/
FootnotesWethankAnmingZhangandKenSmallforhelpfulcomments.Anyshortcomi FootnotesWethankAnmingZhangandKenSmallforhelpfulcomments.Anyshortcomi

FootnotesWethankAnmingZhangandKenSmallforhelpfulcomments.Anyshortcomi - PDF document

briana-ranney
briana-ranney . @briana-ranney
Follow
422 views
Uploaded On 2016-06-24

FootnotesWethankAnmingZhangandKenSmallforhelpfulcomments.Anyshortcomi - PPT Presentation

HahnRW1984MarketpowerandtransferablepropertyrightsQuarterlyJournalofEconomics99753765JohnsonTSavageI2006DeparturedelaysthepricingofcongestionandexpansionproposalsatChicagosOHareair ID: 375856

Hahn R.W. 1984.Marketpowerandtransferablepropertyrights.QuarterlyJournalofEconomics99 753-765.Johnson Savage 2006.Departuredelays thepricingofcongestion andexpansionproposalsatChicago'sO'Hareair

Share:

Link:

Embed:

Download Presentation from below link

Download Pdf The PPT/PDF document "FootnotesWethankAnmingZhangandKenSmallf..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.


Presentation Transcript

FootnotesWethankAnmingZhangandKenSmallforhelpfulcomments.Anyshortcomingsinthepaper,however,areourresponsibility.1SeeBrueckner(2008),BruecknerandVanDender(2008),Verhoef(2008),DanielandPahwa(2000),Daniel(2001),DanielandHarback(2008a,b,c),ZhangandZhang(2006),BassoandZhang(2007),JohnsonandSavage(2006),andMorrisonandWinston(2007).2Arelatedconcernaboutmanipulationarisesinmarketsforpollutionpermits,wherelargepolluterscouldmanipulatethepermitpricetotheiradvantage.SeeHahn(1984)foranearlystudyinvestigatingthisissue.Notethatasimilarphenomenoncouldemergeinthemarketforairportslots,whichrepresentanalternateinstrument(analogoustopollutionrights)fordealingwithairportcongestion(seeBrueckner(2008)andVerhoef(2008)foranalysesofslotmarkets).3Notethattheadjective\classical"isusedtodenoteaparametrictollthatissetunderthecorrectpredictionoftheoptimum.Thetermisthereforenotusedtodistinguishtollswithandwithoutpartialinternalizationofcongestion.4ItshouldbenotedthataclassicalPigouviantollisnotsubjecttothekindofmanipulationdescribedinthissection.SinceaclassicaltollcomesfromevaluatingthePigouvianruleatthesocialoptimum,theonlywaycarrierscoulda ectthetolllevelisbysomehowchangingtheregulator'sperceptionoftheoptimum.Suchbehaviorisconceivable,forexample,inasettingwheretheregulatorreliesonthecarriersforinformationonthecostparameter,usedincomputingf.Carrierscouldmisrepresenttheircosts,thusin\ruencingtheregulator'scomputedfandultimatelythelevelofthetolltheypay.Explorationofsuchanalternativesetting,however,isbeyondthescopeofthepaper.5Notethattheregulator'sexpectationsarealsocon rmedintheclassicalPigouviancase.Theregulatorusesthesocially\rightvolumefincomputingthetoll,andthis\rightvolumeisindeedchosenbythecarriers.6In(25), equalsleft-handsideof(24)evaluatedatf1andf2,whileequalstheleft-handsideof(23)evaluatedattheoptimumminus (f1+f2).7Ifthe rmsproducethesamegood,thecommonmarketdemandisassumedtobeperfectlyelastic.21 Hahn,R.W.,1984.Marketpowerandtransferablepropertyrights.QuarterlyJournalofEconomics99,753-765.Johnson,T.,Savage,I.,2006.Departuredelays,thepricingofcongestion,andexpansionproposalsatChicago'sO'Hareairport.JournalofAirTransportManagement12,182-190.Mohring,H.,Harwitz,M.,1962.HighwayBene ts:AnAnalyticalFramework.North-westernUniversityPress,Evanston,IL.Morrison,S.A.,Winston,C.,2007.Anotherlookatairportcongestionpricing.AmericanEconomicReview97,1970-1977.Pels,E.,Verhoef,E.,2004.Theeconomicsofairportcongestionpricing.JournalofUrbanEconomics55,257-277.Small,K.A.,Verhoef,E.T.,2007.TheEconomicsofUrbanTransportation.Routledge,London.Verhoef,E.T.,2008.Congestionpricing,slotsalesandslottradinginaviation.DiscussionpaperTI2008-030/3,TinbergenInstitute,Amsterdam-Rotterdam.Zhang,A.,Zhang,Y.,2006.Airportcapacityandcongestionwhencarriershavemarketpower.JournalofUrbanEconomics60,229-247.20 ReferencesBasso,L.J.,Zhang,A.,2007.Congestiblefacilityrivalryinverticalstructures,JournalofUrbanEconomics61,218-237.Brueckner,J.K.,2002.Airportcongestionwhencarriershavemarketpower,AmericanEconomicReview92,1357-1375.Brueckner,J.K.,2005.Internalizationofairportcongestion:Anetworkanalysis.Interna-tionalJournalofIndustrialOrganization23,599-614.Brueckner,J.K.,2008.Pricevs.quantity-basedapproachestoairportcongestionmanage-ment.Unpublishedpaper,UCIrvine.Brueckner,J.K.,VanDender,K.,2008.Atomisticcongestiontollsatconcentratedairports?Seekingauni edviewintheinternalizationdebate.JournalofUrbanEconomics64,288-295.Buchanan,J.M.,1969.Externaldiseconomies,externaltaxes,andmarketstructure.Amer-icanEconomicReview59,174-177.Daniel,J.I.,Pahwa,M.,2000.Comparisonofthreeempiricalmodelsofairportcongestionpricing.JournalofUrbanEconomics47,1-38.Daniel,J.I.,2001.Distributionalconsequencesofairportcongestionpricing.JournalofUrbanEconomics50,230-258.Daniel,J.I.,1995.Congestionpricingandcapacityoflargehubairports:Abottleneckmodelwithstochasticqueues,Econometrica63,327-370.Daniel,J.I.,Harback,K.T.,2008a.(When)Dohubairlinesinternalizetheirself-imposedcongestiondelays?JournalofUrbanEconomics,63,583-612.Daniel,J.I.,Harback,K.T.,2008b.Pricingthemajorhubairports.JournalofUrbanEconomics,forthcoming.Daniel,J.I.,Harback,K.T.,2008c.PricingCanadianairports.Unpublishedpaper,UniversityofDelaware.Forbes,S.J.,2008.Thee ectofairtracdelaysonairlineprices.InternationalJournalofIndustrialOrganization26,1218-1232.19 willnolongerguidethemarkettothesocialoptimum.Toaddressthisproblem,thepresentpaperhasderivedalternative,manipulabletollrules,whicharedesignedtoachievethesocialoptimumwhenagentsanticipatethefullimpactoftheiractionsontollliabilities.Theanalysisshowsthat,althoughthemarginaltollsdonotdi erbetweentheconventionalPigouvianandmanipulablecases,theaverageandtotaltollliabilitiesgenerallywillbedi erent.Inaddition,revenuesfrommanipulabletollingarelowerthanunderparametricPigouviantolling,furtherreducingthescopeforself- nancingofcapacity.Theapproachusedinthepapercanbeex-tendedtoderivemanipulablePigouviantaxesinothercontextswherelargeagentsgenerateexternalitiesandunderstandtheimpactoftheiractionsoncorrectivetaxliabilities.18 two rmsgeneratepollutionasaby-productofproduction,withtheexternalitydamageafunctionofthesumofthesetwolevels(thisdamageissu eredbyotheragents,notthe rmsthemselves).A rm'spollutionleveldependsonitsoutputalongwiththelevelofacostlyabatementactivity,andthedemandforeach rm'soutputisassumedtobeperfectlyelastic.7Inthiscase,thesocialoptimumrequiresequalitybetweenthemarginalpro tfromanextraunitofoutputandthemarginalpollutiondamagefromthatunit(equaltotheadditionalpollutiontimesmarginalpollutiondamage).Inaddition,themarginalbene tfromabatement(equaltoabatementproductivitytimesmarginalpollutiondamage)mustequalthemarginalabatementcost.Together,thesetwoconditionsdeterminethe rms'optimaloutputandabatementlevels.Providedthatthe rmsviewanypollutiontaxasparametric,theoptimumcanbegeneratedbyaPigouviantaxequalsetequaltomarginalpollutiondamage,evaluatedatthesociallyoptimalpollutionlevelundertheclassicalapproachorattheexpectedequilibriumlevelunderthealternateapproach.When rmsmanipulatethePigouviantax,theoptimumfailstoemerge,andamanipulabletaxisthenrequired.Theformofthistaxcanbepredictedgiventhepreviousanalysis.Inparticular,each rmshouldpayataxliabilityequaltototalpollutiondamage,adjustedbysomeconstant.Whensuchatax,whosestructureisunderstoodbythe rms,issubtractedfrompro t,optimaloutputandabatementlevelsarethenchosen.Itshouldnotedthat,sincepollutiondamageissu eredbyagentsotherthanthe rmsthemselves,themanipulabletaxstructuredoesnotshareakeyfeatureofthepreviouscongestiontolls:chargingfordamagetotheother rm.Insteadeach rmischargedfordamagedoneto(outside)pollutionvictims,adjustedbyconstant.Amanipulabletaxwouldhaveasimilarforminmanyothercontexts.6.ConclusionTherecentliteratureoncongestionpricingwithlargeagentscontainsaremarkablein-consistency:thoughagentsarelargeenoughtorecognizeself-imposedcongestionandexertmarketpoweroverprices,theydonottakeintoaccounttheimpactoftheirownactionsonthemagnitudeofcongestiontolls.Whenlargeagentsareconfrontedwithtollsderivedunderthisparametricassumptionbutunderstandtheruleusedtogeneratethem,thetollsystem17 levelandderivativeofthedesiredaveragetollfunctiont().Theseequationscanbesolvedfortheseunknownquantities,yieldingt(f1+f2)=f12f1f2c01(f1;f2)f22f1f2c02(f1;f2)(23)t0(f1+f2)=f1f1f2c01(f1;f2)+f2f1f2c02(f1;f2)+D0(f1;f2):(24)Althougharestrictedtollfunctioncouldbederivedinthebasicmodelby ndingthesolutiontoadi erentialequation,asimilaroutcomeisnotfeasiblehere.Toseethediculty,observethattheleft-handof(24)isnotequaltothederivativeofleft-handsideof(23),whichmeansthatafunctiont()satisfyingthesetwoconditionsforallvaluesoff1andf2doesnotexist.Whilethisconclusionmeansthatarestrictedmanipulabletollrulelikethatinsection3isinfeasible,itispossibletogeneratea\fake"manipulablerule,onethatiscomputedusingknowledgeofthesocialoptimum.Inthissense,thetollresemblestheclassicalPigouviantollin(19),whosecomputationrequiresalltheinformationneededtoderivethesociallyoptimalvaluesf1andf2.Inparticular,itcanbeveri edthatthelinearaveragetollfunctiongivenbyet(f1+f2)=+ (f1+f2)(25)yieldsamanipulabletollrulethatgeneratesthesocialoptimum,wheretheslopeandintercepttermsand arederivedusing(23)and(24)andthemselvesdependonf1andf2.6Atollrulewithacommonaveragetollper\rightmaybeappealingonequitygrounds,butdespitethenegativeconclusionjustderived,equitycouldbeachievedinadi erentfashion.Speci cally,sincetheairportauthoritycanadjusttollliabilitiesunderthegeneralmanipulabletollbyvaryingtheKitermsin(21),anydistributionalgoalcanbemeteventhoughthedesiredtypeofrestrictedtollfunctionisnotavailable.5.ManipulableTaxesinOtherContextsTheprinciplesoftheprecedinganalysiscanbeappliedinothercontextswheremanip-ulationofPigouviantaxesmayoccur.Consider,forexample,thepollutioncontext,where16 ingservicebetweenacommonsetofnuncongestedendpointsandthesamecongestedairport(whichmightbeahub).Sincetheyinvolvedi erentendpointcities,suchmarketsmayhavedi erentdemandfunctions,leadingtomarket-powerdistortionsofdi erentmagnitudes.Thesedistortionscanthenbeaddressedbymarket-levelsubsidies,whilecongestionisaddressedbyseparateairport-leveltolls,eitherofPigouvianormanipulableform.Withtwoasymmetriccarriers,therequiredsubsidieswouldbebothcarrierandmarket-speci c,yielding2ndi erentsubsidiesinthevariousmarkets.Inaddition,twodi erent,carrier-speci ctollswouldbeleviedatthecongestedairport.Thesubsidiesandtollswouldseparatelyrepresentthemarket-powerandcongestionterms,suitablygeneralizedtomultiplemarkets,fromeitherthePigouviantollformula(20)orthemanipulableformula(22).A nalpointregardingthecorrectionofthemarket-powerdistortionconcerns rst-degreepricediscrimination.Undersuchbehavior,everyticketwouldbesoldatapriceequaltotheconsumer'smaximumwillingnesstopay,andthemarket-powerdistortionwouldvanish.Althoughpricediscriminationinfaresettingisstandardpracticeintheairlineindustry,itdoesnottakeonthisidealform,sothatsomedownwardadjustmentofcongestiontolls(orthepaymentofseparatemarket-levelsubsidies)willremainnecessarytocompensateforexcessivefares.4.2.FeasibilityofamanipulablerulewithacommonaveragetollAsseeninsection3.2,arestrictedmanipulabletoll,wherebothcarriersarechargedthesameaveragetoll,isfeasibleinthebasicmodel.Anaturalquestioniswhethersuchatollisfeasibleinthegeneralsettingunderconsideration.Onemightexpectthatcarrierasymmetrymakesarestrictedtolllikethatderivedaboveinfeasible,andtheensuinganalysisshowsthatthisconjectureisindeedcorrect.Eventhoughthisresultmaycomeasnosurprise,theanalysisfurtherilluminatestheissuesinvolvedinderivingmanipulabletollrules.Asbefore,whencarrierspayacommonaveragetollt(),themarginalmanipulabletollisgivenby@Ti(f1;f2)=@fi=t(f1+f2)+fit0(f1+f2),sothatthemarginal-tollcondition(20)mustholdwiththelatterexpressionreplacingtheleft-handsideexpression.Inthepreviousanalysis,theanalogto(20)(namely,eq.(10))reducedtothesinglecondition(11)givensymmetryofthecarriers.However,withasymmetry,(20)remainsastwoseparateequationsinvolvingthe15 Integratingthe rsttermsbypartsandimposingthepreviousrequirementKi=fici(fi)ontheconstantsofintegration,(21)reducestoTi(f1;f2)="Zfi0D(x+fi)dxfiD(f1+f2)#+fi(ci(f1+f2)ci(fi));i=1;2:(22)Notethat(22)equalstheincreaseintheothercarrier'scongestioncostduetocarrieri'soperationsminustheadditiontoconsumersurplusfromtheseoperations.Withtollliabilityrule(22)inplace,airlinei'spro tfunctionbecomese ectivelyequaltosocialsurplusabovethelevelachievedifonlyfiweresupplied.Thisfactcanbeveri edbysubstitutionof(22)intotheexpressiongivingpro tnetofthetollandcomparingwith(16).Theintuitionissimple:ifatollmanipulatoristobeseducedtobehavesoastomaximizesocialsurplus,thetollliabilityshouldbede nedsothatpro tinclusiveofthisliabilityshouldvaryperfectlyinparallelwithsocialsurplus.Inhindsight,itwouldhavebeensurprisingifthiswouldnothavebeentrue.Thedistributionalimpactofthetollschedule,rewardingthecarrierwiththefullincreaseinsocialsurplusitcreatesgiventhecompetitor'soutput,mayofcoursebeconsideredundesirable.TheKitermsmaythenbeusedtoadjustthisimpactbysettingthematvaluesdi erentfromKi=fici(fi).Butwhattheanalysisshowsisthatpost-taxpro tsshouldvaryperfectlyinparallelwithsocialsurplusforthemanipulabletolltobeoptimal,apropertythatisindependentofthechoiceofKi.ThisrequirementleavestheKitermsasinstrumentstoaddressdistributionalconcerns.AsinthecaseofthePigouviantollfrom(20),thetotaltollliabilityin(22)maybepositive(anettax)ornegative(anetsubsidy).Thelatteroutcomeemergeswhenconsumersurplusislargerelativetothecongestionexternality(requiringsucientlyinelasticdemand).Thisambiguityre\rectstheuseofasingletollinstrumenttoaddresstwoopposingdistortions:amarket-powerdistortionthatinecientlylimits\rightvolumes,andacongestionexternalitythatinecientlyin\ratesthem.AssuggestedbyBrueckner(2005),however,thesetwodistor-tionscouldbeaddressedbydi erentinstrumentsinasituationwherethedistortionscanbeseparated.Suchasituationariseswhenthecarriersservemultiplemarkets,witheachprovid-14 Using(1),the rst-orderconditionforpro tmaximizationintheabsenceoftollsisD(f1+f2)+fiD0(f1+f2)ici(f1+f2)fici0(f1+f2)=0;i=1;2:(18)Notethatthesecondtermin(18)arisesfromtheexploitationofmarketpower,whichallowsthecarrierstoraisethefarebylimiting\rights.WhenaclassicalPigouviantollisusedtoeliminatethedi erencebetween(17)and(18),carrieri'stollequalszi=fiD0(f1+f2)+fici0(f1+f2);(19)asseeninPelsandVerhoef(2004).Theasterisksagaindenotesociallyoptimalvalues,whicharenowasymmetric,asarethePigouviantollsthemselves.Notethatthe rsttermin(19)isnegativewhendemandisimperfectlyelastic,indicatingthatthetollisadjusteddownwardtomitigateover-pricingbythecarriers.Thiscorrectionvanishesascarriersbecomein nitesimallysmallandfiapproacheszero.ThedownwardadjustmentofPigouviantaxesundermarketpowerwasoriginallyderivedbyBuchanan(1969),anditfurthererodesthescopeforself- nancingofoptimalairportcapacity.Observethatthemarket-poweradjustmentin(19)islargerforbiggercarriers,o settingtheirgreaterincentivetorestrictoutputinordertoraisethefare.Asbefore,themanipulabletollruleTi(f1;f2)forcarrierishouldbesetsothatthemarginaltollcoincideswiththePigouviantollrule.Using(19),thisrequirementyields@Ti(f1;f2)@fi=fiD0(f1+f2)+fici0(f1+f2);i=1;2:(20)Themanipulabletollruleisfoundbyintegrating(20),whichyieldsTi(f1;f2)=Zfi0xD0(x+fi)dx+fici(f1+f2)+Ki;i=1;2:(21)13 3.5.AirportcostrecoveryThewell-knownself- nancingtheoremfromroadpricingtheorysaysthat,whenroadsarebuiltwithneutralscaleeconomiesandanadditionalzero-degreehomogeneityassumptionholdsforthecongestionfunction(i.e.,adoublingoftrac\rowandcapacitywouldleavethetraveltimeconstant),thecostoftheoptimal-sizeroadisexactlycoveredbyPigouviancongestiontollrevenue(MohringandHarwitz,1962).Thisresultfailstohold,however,intheairportcontextduetointernalizationofcongestion,whichgeneratessmallertollsthanintheroadsetting(foraformalderivation,seeBrueckner(2002,2008)).Giventheconclusionsderivedabove,therevenueshortfallisevenlargerunderarestrictedmanipulabletollrule.Inotherwords,sincetheaveragemanipulabletollissmallerthanthePigouviantollinequilibrium,andsincethelattertollitselfalreadyfailstocoverairportcost,alargerde citemergesinthemanipulablecase.SincethegeneralmanipulabletollalsogeneratesasmallerliabilitythanthePigouviantollwhenittakestheformin(11)andc()isconvex,costrecoveryagainfailsinthiscase.Thede citunderbothtypesofmanipulabletollsmustbecoveredbyadditionallump-sumcharges.4.TheGeneralModel:AsymmetricCarriersandImperfectlyElasticDe-mand4.1.TheunrestrictedmanipulabletollConsidernowthemoregeneralcasewherecarriersneednotbesymmetricanddemandisnotnecessarilyperfectlyelastic.Socialsurplusisnowde nedbyW=Zf1+f20D(x)dx2Xj=1fj(j+cj(f1+f2));(16)andthesocialoptimumrequiresD(f1+f2)ici(f1+f2)fici0(f1+f2)fici0(f1+f2)=0;i=1;2;(17)wherethesubscriptionc()denotesthefunctionbelongingtocarrieri'scompetitor.Notethat(17)di ersfrom(2)inthelackofsymmetryandtheappearanceofD()inplaceofp.12 Butrecognizingthatthe\rightvolumesaresymmetricinequilibrium,f1=f2=F2holds,whereFisthetotal\rightvolume.Makingthissubstitution,(12)canberewrittenast(F)+F2t0(F)=F2c0(F):(13)Thisconditionisalinear, rst-orderdi erentialequationintheunknownfunctiont(F).Giventhepresenceofthe1=2factor,theleft-handsideexpressioncannotbeintegrated,ap-parentlypreventingderivationofageneralsolution.However,asolutioncanbederivedunderafairlygeneralfunctionalformforc().Inparticular,ifc(F) F,where1,thenat(F)functionwiththesameexponentbutadi erentmultiplicativefactorcansatisfy(13).Lettingt(F)= F,(13)reduces,afterdi erentiationandsubstitution,tothefollowingrequirement: F+F2 F1=F2 F1;(14)whichissatis edwhen = =(2+).Thus,themanipulableaveragetollfunctionforthisspecialcaseisgivenbyt(F)= 2+F(15)Whencarrierifacesatollliabilityoffit(f1+f2)=fi (2+)(f1+f2),theresultingpro t-maximizing\rightvolumesaresociallyoptimal.TheequilibriumtollpaymentunderthisrestrictedmanipulablefunctionissmallerthaninthePigouviancase.Undertheassumedformofc(),themarginalPigouviantollper\right(eitherintheclassicalcaseorthenon-classicalequilibrium)is(F=2)c0(F)=( =2)F,whichexceedstheaveragemanipulabletollinequilibrium,givenby(15)evaluatedatF(thesociallyoptimaltotal\rightvolume).Thisrelationshipalsoholdsingeneral,ascanbeseenfrominspectionof(13).Sincetheright-handsideevaluatedatFisthePigouviantollper\right,itfollowsthatt(F)mustbesmallerthanthismagnitudeaslongast0()ispositive.Intuitively,theaveragemanipulabletollislessthanthePigouviantollwhentheaveragetollisincreasingsincethise ectprovidesanadditionaldeterrentinlimitingacarrier's\rightvolume.11 plusaconstant.WhileKicouldbesetata xednumericalvalue,itneedonlybeindependentoffiandthuscouldbeafunctionoffi,theothercarrier's\rightvolume.Giventhispossibility,anaturalchoiceistosetKi=fic(fi),sothatTi(f1;f2)=fic(f1+f2)fic(fi):(11)Then,carrieri'smanipulabletolliszerowhenfi=0andequalstheincreaseintheothercarrier'scongestioncostduetothe\rightsoperatedbycarrieri.ItisinterestingtocomparethetollliabilitiesinthemanipulableandPigouviancases.Whenthemanipulabletolltakestheformin(11),thetollliabilitycanberewrittenasRfi0fic0(x+fi)dx,whileunderPigouviantollingitequalsfific0(fi+fi),withbothexpressionsevaluatedatthesocialoptimumtogenerateequilibriumvalues.Ifc()isstrictlyconvex,thesecondexpressionislargerforanyvaluesoffiandfi,implyingthatthetollliabilityunderPigouviantollingexceedsthatundermanipulabletolling.However,giventhefreedomtoadjustKi,adi erentchoicecanmakethemanipulabletoll'sliabilityexceedthatofthePigouviantoll.3.4.ArestrictedmanipulabletollThemanipulabletollin(10)hasaformthatmightbeviewedasunappealinginapracticalsense,giventhatitembodiesachargethatdependsontheothercarrier'scongestioncost.Toeliminatethisdrawback,adi erenttollrulecanbederived,subjecttoamorenaturalrestrictionontherule'sform.Inparticular,supposethatacarrier'stollliabilityisrequiredtoequalitsown\rightvolumetimesafunction,commontobothcarriers,thatdependsonthetotal\rightvolumeattheairport.Stateddi erently,therequirementisthattheaveragetollpaidbyacarrier,equaltoitstollliabilitydividedbyitsown\rightvolume,begivenbyacommonfunctionthatdependsonthe\righttotal.Lettingthisaveragetollfunctionbewrittent(f1+f2),carrieri'stollliabilityisthengivenbyfit(f1+f2).Thus,therestrictionimpliesT1(f1;f2)f1t(f1+f2)andT2(f1;f2)f2t(f1+f2).Underthisrestriction,@Ti(f1;f2)=@fi=t(f1+f2)+fit0(f1+f2)holds,sothat(9)becomest(f1+f2)+fit0(f1+f2)=fic0(f1+f2):(12)10 volumechoices,whichcanbeexploitedindecisionmaking.Althoughthetollruleisgivenforthecarrier,thetolllevelhasthus,intentionally,become\manipulable."LetTi(f1;f2)denotethemanipulabletollruleforcarrieri,whichgivesitstotaltolllia-bilityasafunctionofboth\rightvolumes(carrieri'stollper\right,ti,isthereforeequaltoTi(f1;f2)=fi).Facedwiththisfunction,carrieri's rst-orderconditionforchoiceoffiispc(f1+f2)fic0(f1+f2)@Ti(f1;f2)@fi=0;i=1;2:(8)ThegoalistochoosetheTifunctionssothatthesolutionstothetwoconditionsin(8)coincidewiththesocialoptimum.Thisgoalcanbeachievedifthemanipulabletollruleischosensothat(8)isthesameasthesocialoptimalityconditionin(2).Inspectionofthetwoconditionsshowsthatthiscoincidencerequires@Ti(f1;f2)@fi=fic0(f1+f2):(9)Theleft-handsideof(9)givesthemarginaltollasperceivedbyatoll-manipulatingcarrier:theadditionaltollliabilityresultingfromaddinganextra\right.Notethattheright-handsideof(9),whenevaluatedatthesocialoptimum,issimplyequaltotheleveloftheclassicalPigouviantollz,whichitselfrepresentsamarginaltollgiventhetollliabilityofzfi.Thus,(9)impliesthatthemarginaltollfromthemanipulablecase(whenevaluatedatthesocialoptimum)isequaltothemarginalPigouviantoll.Despitethiscoincidenceofmarginaltollsinthetwocases,itwillbecomeclearbelowthattheaveragetolls(zintheclassicalPigouviancasevs.Ti(f1;f2)=fiinthemanipulablecase)willnotbeequal,norwillthetotaltollliabilities.Integratingbothsidesof(9)withrespecttofidirectlyyieldsthemanipulabletollrule,whichisgivenbyTi(f1;f2)=fic(f1+f2)+Ki:(10)whereKiisaconstantofintegration.ThusT1(f1;f2)=f2c(f1+f2)+K1andT2(f1;f2)=f1c(f1+f2)+K2,sothatacarrier'stollliabilityequalstheothercarrier'stotalcongestioncost9 elicitsthesesame\rightvolumesaschoices.Thus,theregulator'sexpectationsincomputingthetollarecon rmed.Sincethemodi edversionof(5)coincideswiththeoptimalitycondition(2),theresultingequilibriumisecient,justliketheonebasedontheclassicalPigouviantoll.4Nowconsiderthecasewheretheairlinesknowthetollrule,perhapsbecausetheparametrictollhasbeenadjustedanumberoftimesinresponsetochangesintheequilibrium(duetovariationinpor,forexample)orbecausethetollruleispublic.Inaddition,supposethatthecarriersusethisknowledgetotheiradvantage.Speci cally,carrieriwouldviewpro tasbeinggivenby(p)fic(f1+f2)fific0(f1+f2)fi;(6)butthetollper\rightexpression(fic0(f1+f2))appearingthelasttermwouldnolongerbeviewedasparametric.Carrierswouldthenoptimizetakingaccountofthee ectoftheirdecisionsonthetollpaidper\right.Computingtherelevant rst-orderconditionandevaluatingatthesymmetricequilibrium,theresultingcommon\rightvolume,denotedef,satis espc(2ef)2efc0(2ef)ef2c00(2ef)=0:(7)Notethatcon rmationoftheregulator's\right-volumeexpectationsisagainensuredby(7).Theregulatorexpectssymmetric\rightvolumesofefandthuschargesatollper\rightofefc0(2ef).But(7)ensuresthatcarrierchoices,whichnowinvolvemanipulationofthetoll,generateacommon\rightvolumeofef,con rmingtheseexpectations.Thekeyobservationfrom(7),however,isthatsincec00�0,theequationissatis edatanefvaluesmallerthanf.Thus,manipulationofthe(non-classical)Pigouviancongestiontollleadstoaninecientlylow\rightvolume.53.3.AmanipulablecongestiontollTocircumventmanipulativebehavior,theplannercouldabandonthePigouviantollandinsteadannouncetothecarriersacompleterulethatdeterminestheirtollliabilitiesasafunctionof\rightvolumes.Thistollruleisdesignedtobemanipulatedinthesensethatcarriersaregivenfullinformationabouttheconnectionbetweentollliabilitiesandtheir\right-8 Whenzfiissubtractedfromcarrieri'spro t,thenew rst-orderconditionsarepc(f1+f2)fic0(f1+f2)fc0(2f)=0;i=1;2:(5)whichyieldthesolutionf1=f2=f.Intheroadcontext,whereinternalizationofcongestiondoesnotoccur,thePigouviantollwouldequalthefullmarginalcongestiondamageevaluatedatthesocialoptimum,givenherebyMCD=2fc0.Butinternalizationmeansthatatollthislargeisnotrequired,withzinsteadequaltoMCD=2.3.2.IncentivesformanipulationWhilethe\classical"Pigouvianapproachembodiedin(4)assumesthattheregulatorhassucientinformationtocomputetheoptimum,anactualapplicationofthePigouvianmethodmightbebasedonincompleteinformation.Theregulator,forexample,maylackinformationonthecostparameterorthefullpricep.Assuming,however,thattheregulatorknowsthecongestioncostfunctionc(),avariantoftheclassicalapproachcanbeenvisioned.Itgeneratesthesameoptimaloutcomeastheclassicalapproachwhencarrierstreattollsasparametric,buttheapproachalsoinvitestollmanipulation.Underthisalternateapproach,theregulatorcomputesthecongestiontollper\rightusingthePigouvianrulefic0(f1+f2).Ratherthanevaluatingtheruleatthesocialoptimum,whichtheregulatorisunabletocompute,theruleisevaluatedatthe\rightvolumesthereg-ulatorexpectstoemergeinequilibrium.Theseexpectationsmustbecon rmed,however,bycarrierchoices.Giventheirperceptionofthetoll(asparametricor,alternatively,subjecttomanipulation),the\right-volumechoicesofthecarriersmustmatchtheregulator'sexpecta-tionsregardingthesevolumes.Ifnot,theregulatorwouldadjusttheparametrictollsuntilexpectationsarecon rmed(achievingequilibrium).Inthecasewherethecarriersviewthetollasparametric,thisequilibriumrequirementisstraightforward.Inparticular,thePigouviantollper\rightfic0(f1+f2),whichisviewedasparametric,issubstitutedinplaceoffc0(2f)in(5).EvaluatingthePigouvianruleatthe\rightvolumesthatsatisfy(5)yieldsatollthat,whenperceivedasparametricbythecarriers,7 3.1.Socialoptimum,laissez-faireequilibrium,andPigouviantollsWithperfectlyelasticdemand,consumersurplusiszeroandsocialwelfareequalstotalpro t,givenby[pc(f1+f2)](f1+f2).Maximizingthisexpressionbychoiceoff1andf2yieldstwoidentical rst-orderconditions,givenbypc(f1+f2)(f1+f2)c0(f1+f2)=0:(2)Thisconditiondeterminesasymmetricoptimal\rightvolume,givenbyf1=f2=f.Notethat(2)saysthatextra\rightsshouldbeoperateduptothepointwherethefullpriceminusoperatingandcongestioncostsper\rightequalsthemarginalcongestiondamagegeneratedbyanextra\right,givenby(f1+f2)c0(f1+f2)MCD.Thisexpressionequalsmarginalcongestioncostper\right,givenbytheslopeofthecongestioncostfunction,timesthenumberof\rightsexperiencingtheadditionalcongestion.Thelaissez-faireequilibriumisgeneratedbypro tmaximizationunderCournotbehavior.Carrieri'spro tequals[pc(f1+f2)]fi,andthe rst-orderconditionformaximizationofthisexpressionispc(f1+f2)fic0(f1+f2)=0;i=1;2:(3)In(3),firatherthanf1+f2multipliesc0,indicatingthatcarrieridoesnottakeintoaccountthecongestiondamageimposedontheothercarrierwhenitschedulesanextra\right.Therefore,thesymmetricequilibrium\rightvolumes,equaltobf,aretoolarge,satisfyingbf�f.Notethat,eventhoughitignorestheimpactontheothercarrier,carrieridoesinternalizethecongestionitimposesonitself,viewingitsowncongestiondamage(fic0)aspartofthecostofoperatinganextra\right.APigouviantoll,ifviewedasparametricbythecarriers,canremedythisineciency.Thetollchargeseachcarrierforthecongestiondamagethatitdoesnottakeintoaccount,equalingfic0(f1+f2)per\rightforcarrieri,whereidenotestheothercarrier.Evaluatingthisexpressionatthesymmetricsocialoptimum,theclassicalPigouviantollisthengivenby3z=fc0(2f):(4)6 givenbyD(f1+f2)h(f1+f2).Thus,congestiongeneratesafarediscount,asveri edintheempiricalstudyofForbes(2008).Carrieri'srevenueisthenequalto[D(f1+f2)h(f1+f2)]fi,i=1;2.Inadditiontoraisingpassengertimecosts,airportcongestionincreasesairlineoperatingcosts,withthee ectallowedtodi eracrosscarriers.Congestioncostper\rightforcarrieriisgivenbygi(f1+f2),i=1;2,withthesefunctionsagainassumedtobeincreasingandconvexovertherelevantrange.Anairlinealsoincursoperatingcoststhatareunrelatedtoairportcongestion,coststhatagainmaydi eracrosscarriers.Assumingconstantreturns,thesecostsaregivenbyifi,whereiisoperatingcostper\right,i=1;2.Combiningtheaboveelements,airlinei'spro tisgivenbyi=[D(f1+f2)h(f1+f2)]fiifigi(f1+f2)fi=D(f1+f2)fiifici(f1+f2)fi;i=1;2(1)whereci(f1+f2)h(f1+f2)+gi(f1+f2)givespassengerplusairlinecongestioncost.Notethatthetwotypesofcongestioncostsenterthepro tfunctionsymmetricallybecauseanincreaseinpassengertimecostsimpliesanequallylargedecreaseinthefarethatcanbechargedforagivenoutputlevel.Firmpro tsarethereforeequallysensitivetoh(f1+f2)andgi(f1+f2).Socialwelfareismeasuredbytotalpro t,1+2,plusconsumersurplus,whichisgivenbyRf1+f20D(x)dx(f1+f2)D(f1+f2).Useofthiswelfarefunctionrequirestheabsenceofincomee ectsondemand,inwhichcaseconsumersurplusisanexactmeasure.3.BaseCase:SymmetricCarriersandPerfectlyElasticDemandItisusefultobeginbyconsideringabasecasewherefurthersimpli cationsareimposedonthemodel.Accordingly,supposethatdemandisperfectlyelastic,sothattheDfunctionfromaboveisequaltoaconstant,denotedp.Inaddition,letcostsbesymmetricacrosscarriers,sothatoperatingandcongestioncostsin(1)losetheirisubscripts.Afterderivingresultsforthisbasecase,theanalysisreturnstothegeneralcaseinsection4.5 demandisperfectlyelasticandcostsaresymmetric.Section3.1derivesthesocialoptimumforthiscase,andthencomputesthePigouviancongestiontollsrequiredtosupportit,assumingthatthetollsareviewedasparametricbythecarriers.Then,section3.2explorestheincentivesfortollmanipulation,relyingonthenon-classicalversionofthePigouvianapproachsketchedabove.Manipulabletolls,requiredinresponsetomanipulativebehavior,areanalyzednext.Theanalysisinsection3.3derivesthemostgeneralformforsuchtolls,andsection3.4considerstheirformunderaplausiblerestriction,whichrequiresacarrier'stollliabilitytoequalitsown\rightvolumetimesatollper\rightthatiscommonacrosscarriers.Section3.5considerstheissueofairportcostrecoveryunderthedi erenttollregimes.Section4focusesonthegeneralmodel,inwhichcarriersareasymmetricanddemandisnolongerperfectlyelastic,leadingtoinecientpricingmark-upsascarriersexercisemarketpower.Sections4.1and4.2respectivelyconsidergeneralandrestrictedtollschedulesforthegeneralcase,withthelatterinvolvingacommontollper\rightacrossthenow-asymmetriccarriers.Section5considersapplicationofthelessonsoftheanalysistootherexternalitycontexts,andsection6o ersconclusions.2.TheModelTheanalysisfocusesonasingletravelmarketwithacongestedairportatoneendpoint.Incontrasttosomeearlierpapers,themodeldoesnotdistinguishbetweenpeakando -peakperiods,sothatcongestionisalwayspresent.Themarketisservedbytwoairlines,denoted1and2,whichinteractinCournotfashion.Letfidenotethenumberof\rightsoperatedbycarrieri,andletthenumberofpassengersper\rightbeconstantandnormalizedtounity,sothatf1+f2representsboththetotal\rightvolumeandthetotalnumberofpassengers.Thedemandfor\rightsisgivenbytheinversedemandfunctionD(f1+f2),whichgivesthemarginalwillingnesstopayfortravel.Passengervolumeisdeterminedbyequatingthiswillingnesstopaytothe\fullprice"oftravel,whichincludestheairfareandthevalueoflostpassengertimeduetoairportcongestion.Withcongestiondependingontotal\rightsattheairport,timecostperpassengerisgivenbyh(f1+f2),afunctionthatisassumedtobeincreasingandconvexovertherelevantrangeof\rightvolumes(itmaybezeroatlowvolumes).Sincetheairfareplustimecostequalsthefullprice,itfollowsthatthefareis4 atomistic,raisingconcernsaboutmanipulation.Thisisthecaseofairportcongestion,whicharisesfromusageofacapacity-constrainedairportbyarelativelysmallnumberofairlines,someofwhichmayaccountforanappreciableshareofthetotal\rights.Forconcreteness,thepaper'sanalysisisdevelopedintheairportcontext.Departingfrompreviousstudies,arecentliteratureonairportcongestionrecognizesthenon-atomisticnatureofairlineswhileexploringaparticularconsequenceofthisalterationofthestandardroad-orientedmodel:internalizationofcongestion.Internalizationoccursbecauseanon-atomisticcarrier,inschedulinganextra\right,takesintoaccounttheadditionalcongestioncostsimposedontheother\rightsitoperates.Asaresult,aPigouviancongestiontollneedonlychargeanairlineforthecongestionimposedonothercarriers,excludingthecongestiontheairlineimposesonitself.Oneimplicationofthisruleisthat,whencarriersareasymmetric,theyshouldpaydi erenttolls.Acarrierwithalarge\rightshareshouldpayalowtollgiventhatitinternalizesmostofthecongestionfromitsoperationofanextra\right,whileasmallcarrier,whichinternalizeslittleofthecongestionitcreates,shouldpayahightoll.Usingasimulationmodel,Daniel(2005)wasthe rsttorecognizethepotentialforinternalizationofcongestion,whileBrueckner(2002,2005)andPelsandVerhoef(2004)exploredtheimplicationsofinternalizationusingsimpleanalyticalmodels.Aburgeoningliteraturehasfollowedtheseinitialstudies.1Asthisdiscussionindicates,therecentanalysisofairportcongestionpresumestheuseofPigouviancongestiontolls,eventhoughthenon-atomisticnatureofairlinessuggeststhepotentialformanipulationofsuchtolls.Thepresentpaperismeanttoredressthisomission.Byanalyzingthenatureofthemanipulablecongestiontollsinanairportcontext,itaddsamissingcomponenttothenewtheoryofcongestionpricinginthepresenceofnon-atomisticagents.Theresults,however,applymoregenerallytothetheoryofcorrectivetaxation,showingthatmanipulabletaxesmayneedtoreplacePigouvianchargesinothercontexts.2Theplanofthepaperisasfollows.Section2presentsthemodel,whichtakesthesimplestpossibleform.Twoairlinesserveasingletravelmarket,withacongestedairportatoneendpoint.Whilethegeneralmodelhasanelasticdemandfortravelandcostfunctionsthatpotentiallydi eracrosscarriers,section3beginsthediscussionbyconsideringabasecasewhere3 forexample).Then,ifthemarginaldamagefunctionisincreasing,thePigouviantaxwillvaryaswell,exhibitingpositivecorrelationwithboththeoverallactivitylevelandindividualactivitylevelswithinthesmallsetofagents.ThiscorrelationmayrevealtotheagentsthenatureoftheruleusedtocomputethePigouviantax,showingthatthetaxperunitdependsontheoverallactivitylevel.AnotherreasonwhyagentscouldbeawareoftheruleunderlyingthecalculationofPigouviantaxesisthattheprinciplesaccordingtowhichthetaxlevelisdeterminedmayhavetobemadetransparentandpubliclyaccessibleforlegalreasons.Agentscanthenobservetheruledirectly.Ineithercase,withthisknowledge,agentsmaythenattempttomanipulatethetaxregime,furtherrestrictingtheiractivitylevelsinordertodepressthemagnitudeoftheper-unittaxthattheypay.Suchbehavior,whichisshownbelowtoberelevantunderthealternate,non-classicalapproach,underminesthePigouviantaxregime,makingitincapableofachievingtheoptimum.Thepurposeofthepresentpaperistoexploreanalternativeapproachtocorrectivetax-ationthatrecognizesthepotentialexistenceofsuchmanipulativebehavior.Ourapproachconfrontsmanipulationhead-onbyreplacingtheconventionalPigouviantax,designedtobetreatedparametricallybyindividuals,withataxrulethatisdesignedtobemanipulated.Inotherwords,theplannerannouncesarulethatgiveseachagent'staxliabilityasafunctionofhisorherownactivitylevelandthelevelsofotheragents.Eachagentthenoptimizeswithfullknowledgeoftheruleusedtocomputethetaxliability,withtheoptimizationbeingcarriedoutconditionalonthechoicesofotheragents.Thus,insteadofusingaPigouvianrulethatoperatesbehindthescenestogenerateataxvaluemeanttobeviewedasparametric,ourapproachpresentsatransparent,manipulablerulethatisdirectlyexploitedbytheagentsinchoosingtheiractivitylevels.Thegoaloftheanalysisistoderivetheformofsuchmanipulabletaxrules,providingacomparisontothePigouviancase.Theanalysisfocusesoncongestionexternalities,wherePigouviantaxestaketheformofcongestiontolls.Traditionalanalysesofroadcongestionpricing(seeSmallandVerhoef(2007))aremostlyimmunetothemanipulationcritiquefromabove,aconsequenceofthefactthatroadusersaretypicallyatomisticandthusunabletomanipulatePigouviancongestiontolls.However,arecentliteraturefocusesonacasewheretheagentsgeneratingcongestionarenon-2 ManipulableCongestionTollsbyJanK.BruecknerandErikT.Verhoef*Pigouviantaxesarestandardinstrumentsfordealingwithnegativeexternalities.Theproblem,ofcourse,isthattheactivitygeneratingthenegativeexternale ectoccursattoohighalevel.ThePigouviantaxisdesignedtoraisethepricefortheactivity,reducingitsoverallleveltoonethatissociallyoptimal.Tocomputethetax,themarginalexternalitydamagefromtheactivityisevaluatedatthesocialoptimum,withthetaxoneachunitoftheactivitythensetattheresultingvalue.Facedwiththetax,theo endingagentsrestricttheirindividualactivitylevels,leadingtoanoptimaloveralllevel.Analternativetothis\classical"Pigouvianapproach,whichisusefulwhentheregulatorknowsthedamagefunctionbutlacksalltheinformationneededtocomputetheoptimum,istochargeatollequaltomarginaldamageevaluatedattheexpectedequilibriumactivitylevel.Inequilibrium,thisexpectationiscon rmed,andthesociallyoptimalactivitylevelagainemerges.Inatypicalapplication,themarginalexternalitydamageisitselfafunctionoftheoverallactivitylevel,usuallyanincreasingone.However,thecomputedtaxisjustascalarvalue,generatedbyevaluatingthemarginaldamagefunctionatthesociallyoptimalactivitylevel(ortheexpectedequilibriumlevel,underthealternateapproach).Thus,whilethePigouviantaxruleinvolvestheleveloftheactivity,theper-unittaxultimatelychargedisjustanumbergeneratedbyevaluatingtheruleatparticularvalue.ThePigouvianapproachassumesthatagentstreattheresultingtaxasparametric,inde-pendentoftheirchosenactivitylevels.Thisviewisappropriatewhentheexternalityisjointlygeneratedbymanyagents,eachofwhommakesasmallcontributiontotheoverallleveloftheo endingactivity.However,whentheexternalityisgeneratedbyjustafewlargeagents,aparametricviewofthePigouviantaxmaybelessplausible.Tounderstandthispoint,imaginethatthesociallyoptimalactivitylevelvariesfromperiodtoperiodasaresultofchangesintheeconomy'sparameters(shiftsindemandorcostcurves,1 ManipulableCongestionTollsbyJanK.BruecknerDepartmentofEconomicsUniversityofCalifornia,Irvine3151SocialSciencePlazaIrvine,CA92697e-mail:jkbrueck@uci.eduandErikT.VerhoefDepartmentofSpatialEconomicsVUUniversityDeBoelelaan11051081HVAmsterdamTheNetherlandse-mail:everhoef@feweb.vu.nlJanuary2009AbstractTherecentliteratureoncongestionpricingwithlargeagentscontainsaremarkableinconsis-tency:thoughagentsarelargeenoughtorecognizeself-imposedcongestionandexertmarketpoweroverprices,theydonottakeintoaccounttheimpactoftheirownactionsonthemag-nitudeofcongestiontolls.Whenlargeagentsareconfrontedwithtollsderivedunderthisparametricassumptionbutunderstandtheruleusedtogeneratethem,thetollsystemwillnolongerguidethemarkettothesocialoptimum.Toaddressthisproblem,thepresentpaperderivesalternate,manipulabletollrules,whicharedesignedtoachievethesocialoptimumwhenagentsanticipatethefullimpactoftheiractionsontollliabilities.