completelyunprecedentedway.Trafficacrossthecordondecreasedover20%,meaningthattrafficwasdowntolevelsnotseenthe1970’s,reducingqueuingtimesby30‐50%.Onsomelinksandroutes, theeffectswereeven larger.Despite this, circumferential trafficdidnotincrease,andpublictransportridershipincreasedbyjustafewpercent.

14 Total traffic reduction is less because professional traffic decreased much less than private traffic.


28

One way to judge whether the transport model was “good enough” is to considerwhetherthesystemhadbeendesigneddifferently,differentpreparatorymeasureshadbeen undertaken, or the scheme been abolished altogether if the forecast had beenperfect in all respects. Generally speaking, the answer is a qualified “no”. Thequalification is that the model’s deficiencies, in particular the lack of dynamiccongestionrepresentationanddeparturetimemodelling,mayhavehadbecomemoreofaproblemifthesystemdesignhadbeenmorecomplex,withmorechargingpointsandmore fine‐tunedtime/placedifferentiation.Butgiventhat thesystemdesignwasconstrained to be relatively simple, themodelwas good enough to answer themostimportant design questions: what traffic reduction was needed to reduce queuessignificantly,what charge levelswereneeded to accomplish this reduction, andwhatthesecondaryconsequencesintermsofpossibletrafficreroutingandtransitcrowdingwould be. One of the authors headed the design process, and can confirm thewide‐spread observation that the effects of a significant change in a complex transportsystemareusuallytoocomplexandmulti‐dimensionaltoforeseewithoutthehelpofamodel. Themost important advantage of using a transportmodelmay not be that itgivesexactanswers,butthatitgivescoherentanswers.Duringthedesignprocess,themodelrepeatedlygaveresults thatweresurprisingat first,butwereself‐evidentandeasy to explain intuitively after some thinking. The point is that these “intuitive”explanationsandconclusionshadnotbeenrealizedbeforethemodelresultshadbeenproduced. In this sense, themodel turnedout tobean indispensable tool for systemdesign and evaluation, and several design suggestions were discarded after modelresultshadshownthattheywouldnotworksatisfactorily.

5.3 General advice on designing congestion charges

ThegoalshavetobeexplicitandrelevantFirst, the system needs a goal. The goal may be to reduce congestion reduction,improveairquality,yieldrevenues,oracombinationofsuchgoals.Whateverthegoalsare, they need to be explicit. Moreover, they should be quantified, at least to someextent.Thisquantificationusuallyhastobedoneincooperationbetweenpolicymakersand traffic experts: setting up relevant goals and targets is harder than mostpolicymakersrealise.Goalsmustaboveallberelevantandconsistent.Specifically,oneshould at this stagenot specifically strive tomake them easy to communicate to thepublic.Communicationisimportant,butcomeslater.Thegoalssetatthisstagearetheones thatwillbeusedduring thedesignprocess,and theyneed tobeconsistentandrelevant,notnecessarilyeasytoexplainorsoundgood.Anexampleofaconsistentandrelevant goal thathappens tobe ratherdifficult to communicate is “achievemaximalsocial benefit from congestion reduction” (perhaps given some restriction on chargelevels).Acommonexampleofagoalpopularamongpolicymakersandcommunicatorsis “getting more people to choose public transit”. This is not a relevant goal forcongestioncharges,whichshouldbeobvious.(Therelevantgoalistomakelesspeoplechoose car during congested hours. If they instead choose transit, that’s fine; if theyprefer to adapt by cancelling trips or changingdeparture times or destinations, thenthis is just as fine.) Choosing ill‐formulated goals and targets will very likely causeproblemsduringthedesignprocess,attheveryleastcausingconfuseddiscussions.

Designingthechargesisajobforexperts.Designingachargingsystemis,asarule,averydifficulttask–howdifficultdependsonthe topology of the city. (For example, Stockholm is reasonably easy,with theworstcongestionproblemslocatedalonganaturalcordon,whileGothenburgisdifficult,withcongestionproblemsspreadingoutfromacomplicatedmultiple‐arterialjunction.)Itisabsolutelynecessarytohavesufficienttime,andaccesstoareasonablygoodtransport


29

model.Ifonehasaccesstodesignoptimisationtools,thiscancomeinveryhandy.Evengiventhis,itwillbedifficult.Inparticular,intuitionandpriorknowledgewillingeneralnot be sufficient, even for experienced traffic planners: transport systemsare simplytoo complex. There will almost certainly be surprises, and the first attempt at achargingsystemdesignwillmostlikelynotbeoptimalorevengood–itmayevenmakecongestion worse overall by “moving congestion around”. The system design is aniterativeprocess,where involvementofpoliticiansdoesnothelp.This iswhy it is soimportant that goals are stated clearly at the outset. Design and goal‐formulating is,ideally,an iterativeprocessaswell: it is likely thatsomegoalsor (more likely)somedesign restrictions were forgotten at the outset. But this does not change the basicpremise that while formulating goals and restrictions is a job for policymakers,designing thedetailsof thesystem– locationsand levelsof thecharges– isa job forexperts.Anill‐designedsystemmaynotonlybesuboptimal–itmaylikelycausemoreproblemsthanbefore.

Youneedagoodtransportmodel.Mosttransportmodelsareconstructedforotherpurposesthanmodellingtheimpactofcongestioncharging.Certain shortcomingsofmost currentmodelsbecomeespeciallyimportant in the context of congestion charges, and one needs to be aware of them.First, the value of time differs between vehicles, but this is often neglected in theassignment step. Technically, one must use multi‐class assignment, i.e. divide trafficintoseveral“classes”,eachwithavalueoftimeoftheirown.Thevalueoftimeofeachclasswilldecidewhether it isworthtakingadetour toavoidacharge.Dependingonthe network topology, the value of time distribution over classes may affect resultsstrongly.Often,thereis littleevidenceonthevalueoftimedistribution,sosensitivityanalyseswill play an important role. Second, departure time choices and schedulingconsiderations are often sketchily implemented, if at all. Obviously, this willunderestimate the impact of a time‐differentiated charging system, since theopportunity to adapt by changing departure time is not reflected in themodel. Lessobvious, onemay underestimate traffic decreases during non‐charged hours – sincethosetripsispartlymadeupof“returntrips”,i.e.thesecondlegofatripwhosefirstlegwas during the charged time period. Third, static assignmentmodelswill in generalunderestimate travel times in the presence of severe congestion. They will, amongother things, by definition neglect the effect of spillback congestion. Thismeans thatduring the design process, it may be better to focus on traffic decreases in knownbottlenecks, rather than to focus on actual travel times from a static traffic model(althoughtraveltimesneedtobeusedaswell).

Trytogetpoliticalandlegalpossibilitiestoadjustthesystemonceitisinplace.Evenwithcarefulplanning,surprisesarelikelytoappearwhenthesystemstarts.Inthebestcase,surprisesarepositive(inStockholm,traveltimeimprovementswerelargerthananticipated,forexample).Buttheremaybenegativesurprisesaswell:unexpected“rat‐running”, forexample.Becauseof this, it isgoodifonecangetpoliticaland legalleewaytomakeminoradjustmentstothesystemwithaminimumofdelayandhassle.Politically,thiswillbeeasierifgoalsareclearlyformulated:ifso,thenitwillbeeasiertoseeiftheyaremetornot,andifnot,thesystemcanbechanged.Thelegalproblemmaybehardertosolve.InSweden,forexample,thecharges(whichareformallyastatetax)have tobedecidedbyparliamentarydecisions,which requires a lotof timeandpoliticaleffort.

Thereisaconflictbetween“effective”and“easilycommunicated”design,buterringtowardstoosimpleseemsmorecommon.Policymakersoftenstressthattheywantadesignthatis“simpletounderstand”.Whileit is an important consideration that the systemmust be sufficiently simple for thepresumptiveuserstounderstand,policymakersoftenseemtounderestimatepeople’s


30

cognitiveability.TheSingaporesystemandtheUS“valuepricing”roads, forexample,appearcomplexatfirstglance.Thechargeisfinelydifferentiatedbytimeandlocation,andontopofthatmaychangeseveraltimeseachyear.Despiteitsapparentcomplexity,itturnsoutthatusersareabletograspandadapttothesystem.Forcingthesystemtobetoosimpletooearlyinthedesignprocessislikelytocausedesignrestrictionsthataredifficulttosolve.Thereluctanceofmanypoliticiansandplannerstoconsider“toocomplicated systems” can lead to the pointwhere the system becomes so simplifiedthatitwillnotdeliverthepromisedcongestionreduction.Thiswillnotonlybeawasteofresources–itwillalsoleadtolowacceptabilityofthecharges.

5.4 Information and technical design

In Stockholm, vehicles are registered automatically by cameras that photograph thenumber plates. During the trial, the main means of identification was transponders(“tag‐and‐beacons”orDSRC,dedicatedshort‐rangecommunication).Whenthechargeswere reintroduced, the automatic camera identification worked so well that it wasdecided to abolish the transponders. From the beginning, each daily charge washandled as one payment (one invoice per day), since it was believed that this wasimportanttoemphasizethateachpassagedidinfactcauseapayment.Aftersometime,however, paymentswere aggregated tomonthly invoices,without any consequencesfortrafficvolumes.Obviously,thishasreducedbothoperationscostsandthehassleforusers. In particular, the initial designwith a very large number of invoices and veryshort time allowed for payment (initially five days) caused major problems forbusinesses.CameraidentificationisusedinLondonaswell,althoughautomaticidentificationratesarereportedtobeconsiderablylowerinLondon.ThemaindifferenceisthatinLondon,it is the responsibility of the driver to pay the correct sum, and the systemmerelychecks that vehicles passing through the charging zone have paid. In Stockholm, thesystemautomaticallyissuesthechargeonceithasregisteredavehicle;ifthevehicleisnotidentified,thereisnoneed(orevenpossibility)topay.Thekeyargumentinfavourof“issuing/registration”ratherthansimply“checking”wasthatthismadeiteasiertoapplydifferentchargesaccordingtothetimeofday.The technical system was procured by the National Road Administration in a complicated tendering process eventually won by a consortium led by IBM. Information to the public was handled by the Road Administration. The system turned out to work very well from the start, both from a purely technical point of view and from an informational point of view; people knew what to do, how to pay, etc. Payment compliance was high, and the number of complaints was much lower than expected. On an average day in May 2006, 371,300 journeys took place over the charge cordon. 19 percent of passages were made with exempted vehicles (buses, taxis etc.), and an additional 9 percent were exempted due to the “Lidingö exception”. 267 500 passages were hence charged, resulting in 115 100 tax decisions (one tax decision was made per day and vehicle) and yielding toll revenues of SEK 3.2 million. Of the 115,100 daily tax decisions, only 100 were investigated by the Swedish Tax Agency and five were appealed. The Swedish Road Administration customer-service unit received on an average day in May 2,200 calls, in contrast to an expected number of 30,000 calls. Based on this, our assessment is that the system and the information generally worked well from a user’s perspective.

5.5 General advice on the technical system

The procurement and the design of the technical system in Stockholm is analysedHamilton(2011)inoffersomeimportantlessons.Someofthegeneralconclusionsarethatthepoliticalriskwhendecidingtoimplementcongestionchargingaffectsthecost


31

of the system, that an agency spending to reduce that risk can be seen as rationallypayinganinsurancepremium,andthatpublicacceptanceisaffectedbothbythecostofthe system and its functionality at launch. In addition, there are several designconsiderations,summarizedbelow.

GetthelegalconditionsclearearlyEarlyinthetechnicaldesignprocess,onemustknowthelegalconditions.Forexample,what is an acceptableproof that a vehiclehaspassed a gantry?Whatpossibilities toappealmustexist?Theanswerstosuchquestionswillhaveimportantrepercussionsonthe technical design, for example whether transponders can be the sole means ofidentificationornot.In Stockholm, a problem occurred that hopefully should be rare: midway in theprocurement process, the legal status of the congestion charge changed from a“municipalenvironmentalcharge”toastatetax(alegalinvestigationconcludedthatitwas illegal for a city to charge moving vehicles on existing roads). This had manyeffects,includingthattheresponsibilityoftheprocurementhadtobechangedfromtheCityofStockholmtothenationalgovernment.Thisincreasedthecostforestablishingthesystemconsiderably.

Choosecost‐efficientserviceleveltargets.Considerwhatthecost‐efficienttargetsofservicelevelsare,givenwhatthegoalsofthesystemareandhowdifferentservicelevelsaffecttheintendedfunctionofthesystem.Goingfrom,say,95%to99%orfrom99%to99.9%onanygivenservicelevelmaybeasignificantcostdriver.InStockholm,the“uptime”ofthesystem(measuredastheshareof“lane‐minutes”thesystemwasactuallyregisteringpassages)wasrequiredtoexceed99.9%.Tomeetthishighrequirement,theprimecontractordesignedasystemwhere(almost) every component was duplicated, spare parts were obtained in largequantities,trainedstaffwasmadeavailabletodoon‐siteservicewithshortnotice,andtechnical IT support was initially on standby 24/7. Obviously, this increasedinvestment and operations costs. Moreover, it should be obvious that lowering theuptime requirement to, say, 95%, would not affect the traffic‐reducing effect of thecharges.Afterall,thetravellersaremakingtheirtraveldecisionsbasedonthefactthattheyarehighlylikelytohavetopayiftheygobycaracrossthechargingcordon.Fromthis perspective, uptime requirements could havebeen relaxed substantiallywithoutlosinganyoftheultimateeffectonthetrafficsituation.Thisillustratestheprincipleofhavingcost‐efficiencyinmindwhenformulatingtechnicalsystemrequirements.

Choosecost‐efficientpaymentchannelsEachpayment transactioncomesat a cost,both in termsof convenience for theuserandasafeefromthefinancialserviceprovider.Hence,allowingforaggregatedmonthlypayments rather than paying each passage individually will reduce operating costs.Cashovercounter(inshops,forexample)mightbenecessaryforuseracceptance,butitisprobablythemostexpensiveformofpayment.

HandlingtranspondersisexpensiveTransponder (or “tag‐and‐beacon”) technology is efficient in many ways, not leastbecause it allows complex charging structures andmakes it easy for the driver. Theproductionofmanytranspondersmaybeasignificantcostdriver,though,butlesswellknownisthatitisoftenamajorcosttoadministratetransponders.Newcarsneednewtransponders, cars change owners, and transponders are lost, stolen, and broken. InNorway,whereover40differentroadtollschemesareinoperation,transpondersareusedinsome,whileothersaremanagedbymannedtollbooths.Andeventhere,wherethecomparisontechnologyishighlymanual,thereisaslightproductivityadvantageforthose not using transponders (Odeck, 2008). With today’s technology, cameras andautomatic number plate recognition (ANPR) can potentially reach a very high


32

identificationratio,whichofferamplecompetitionforanytransponder‐basedsolution.The Stockholm system started out as a transponder‐based system,with ANPR as anadd‐onforlegalreasons,buthasreliedonANPRexclusivelysinceafewyears.

Whendoingafunctionalprocurement,makesuretoaligncostandriskresponsibilitiesInStockholm,thecongestionchargingcallcentrewasinitiallyvastlyoversized,whichwasamajorcostdriver initially.Partof thereason for thiswasthat if thecallcentrewould not meet its service quality targets (e.g. maximal answering times), then theprime contractorwould be financially penalized,while itwas buyer that carried thecostofcallcentrestaff.Hence,riskandcostwerebornebydifferentparties15,andthecontractorhadno incentive to increase itsownriskbycuttingdownonresources. Ifprocuringasystemasafunction,oneshouldmakesurethatthepartycarryingtheriskisalsotheonetakingthecostforriskmitigation,inallareasoftheoperation.

Highpoliticalriskswillweakenthepublicnegotiationposition,andwillincreasecostsbyhavingthecontractorrequireariskpremiumInStockholm,thestakeswerehighforalmostallactorsinvolved.Individualcareersaswell as the prosperity of private firms andpolitical coalitionswas at risk, or at leastperceivedasbeingso.Thisdominatedthecontextinwhichtheprojectwascarriedout,and it was under the influence of this risk environment that decisions were made.Thereweremanyunknownfactorsthatwerethoughttokilltheprojectontheirown.Aboveall,ifthesystemdidnotwork,orwasperceivednottowork,rightfromthestart,itwouldalmostcertainlybeabolishedimmediately.Thisisatleastapartialexplanationof cost drivers such as the oversized call centre, the excessive service levelrequirements etc. It all goes back to the intense political pressure and high politicalstakes: the outcome of the next election would depend on the outcome of the trial,perhapsnotonly in thecitybutalsoonthenational level.Thismeant that thepublicnegotiationpositionwasweak–thesystemhad towork,andithad tobe finishedontime.Obviously,suchasituationcreatesopportunitiesforacontractortochargemoremoney.Forthecontractors,afailure–evenifitwasnotduetomistakesoftheirown–could be potentially disastrous for future business. Thismeans that contractorswillrequireariskpremiumtoevenengageintheworkofconstructingthesystem.Hence,thelessonisthatastablepoliticalenvironmentandampletimetoplanandimplementthe system will keep costs down. Conversely, the extremely tense and uncertainpoliticalsituationinStockholmatthetime,andthelegaluncertaintiesthatrequiredalarge number of changes and redesigns during the development process, allcontributedtoasubstantiallyhigherinvestmentcostthanwouldotherwisehavebeennecessary.Themaincontractorarguesthatasimilarsystemcouldnowbebuiltforhalfthecostorless.

6 COST‐BENEFIT ANALYSIS

Evenifitiswellestablishedthatcongestionpricingwillyieldasocialsurplus,itisnotevidenteitherthatitwillbeenoughtocoverinvestmentandoperationalcosts,orthatareal congestion pricing system,with all its practical and political limitations, will besociallyprofitable.Eliasson(2009)providesacost‐benefitanalysisofthecharges.

6.1 Technology costs

The total cost for the charging systemwas approximately SEK 1.9 billion, includingoperations costsduring the firstyearof operation.1.05billionwas incurredprior to

15 To be fair, it should be pointed out that this misalignment of costs and risks was an exception in the Stockholm procurement.


33

the start of operations. A significant part of this startup costwas costs for extensivetesting. The system would only be operational for 7 months, making it absolutelynecessarythateverythingworkedrightfromthestart.Thestartupcostalsoincluded,in addition to purely technical investments, system development in a wide sense,educating and training staff, testing, public information, etc., and certain otheradditionalminorcosts,suchasthosefortrafficsignals,andtheservicesoftheSwedishEnforcementAgencyandtheSwedishTaxAgency.Therestofthecosts(850millionSEK)wererunningcostsandadditionaldevelopmentcostsduring2006.Far fromall costs incurredduring2006werepure runningcosts:thesystemwasimprovedinseveralwaysduringthespringof2006.AlsoincludedaretheSwedishRoadAdministration’s costs for closingdown the systemandevaluatingthe results during the second half of 2006. The investment and startup costs wereconsiderably increased by the uncertain and heated political situation (as discussedabove).Themaincontractorarguesthatasimilarsystemcouldnowbebuiltforhalfthecostorless.Actual running costs decreased significantly by eachmonth of 2006,when it quicklybecameobviousthatthingsinfactwentbetterthanplanned:thenumberofcomplaintsandlegalactionswereforexampleconsiderablylowerthanwhathadbeenanticipated,reducingcostsforlegalandtaxadministration.Further,thenumberofcallstothecallcenter (the single biggest item in running costs) turned out to be around 1/20 thanwhathadbeenanticipated–around1500callsperdayinsteadof30000perday.Thismeant that the call center was very much oversized, and during the spring, it wasdownsized – a considerable reduction of running costs. Thismeans that investmentscostscouldprobablyhavebeenreducedquitesubstantially if theconditions(andnotleastthetimeconstraints)hadbeendifferent.Thispointmaybeespecially importantto note for other cities considering similar schemes. In 2006, the National RoadAdministration estimated future running costs to around 220 million SEK per year(around25millionEuros).Sincethen,theresponsibilityforthesystemhasbeenmovedto the National Transport Agency system has been developed in various ways, inparticulartoincorporatethecongestionchargingsysteminGothenburgwhichstartedin January 2013, and running costs have decreased further. At the time of writing(February2014),theTransportAgencyestimatesyearlyrunningcoststoaround250MSEKforbothsystems.Thecost‐benefitanalysispresentedhere,however,usestheoldestimateforrunningcosts.Yearlyrevenuesforthefirstyearsofoperationwerealittlemorethan800MSEK.Sinceafewexemptionshavebeenabolished,revenueshaveincreasedsomewhattoaround850MSEKin2013.Comparedtotheforecastbeforethestart,revenuesturnedoutotbearound14%lessthanpredicted.5%ofthiswasduetothetrafficreductionbeinglargerthan predicted. The major part of the revenue shortfall, though, was due to anunderestimationofhowmanyvehiclesthatwouldbeexempted.Thiswaspurelyduetoa lackof data on the traffic composition,which shouldhavebeen avoidable. Instead,neglecting to conduct reasonably simple traffic surveys beforehand meant that theexemptions became more costly than expected, and also created problems for thepredictedcashflows.

6.2 Benefits

Mostofthebenefitsareaccessibilitybenefits,i.e.traveltimesavingsandreducedtraveltime variability. From an economic point of view, it should be noted that it is thesebenefitsthataretranslatedintoeconomicproductivityandgrowth.Forexample,travel


34

time savings are partly converted into more working hours (increasing economicproduction) and partly to better matching on the labour market. These effects fallpartlyoutsidestandardcost‐benefitanalysis,sincetheyusuallydonotaccountfortaxwedges and agglomeration effects (Anderstig, Berglund, Eliasson, Andersson, &Pyddoke,2012).Secondarybenefits includereducedemissionsandaccidents. Inaddition, therevenueitself may generate benefits if they are used to reduce distortionary taxes orinvestmentswithpositivebenefit‐costratios.TheCBA inTable2 (Eliasson, 2009) shows that the Stockholm systemyields a largesocial surplus16, well enough to cover both investment and operational costs. Theannualsocialsurplusisaround650MSEK(afterdeductingoperatingcosts).Allmajoreffectsareprimarilybasedonmeasurements,themostimportantsourcesbeingtraveltimeandtravelflowmeasurements. Loss/gain

Consumer surplus Shorter travel times 536

More reliable travel times 78 Loss for evicted car drivers, gain for new car drivers -74

Paid congestion charges -804 Increased transit crowding -15

Consumer surplus, total -279

Externalities

Reduced greenhouse gas emissions 64 Health and environmental effects 22

Increased traffic safety 125

Externalities, total 211

Government costs and revenues

Paid congestion charges 804 Increased public transit revenues 138

Decreased revenues from fuel taxes -53 Increased public transport capacity -64

Operational costs for charging system (incl. reinvestment and maintenance) -220

Government costs and revenues, total 606

Tax effects etc.

Marginal cost of public funds 182 Correction for indirect taxes -65

Net social benefit, excl. investment costs 654 Table2.Costsandbenefitsofthecharges,MSEKperyear.

16 The value of travel time was assumed to be 122 SEK/h per vehicle (65 SEK/h per person for private trips, 1.26 persons per private car and 190 SEK/h per person for business trips and distribution traffic). The value of reliability was assumed to be 98 SEK per hour of standard deviation (i.e. 80% of the value of travel time). CO2 emissions were valued at 1.50 SEK/kg. The Swedish value of a statistical life is 17.5 MSEK, the value of a severe injury 3.1 MSEK and a light injury 0.18 MSEK. The standard Swedish estimates of marginal cost of public funds and correction for indirect taxes are 1.3 and 1.23 respectively. The CBA is described in detail in Eliasson (2006, 2008).


35

The total start‐up cost of the system was 1900 MSEK, including informationcampaigns, extensive system tests and so on. Together with marginal cost of publicfundsandcorrection for indirect taxes, thisgivesa total social start‐up costof2900MSEK.Hence,thestart‐upcostis“recouped”intermsofsocialbenefitsinabout4years.The estimated yearly operational cost of the system (220 MSEK) includes not onlyrunningcostsbutalsonecessaryreinvestmentsandmaintenancesuchasreplacementofcamerasandotherhardware.Consumer surplus is negative, as expected, but the value of the time gains is highcomparedtothepaidcharges–timegainsamounttoalmost70%ofthepaidcharges,whichisveryhighcomparedtomosttheoreticalormodel‐basedstudies.Thisismainlydueto”networkeffects”,i.e.significantamountsoftrafficthatdonotcrossthecordonandhencedonotpayanychargebutstillgainfromthecongestionreduction.

7 EQUITY EFFECTS

Oneofthemajorcriticismsofcongestionpricingisthatitisregressive,sincethosewiththeleastincomehavethehardesttimeaffordingtopaythetoll.Atthesametime,thereis a strongcounter‐argument tobemade that congestionpricing canbe regressive ifthe drivers are, by and large, thosewith the highest incomes, and thosewith lowerincomes take transit. An advantage of the approachhere is thatwe can examine theaverageeffecteitheronincomecategoriesasawhole,oronincomecategoriesamongdrivers,transitriders,oreventhosewhoswitch.Thereareanumberofstudiesontheequity effects of the Stockholm charges (Eliasson & Levander,, 2006; Eliasson &Mattsson, 2006; Franklin et al., 2010; Karlström & Franklin, 2009), yielding similarconclusions. The analysis and results presented here is taken from Franklin et al.(2010).The average effects for income and toll effect sub‐groups are shown in Table 3.Asterisksmarktheincomecategorieswhoseeffectsaresignificantlydifferentfromthepooled income categories. The most notable result here is actually the absence ofsignificant differences. The only significant trend here is among the un‐tolled,wherethe income categories are significantly different from the pooled income categories.Thereisalsoatrendof increasingbenefitwithincome,butanimportantnotehereisthatallfiveincomecategoriesdidindeedseeanaveragebenefit.Inotherwords,fortheonlysubgroupwhereincomeplayedasignificantrole,everyonewasbetteroffanyway.Table3.AverageWelfareEffectsbyRefundScenario

Group

AverageWelfareChange,SEK/year/person

NoRefunds

RefundScenarioLumpSum TaxReduction

All –78 +180 +173ByIncomeCategory:<25000 –50* +244* +12425‐40000 –138 +110 +4540‐55000 –39 +224 +19255‐70000 –12 +234 +273>70000 –152 +84 +322

Note: * = significantly different (>95%) from the average welfare change forpooledgroups.


36

Also, although there are not significant differences between income levels, there aresomedifferences in theaveragenumbers, but thesedonot followa clear trend fromlow tohigh incomes.The reason for this is likely thatother factors, suchasdistancetravelledandamountoftollpaid(basedontimeofday),andnumberoftimescrossingthetollcordon,variedmorewidelywithineachincomecategoriesthanbetweenthem.Itappearsthat individualcircumstanceshaveagreatereffectonwelfarethanincomelevel.Theuseofrevenuescanbeextremelyimportanttotheequityeffectsofatollsystem,asevidencedbypriorstudies(Eliasson&Mattsson,2006).Therefore,whilewecannotbedefinitiveabouttheredistributiveeffectsoftollrevenuedisbursements,wecanatleastidentifysomebounds,bytestingtwotheoreticalscenariosforrefundingtollrevenues.Thefirstisdeliberatelyprogressive,alumpsumrefundtoallinourstudypopulation;thesecond,deliberatelyregressive,withanacross‐the‐boardreduction in the incometaxrate.Adding the effects of each of these refund schemes to the totalwelfare effects foundabove,wearriveattheaveragewelfareeffectsshowninTable3.Importantly,alloftheeffectsarenowpositive,andthisholdstrueforbothalumpsumandataxreduction.Thus,weseeanaffirmationofoneofthecoreargumentsforcongestionpricing:thatbyreducing externalities, we can see a net positive effect. Moreover, the empiricalevidence supports a theoretical result from Small (1983) suggesting that all incomelevelscouldseeabenefitonaverage,aslongasrevenueswereappropriatelyreturned.Incomparingthetworefundscenarios,weseetheexpectedresultthatthelumpsumscenario is progressive, with the lowest income categories receiving the greatestbenefit, while the tax reduction scenario is regressive, with the highest incomecategories gaining themost. Certainly, the refunds themselves for the two scenariosshouldbeprogressiveandregressive,respectively,butwhatthistellsus is thattheseoriginaltendenciesarenotoverwhelmedbythepatternofcostsduetothecongestioncharging system itself, as representedby the "NoRefund" scenario.Treating the tworefundscenariosasbounds,theconclusionwecanreachisthatawiderangeofusesforthe toll revenues could maintain a positive average effect for all income categories,evenifsomeuseswouldbemoreprogressivethanothers.

7.1 Equity, fairness and winners/losers

Theequityanalysisaboveconcentratesonashort‐runwinner/loserperspective,justasmost equity analyses do. This perspective lacks two considerations, however: therelationship between objective equity and the perception of fairness, and therelationshipbetweenobjective(monetary)effectsandsupportforthecharges,whichisnotasimplelinearrelationship.First,aproblemwiththe“winners/losers”perspectiveconcernsthewaythistranslatesto thequestionof “fairness”.Often, if asystemaffectshigh‐incomegroupsmore thanlow‐incomegroups,itisclaimedtobea“fair”system.Hence,“fairness”considerations–whichareknowntoaffectacceptability–areinterpretedasaquestionofidentifying“winners and losers”. In Stockholm, the equity effects were generally speakingprogressive:high‐incomegroupspaidmore than low‐incomegroups,menpaidmorethanwomen,employedmorethanunemployedetc.Butoncethechargesareinplace,andtheshort‐termwinner/loserperspectivefades,anotherperspectivebecomesmoreimportant:whatpriceisactually“fair”tochargeforacartrip?Fromthisperspective,itis “fair” that one pays more to drive on a congested road or to cause emissions indensely populated areas – irrespective of income or place of residence, or what a


37

hypothetical travel patternwould have beenwithout the charges. Thismeans that asystemneedstobeperceivedas“fair” inthissense: itneedstobeconsistentwith itsstated objective. In Stockholm, one of the most common objections to the systemnowadaysisthattrafficwithinthecordonisnotcharged.Althoughtherearetwogoodanswers to this (the congestion is mainly located on the arterials along the cordon;mostofthetrafficinsidethecordoncrossesthecordonatsomepointonthetrip),thisshowshowthedebatehasmovedfrom“whowins/loses”to“what’sfairrelativetotheobjectives”.Second, according to standard transport‐economic theory underlying classical equityanalysis,mostmotoristswouldnot think that the time savedwasworth the chargestheyhadtopay.Theoretically,theincomefromthechargesissufficienttocompensatethe losers, so the standard recommendation in the acceptance literature is thatcongestionchargesmustbepartofa“package”,withinwhichitisclearhowtheincomeisgoing tobespent for theadvantageof thegeneralpublic, if it is going tohaveanychance of being accepted. But the standard theory neglects three importantconsiderations.First:networkeffects.Sincequeuespropagate“upstream”,eventhosenot going through theactualbottleneckwill suffer fromqueues.Pricing traffic in thebottlenecktoreducequeues,allupstreamtrafficwillbenefit–notonlydriversactuallypaying the charge. Second: the effect on the urban environment. Typically, standardanalysisofcongestionchargestakesnoaccountofeffectsforpedestriansorcyclists,ortheeffectontheperceivedurbanenvironment.Third:theself‐selectioneffectontripsandon the value of time. Congestion chargeswill tend to “sort” trips such that tripswithhighvaluewillstayontheroad(andenjoytimebenefits),whilelow‐valuedoneswill be priced off. Not taking this phenomenon into account will underestimate thevalueofthetimebenefits.Fromanacceptanceperspective,theimportantpointisthatindividuals can belong to different valuation “groups” on different days, or differentjourneys.Another potentially important observation is that car drivers apparently changedbehaviour without even noticing it. When motorists were asked if the congestioncharginghadmadethemchangetheirtravellinghabits,thereweretoofewanswering“yes”tocorrespondwiththeactualreductioninmeasuredtrafficvolumes.Oneimportantreasonisthattravelpatternsaremuchmorevariablethanmostpeopleare aware of. It is simply not the same drivers paying the charge each day. On thecontrary:alargemajorityofthedriversare“occasional”drivers,whopaythechargeaoneora fewtimesaweekorevenmoreseldom.Figure11showsthedistributionofpassagefrequencygroupsduringatwoweekperiod.


38

Figure11.Driverspassing thecordonagivenday,groupedbyhowoften thepasscordonduringatwo‐weekperiod.

Thismeansthatmostcarownersareaffectedsomewhat,butveryfewareaffectedalot.Figure12illustratesthis.Anygivenday,around5%ofprivatetrips inthecountyareaffectedbythecharges.Overtwoweeks17,however,43%ofprivatecarsinthecountywill pay at least once. However, over this two week period, only around 2% of carownerswillpayanamountequivalenttotheaveragereturnfareeachday(270SEKfor2weeks),andonlyabout0.4%willpayanamountequivalenttorush‐hourreturnfareeachday(400SEKper2weeks).

Figure12.Shareofprivatecarswhopayvarioustotalchargeamountsduringatwo‐weekperiod.

8 ARE THE RESULTS TRANSFERABLE?

In many respects, each city is unique. Introducing a charging scheme similar to theStockholm system is by no means a guarantee for achieving the same congestionreductionsasinStockholm.Butinamorefundamentalway, itcanbearguedthattheexperiences are indeed transferable to other cities, in the sense that the chargesactuallyaffectedcardrivers in theway thathadbeenpredictedby transportmodels.

17 Due to integrity legislation, statistics can only be aggregated over individuals over two-week periods, not longer.


39

During the design process, three independent transportmodels yielded very similarresults regarding the anticipated effect of the charges. Thiswas despite the fact thethreemodelswereestimatedandcalibratedondifferentdatasetsanduseddifferentmethodologies (although all of them were nested logit models linked to a staticequilibrium model). Despite this, the forecasts were deemed to be unrealistic – thechargescouldnotpossiblyaffecttrafficthatmuch, itwasbelieved.Thelessonhereistwofold:agoodtransportmodelisaninvaluabletoolfordesigninganefficientchargingscheme, and the result is likely to be correct (after accounting for known modellimitations such as underestimating travel times in severe congestion and manymodels’inabilitytoaccountforchangesindeparturetimes).The“transferablelesson”ishencethatawell‐designedchargingschemethatseemstoworkinatrafficmodelisactuallyalsolikelytoworkinreality.Averycommonreactionfromothercitiesisthat“chargesmayhaveworkedinStockholmandLondon,buttheparticularsituationinourcity means that charges won’t work here, despite our transport model sayingotherwise”.Thesameclaimused tobecommon inStockholm, in fact.TheStockholmexperienceisthattransportmodelscanactuallybetrusted.ThesuccessoftheStockholmchargesinreducingcongestionandachievingpublicandpolitical support has attracted great interest from cities around theworld. A naturalquestioniswhetherthepositiveresultsaretransferable–ifcongestionchargeswouldwork just aswell in other cities. Judging from the authors’ experience as advisors tocitiesaroundtheworld,acommonreactionis“itwouldnotworkinourcity”.Ofcourse,all cities have their particular characteristics and local conditions, so a copy of theStockholmsystemwouldnotgiveexactlythesameeffectinanothercity.But in a more fundamental way, it can be argued that the experiences are indeedtransferabletoothercities,inthesensethatthechargesactuallyaffectedcardriversinthewaythathadbeenpredictedbytransportmodels.Theconclusionthatatransportmodelwas able to predict demand responseswith good‐enough accuracy leads to amore qualified answer to the question of transferability: if a congestion chargingsystemispredictedto“work”inagivencity–thatis,reducepeaktrafficinbottleneckswithoutunacceptably adverse side‐effects orhaving touse unacceptablyhigh chargelevels–thenthatislikelytobetrueinrealityaswell,notjustinthemodel.Itshouldbenoted,however,thatthebeneficialeffectsoncongestionandtraveltimesarelikelytobe underestimated by static network models. During the Stockholm design process,three independent transport models yielded very similar results regarding theanticipated effect of the charges. Despite this, the forecasts were deemed to beunrealistic–thechargescouldnotpossiblyaffecttrafficthatmuch,itwasbelieved.A related question is whether the effectiveness of congestion charges are highlydependentonthespecificfeaturesofthelanduseandtransportsystem,implyingthatthegoodexperiencesofthecongestionchargingsystemsimplementedinStockholmorLondon would not be transferable to cities with topological and demographicalconditions,availabilityandattractivenessofnon‐chargedroutesandpublic transportprovisionorsizes.ThisquestionisexploredindetailinastudybyBörjesson,Brundell‐FreijandEliasson(2014). The main conclusion is that although the social benefit of a given chargingsystem(includingthesizeofthecharge)isconsiderablyandnon‐linearlydependentoninitial congestion levels, traffic effects and adaptations are surprisingly stable acrossdifferenttransportsystems.Specifically,thelevelofpublictransportprovisionhasonlysmalleffectsonbaselinecongestion,andthereforeonthetotalbenefitofthecharges.Interestingly, adaptation cost, traffic reduction across the cordon and the share ofdriverspricedofftheroaddivertingtopublictransportarealsosurprisinglyinsensitive


40

tothelevelofpublictransportprovision,contrarytothecommonargumentthatpublictransport provision is crucial for effectiveness and efficiency. Drivers can adapt inmany different ways except for switching to public transport, which explains therobustnessoftheresults.

8.1 Why a “success”?

Thenewspaperquoteintheintroductioncallingthetriala“success”wasfairlytypical.Even if certainly not everybodywas in favour of the charges, the change in generalopinion as reflected in polls andmedia seemed to justify such statements. Butwhatwere thekey factorsbehind this?There isnoconclusiveanswer to thisquestion,butfive main reasons were often mentioned by four key people involved in the trial –Gunnar Söderholm, head of the Congestion Charging Office responsible for (amongotherthings)evaluationandinformation;BirgerHöök,headoftheCongestionChargingUnitattheRoadAdministration,responsibleforthetechnicalsystemandinformationregardingpaymentandtechnology;GunnarJohansson,headoftheIBM‐ledconsortiumthat developed and operated the technical system; and the author, who wasresponsiblefortheearlysystemdesignandchairmanoftheexpertpanelsummarisingandscrutinisingthelargeevaluationpackage.

- Thetechnicalsystemworked.Thatthesystemworkedfromthestartwasofcourseakeyfactor.Thenumberofmisidentificationswasextremelylow,andfromusers’perspectives,everythingworkedseamlessly.Further,theRoadAdministrationmadegreateffortstodevelopacustomer‐friendlysystem.

- Theinformationcampaignhadworked.Apparently,peopleknewwhattodo.Anticipatedproblemswithpeoplewhodidnotknowthattheyshouldpay,ordidnotknowhowtopay,didnotmaterialise.Moreover,theanticipatedproblemofprotestsintheformoflargenumberofcourtappealsorrefusalstopaywasneveraproblem.inspiteofalotoftalkbeforethetrialabout“civildisobedience”intheformofrefusaltopayorappealingtocourt.

- Visiblecongestionreduction.Theimprovementsintraveltimesandtheurbanenvironmentwerevisiblerightfromthestart.Theastonishmentofseeingalmostemptystreetsduringrushhours,inparticularduringthefirstmonths,cannotbestressedenough.Afterthat,thepotencyofroadpricinghadbeenoverwhelminglyproved,andthenegativeargumentsshiftedfrom“itwon’twork”toother,oftenmoreconstructive,arguments.

- Extensiveandscientificevaluation.Evenifeffectswerevisible,oneshouldstresstheimportanceofbeingabletosupplymediawithhardfiguresaboutthereductionoftrafficvolumesandcongestion.Especiallywhenthedebaterecoveredsomewhatfromtheinitialshockcausedbytheenormousinitialeffects,itwasextremelyimportanttohaveprofessional,independentresearchersandexperts,comingfromdifferentbackgroundsandorganisations,beingabletoexplainandevaluatewhatwashappening.Thesizeoftheevaluationwasitselfanimportantfactor:somanyexpertsandresearcherswereinvolvedinonewayoranotherthatitwasimpossibletowaveitaway.

- Clearobjectives.Thesystemhadclearandmeasurableobjectives–reducingcongestionandimprovingtheenvironmentintheinnercity–andthesystemwasvisiblydesignedwiththeseobjectivesinmind.Moreover,theobjectiveswerefulfilled.

9 REFERENCES

Anderstig,C.,Berglund,S.,Eliasson,J.,Andersson,M.,&Pyddoke,R.(2012).Congestionchargesandlabourmarketimperfections:“Widereconomicbenefits”or“losses”?(


41

No. 2012:4). CTS Working Paper. Centre for Transport Studies, KTH RoyalInstituteofTechnology.

Bartley, B. (1995). Mobility Impacts, Reactions and Opinions ‐ Traffic demandmanagement options in Europe: The MIRO Project. Traffic Engineering andControl,36(11),596.

Bonsall,P.,Shires, J.,Maule, J.,Matthews,B.,&Beale, J. (2007).Responses tocomplexpricing signals: Theory, evidence and implications for road pricing.TransportationResearchPartA:PolicyandPractice,41(7),672–683.

Börjesson, M., Brundell Freij, K., & Eliasson, J. (2014). “Not Invented Here”: On thetransferability of congestion pricing benefits (CTSWorking Paper). Centre forTransportStudies,KTHRoyalInstituteofTechnology.

Börjesson,M.,Eliasson,J.,Hugosson,M.B.,&Brundell‐Freij,K.(2012).TheStockholmcongestion charges—5 years on. Effects, acceptability and lessons learnt.TransportPolicy,20,1–12.

Daunfeldt, S.‐O., Rudholm, N., & Rämme, U. (2009). Congestion charges and retailrevenues: Results from the Stockholm road pricing trial. TransportationResearchPartA:PolicyandPractice,43(3),306–309.

Eliasson, J. (2008). Lessons from the Stockholm congestion charging trial. TransportPolicy,TransportPolicy,15(6),395–404.

Eliasson,J.(2009).Acost‐benefitanalysisoftheStockholmcongestionchargingsystem.TransportationResearchPartA:PolicyandPractice,43(4),468–480.

Eliasson, J. (2010).Soyou’reconsidering introducingcongestioncharging?Here’swhatyou need to know: An FAQ based on Stockholm’s experiences ( No. 2010‐04).InternationalTransportForumDiscussionPapers.

Eliasson,J.(2014).TheStockholmcongestionpricingsyndrome:howcongestionchargeswent from unthinkable to uncontroversial ( No. 2014:1). CTSWorking Paper.CentreforTransportStudies,KTHRoyalInstituteofTechnology.

Eliasson, J.,Börjesson,M.,BrundellFreij,K.,Engelson,L.,&VanAmelsfort,D. (2013).Accuracyofcongestionpricingforecasts.TransportationResearchA,52,34–46.

Eliasson, J., Hultkrantz, L., Nerhagen, L., & Rosqvist, L. S. (2009). The Stockholmcongestion ‐ charging trial2006:Overviewofeffects.TransportationResearchPartA:PolicyandPractice,43(3),240–250.

Eliasson,J.,&Jonsson,L.(2011).Theunexpected“yes”:Explanatoryfactorsbehindthepositiveattitudes tocongestioncharges inStockholm.TransportPolicy,18(4),636–647.

Eliasson,J.,&Levander,,A.(2006).EquityEffectsoftheStockholmTrial.Transekreport.Eliasson,J.,&Mattsson,L.‐G.(2006).Equityeffectsofcongestionpricing:Quantitative

methodologyandacasestudy forStockholm.TransportationResearchPartA:PolicyandPractice,40(7),602–620.

Forsberg, B., Burman, L., & Johansson, C. (2006). Stockholmsförsöket harfolkhälsopotential.Läkartidningen,50,4043–5.

Franklin, J.,Eliasson, J.,&Karlström,A. (2010).TravellerResponses to theStockholmCongestionPricingTrial:WhoChanged,WhereDidTheyGo, andWhatDid ItCost Them? In Saleh & Sammer (Eds.), DemandManagement and Road UserPricing:Success,FailureandFeasibility.AshgatePublications.

Frey,B.S.(2003).WhyAreEfficientTransportPolicyInstrumentssoSeldomUsed?InJ.Schade&B.Schlag(Eds.),AcceptabilityofTransportPricingStrategies.Elsevier,Oxford.

Goodwin, P. (2006). The gestationprocess for roadpricing schemes.LocalTransportToday,444.

Goodwin, P., Dargay, J., & Hanly, M. (2004). Elasticities of Road Traffic and FuelConsumptionwithRespect toPriceandIncome:AReview.TransportReviews,24(3),275–292.


42

Hamilton,C.J.(2011).RevisitingthecostoftheStockholmcongestionchargingsystem.TransportPolicy,18(6),836–847.

Hamilton,C.J.,&Eliasson,J.(2012).Decisivefactorsfortheacceptabilityofcongestionpricing. In C. J. Hamilton (Ed.), Implementing Road Pricing: Standards,Institutions, Costs, and Public Acceptance, Doctoral dissertation. Centre forTransportStudies,KTHRoyalInstituteofTechnology.

Hårsman, B., Pädam, S., & Wijkmark, B. (2000). Ways and means to increase theacceptance of urban road pricing (Final report from the PRIMA project).EuropeanCommission.

Hårsman, B., & Quigley, J. M. (2010). Political and public acceptability of congestionpricing: Ideologyand self‐interest. JournalofPolicyAnalysisandManagement,29(4),854–874.

Henriksson,G.(2009).WhatdidtheStockholmTrialmeanforStockholmers?Gullbergand Isaksson (eds.): Congestion taxes in city traffic. Lessons learnt from theStockholmTrial.NordicAcademicPress.

Jones, P. (2003). Acceptability of Road User Charging: Meeting the Challenge. In J.Schade&B.Schlag(Eds.),AcceptabilityofTransportPricingStrategies.Elsevier,Oxford.

Karlström, A., & Franklin, J. P. (2009). Behavioral adjustments and equity effects ofcongestionpricing:AnalysisofmorningcommutesduringtheStockholmTrial.TransportationResearchPartA:PolicyandPractice,43(3),283–296.

Kottenhoff, K.,&Brundell Freij, K. (2009). The role of public transport for feasibilityandacceptabilityofcongestioncharging:ThecaseofStockholm.TransportationResearchPartA:PolicyandPractice,43(3),297–305.

Odeck,J.(2008).Howefficientandproductiveareroadtollcompanies?:EvidencefromNorway.TransportPolicy,15(4),232–241.

Rienstra, S. A., Rietveld, P., & Verhoef, E. T. (1999). The social support for policymeasures in passenger transport.: A statistical analysis for the Netherlands.TransportationResearchPartD:TransportandEnvironment,4(3),181–200.

Schade, J., & Baum, M. (2007). Reactance or acceptance? Reactions towards theintroduction of road pricing. Transportation Research Part A: Policy andPractice,41(1),41–48.

Schade, J., & Schlag, B. (2003a). Acceptability of road user charging: Meeting thechallenge. In J. Schade & B. Schlag (Eds.), Acceptability of Transport PricingStrategies.Pergamon.

Schade, J., & Schlag, B. (2003b). Acceptability of urban transport pricing strategies.TransportationResearchPartF:TrafficPsychologyandBehaviour,6(1),45–61.

Schade,J.,Schlag,B.,Giannouli,I.,&Beijer,A.(1999).Acceptabilityofmarginalcostroadpricing(AFFORDDeliverable2).EuropeanCommission.

Winslott‐Hiselius, L., Brundell‐Freij, K., Vagland, Å., & Byström, C. (2009). Thedevelopment of public attitudes towards the Stockholm congestion trial.TransportationResearchPartA:PolicyandPractice,43(3),269–282.

Documents

The Stockholm congestion charges: an overview · The Stockholm congestion charges: an overview Jonas Eliasson, KTH Royal Institute of Technology CTS Working Paper 2014:7 Abstract