Agent-Intermediated Electronic Markets in International Freight Transportation

InternationalFreightTransportation

BarrieR.Nault

HaskayneSchoolofBusiness

UniversityofCalgaryAlbertS.Dexter

SauderSchoolofBusinessUniversityofBritishColumbia

August23,2003Abstract

Inmanyindustriesagent-intermediatedmarketsareinefficientbecauseinformationabout

latentdemandandsupplynevergetstomarket.Wedemonstratehowinformationtechnologyintheformofanagent-intermediatedelectronicmarket(EM)alleviatesthis

problembyenhancingtheagent-as-market-makerusingtheinternationalfreight

transportationindustryasanexample.WefindthatanEMincreasesagentparticipationandinvestmenttherebyincreasingdemandandsupply.BecauseoftradeoffsbetweenincentivesforinvestmenttheEMchoosesaprofitallocationbetweenagentsresultinginlimitedagentparticipation.Inaddition,whenpricedependsondemandandsupply

balances,priceandvolumeinthemarketcanincreasesimultaneously.

Keywords:ElectronicMarkets,Intermediation,Transportation

Pleasedonotquoteorcitewithoutauthors’permission.Commentsarewelcome.

c2003byBarrieR.NaultandAlbertS.Dexter.Allrightsreserved.Copyright󰀂

ContactAuthor:BarrieR.Nault,HaskayneSchoolofBusiness,UniversityofCalgary

Calgary,Alberta,Canada,T2N1N4,(403)220-2742,nault@ucalgary.ca

1Introduction

Inmanyindustriesmarketinstitutionsareinefficientbecauseinformationaboutlatentde-mandandsupplynevergetstomarketandmanyfeasibletransactionsbetweenbuyersandsellersarenotcompleted.Characteristicsofsuchmarketsincludeinsufficientnumbersofbuyersandsellers(thinmarkets),heterogeneousandcomplexgoods(specificity),andin-formationasymmetrieswhichmayhold-uptrade.Whenonlypartialinformationaboutdemandandsupplyisavailable,individualdemandandsupplyarenotproperlymatched,andmarketsdonotalwaysclear.Consequently,incentivestobringnewdemandandsupplyinformationtothemarketarereduced.The”traditionalmarket”inmanyindustriesusesanagentasanintermediarybetweencustomerandsuppliertoovercomethesecausesofinefficiency.

Twonewinstitutionsbasedonelectronicmarkets(EMs)haverecentlyemerged.ThefirstsubstitutesanEMfortheagentinthetraditionalmarket.Thissubstitutionoccursinindustrieswherethequalityofthegoodscanbeassessedatlowcost,andthetransferofownershipandpossessionisstraightforward.ThesecondEMsupportstradebetweenagents.Thislattercase,whichwecallanagent-intermediatedEM,isthefocusofourpaper.Thissupportisnecessaryinindustrieswherethequalityofgoodsishardtomeasureand/orthetransferofownershipandpossessioniscomplex.Intheseindustriesspecificdomainknowledgeheldbyagentsisvitaltoqualityassuranceandpost-transactionmarketclear-ingoperations.Thisviewisconsistentwiththetradepressthatsuggeststhecriticallessonlearntinelectronicmarketplacessofaristospecialize(Halper,1997)-thatis,understandingthespecialcharacteristicsofagivenmarketplaceiscrucialforperformance.Itisalsocon-sistentwithtransactioncosteconomics,whichpositsthatbecauseofadministrativecosts,distortions,andinefficienciesfromlow-poweredincentives,mostmarketshaveevolvedtokeeptransactionrisklowandpreservespecializedinformationthemarketneedstofunction

1

efficiently(Williamson,1981,1985).

Somebelievethatinformationtechnology(IT)throughitscosteffectiveenablingofEMswillbeusedtodisintermediatemarkets(Gellman,1996;Hoffman,1995;WigandandBen-jamin,1996).However,inthemostcommonlyusedexamples-on-linesecuritiestradingandairtravelreservations-marketshavenotbeendisintermediated.Rather,theinterme-diaryhasbeen(partially)automated.On-linetradingrequiresabrokertocarryoutthetransactionandmoston-lineairlinesreservationsaredonethroughtravelagentsthathaveanInternetpresence.Indeed,manybelievethattherewillnotbedisintermediation,justnewtypesofintermediaries(BaileyandBakos,1997).AclassicexampleisAucnet-theelectronicauctionforusedcarsinJapan.Aucnetisanagent-intermediatedEMwithdealersactingasagentsfortheircustomersinthebuyingandsellingofcars.Supportingthevalueofagentintermediation,researchhasfoundthattheaveragecontractpriceofsecondhandcarssoldthroughAucnetismuchhigherthanthatoftraditional,non-electronicmarkets(Lee,1998).Newtypesofintermediariesprovideinformationtobuyersandsellers,aggre-gatedemandorsupply,managephysicaldeliveriesandpayments,andprovidetrustandintegritytotheEM(Bakos,1998).Intermediariessuchastrustedthirdpartieshavebeenproposedasawayofprovidingqualityassurance(Ba,WhinstonandZhang,1999).Trustiscriticaltoqualityassurance,andithasbeenfoundtobeanimportantdeterminantofEMadoption(Chircu,Davis,andKauffman,2000).Eveninfinancialmarketswherequalityassuranceoftheprocessratherthanthetradedgoodistheissue,specialistauctionswerefoundtoincreasemarketqualitybyreducingbid-askspreads,transactioncosts,andtimetoexecutionascomparedtoautomatedordermatchingsystems(ClemonsandWeber,1996).IThasalsobeenusedtoexplainrestrictionsinthenumberofsuppliersthatparticipateinbuyer-supplierrelationships.Forexample,BakosandBrynjolfsson(1993)showhowincen-tivescanmotivateabuyertolimitthenumberofsuppliersituses.Thatis,toinducesupplierinvestmentsinIT-relatednon-contractibleassets,commitmentsbybuyersaremorecredible

2

iftheylimitthenumberofparticipatingsuppliers.Banker,KalvenesandPatterson(2000)showthatthenumberofcontracttermsinacontractfallwithincreasinglevelsofITduetomoreeconomicalmonitoring.Ifsuppliersfacerelationship-specificcosts,theincreasesinITcancausethebuyertolimititsnumberofsuppliers.Riggins,KriebelandMukhopadhyay(1994)showthatbecauseofthenegativeexternalities(e.g.,competition)betweensuppliers,participationgrowthininterorganizationalsystemsstallswithoutbuyersubsidies.Finally,WangandSeidmann(1995)showthatasupplier’sadoptionofelectronicdatainterchange(EDI)cangeneratepositiveexternalitiesforbuyerswhileproducingnegativeexternalitiesbetweensuppliers.Thus,itmaybeinthebuyers’interesttosubsidizesuppliers’adoptionofEDI.Regardless,ifsupplieradoptioncostsarehigh,thenpartialadoptionbythesupplierbasemaybeoptimalforthebuyer.

OurExampleAnimportantexamplewheredemandandsupplyinformationinthemarketisincompleteistheless-than-full-loadoceanshippingindustry.Lumpysupply-acontainer-togetherwithless-than-containersizeshipmentsoftenresultsinpartiallyemptycontainersbeingshippedwhilepartialcontainershipmentsthatcouldhavefilledthepartiallyemptycontainersremainonthedock.Theinformationrequiredtocompletetransactionsinoceanshippingisextensive.Individualshipmentshaveaspecificoriginanddestination-possiblywithapreferredrouting,andrestrictionsonwhentheymustdepartandarrive.Availablecontainercapacitymaynotbetherightkindofcapacitybecauseofshipmentdimensionsorspecialshipmentrequirementssuchasrefrigeration.Thismarketcoordinationproblemisaninstanceofthemoregeneralless-than-full-loadproblemthatprevailsinallmodesofthetransportationindustry.Arecentreportindicatesthatofa$921BUSlogisticsmarket,18%ofcapacitymovesempty,leadingtoacollectivelossof$165B(Lynch,2000).

Agents,suchasinternationalfreightforwarders,brokers,non-vesseloperationcommoncarriers(NVOCCs),andthirdpartylogisticsproviders(3PLs),serveasmarket-makers,

3

providingsearch,pricediscoveryandnumerousmarketclearingoperationsforbothshippersandcarriers.Onthedemandside,theyprovideshipperswithfinancialservices(lettersofcredit)andinsurance,oftenassumingsomeoftheliabilityofcarriage.Theyalsohaveexpertiseinmanagingtheloading/unloadingandpick-up/deliveryprocess-inparticular,exportpermitsandcustomsclearanceforinternationalshipments(SherwoodandBruns,1992).Consequently,onthedemandsidetheseagentsneedtoinvestinspecializedknowledgesuchascustomsregulationsinparticularregionsinordertosupporttheaboveactivities.Inaddition,theyneedtoinvestusualdemand-generatingactivitiessuchaspromotionalvisitstocustomers(i.e.,shippers).Onthesupplysidetheyprovidecarrierswithconsolidationofshipmentsandthediscoveryofavailableless-than-fullloadcapacity.Agentinvestmentsonthesupplysideareinvestmentsinspecializedknowledgeofindividualcarrieroperationssuchasmethodsusedtoconsolidatefreight,andcontactsinportsandstagingareas.Agentsarespecializedtoservenichemarkets,forexample,transportingindividualau-tomobilesfromEuropetotheU.S.Asaresult,theoceanshippingmarkethasmanyagentsbecausespecializedknowledgeresideswithindividualagents,andeachagent’sexpertisedif-fersfromanother’s.Moreover,theknowledgeonwhichagentexpertiseisbasedchangescon-tinuously(seeHayek,1945).Nichemarketsandspecializedknowledgeisthedifferentiationthatmitigateshead-oncompetitionbetweenagents.Anexampleofanagent-intermediatedEMinoceanshippingisGTNexus(recentlyrenamedfromTradient)whichservesmerchantimportersandexporters,freightforwarders,3PLs,andcarriers.Testifyingtothecomplex-ityofcommercialshipping,andinourviewtheneedforagentintermediation,”...atypicalcommercialshipmentinvolves9differentparticipants,20separatedocuments,35customer-vendorinteractions,and4modesoftransport.Itallcantakeplaceoverthecourseofseveralweeksormonths,andcancrossmultipleinternationalborders.”(GTNexus,2001).Underutilizationofcapacityoccurspreciselybecausethetraditionalmarketisdecen-tralized-individualagentstrytomatchtheirowndemandtotheirownsupply,combining

4

twoaspectsofeachagent’sactivities-promotiontoshippers(generationofdemand)andprovisionofcapacity(generationofsupply).Thiscouplingisevidentwhenanagentpur-chasesallorpartofacontainerfromacarrierandpromotesthatcapacitytoshippersor,alternatively,arrangesforashipmentandthensearchesfortheneededcapacity.Theprob-lemisthatagentsfrequentlygeneratedemandwhichtheycannotindividuallysatisfywiththeirgenerationofsupply,andviceversa.Nonetheless,thistraditionalmarketstilloperatesmoreefficientlythandirectshipper(customer)tocarrierarrangementsbecausedirectship-pertocarrierarrangementscannottakeadvantageofthemoreextensiveexpertiseanagentprovides.

Theagent-intermediatedEMweproposebroadensthetraditionalmarketbymatchingunsatisfieddemandgeneratedbyoneagenttounderutilizedsupplyprovidedbyanother,freeingagentsfromneedingtomatchtheirowndemandtotheirownsupply.TheEMallowsagentstoanonymouslyexchangeinformationaboutcapacitytheyneedorhavewithoutrevealingtheidentityoftheirshipperorcarrier(priortothetransaction).Itdoesnotrequirethatagentsrevealallthedemandandsupplytheygenerate-whenanagentcanmatchitsowndemandandsupplyitmayoperateasitalwayshasinthetraditionalmarket.Nordoesitprecludeoff-EMarrangementsbetweenindividualagents.Inthesemarketsagentsinvestinfindingdemand(shipments)orsupply(capacity).ThroughtheEMagentscanberewardedforgeneratingeithersideofthetransaction.Indeed,theagent-intermediatedEMtogetherwiththetraditionalmarketcanresultinhigherdemandthanatraditionalmarketalone:becauseofthedifficultiesandadditionalcostsoffindingcapacityforidiosyncraticshipments,manyshippersintraditionalmarketsdonotcompletetransactionstheyotherwisewouldifaccesstotheirspecifictransportationneedswaslessexpensive.

Inourexampleofoceanshippingtherehavebeentworecentregulatorychanges.Priorto1984oceanshippingwashighlyregulatedandcontracttermswerestandardizedandpublic.The(U.S.)ShippingActof1984deregulatedcontractstoallowgreaterflexibility

5

inthetypesoftariffsandcontracts(LewisandVellenga,2000;ShashikumarandSchatz,2000).Inaddition,TheActbeganthemovementtowardsgreaterconfidentialitysuchthatonlytheessentialtermsofacontract(originanddestination)werepublished.Specifically,prices,capacity,andothercontractdetailsarenowconfidential.TheOceanShippingReformActof1998wentevenfurtherthantheearlierActbyremovingfrompublicvieworiginsanddestinationsforthroughintermodalmovements,rates,liquidateddamagesfornon-performance,andservicecommitmentssuchasassuredspace,transittime,andportrotation.Theseregulatorychangeshaveencouragedspecializedcontractsandterms,therebyfosteringtraditionalmarketsandEMswepropose.Becausetraditionalmarketsinvolveasingleagent,thesemarketsprovidefortheconfidentialityofcontractsandterms.TheEMsweproposeuseITspecificallydesignedtoprotecttheconfidentialityandanonymityoftheirparticipatingagentsandtheirclients.

Ourresearchobjectivesaretoexamineagentinvestmentandagentparticipationinagent-intermediatedEMstogetherwithtraditionalmarketsascomparedtoinvestmentandpartic-ipationintraditionalmarketsalone,andtoexaminethepotentialeffectsresultingfromtheadditionofanagent-intermediatedEMonmarketprices.Ourmainresultsareasfollows.Withanagent-intermediatedEM,agentinvestmentishigherandmoreagentsparticipate,increasingdemandandsupplyrelativetothetraditionalmarketalone.However,theopti-malallocationofprofitsbetweenagentsresultsinlimitedmarketparticipationwhereonlyaselectgroupofagentsparticipate.Inaddition,whenpricedependsondemandandsupplybalancesintheagent-intermediatedEM-balancesthatareaffectedbytheEMthroughincreasedagentinvestment-wefindthatbothpriceandtotalvolumemayincreasesimul-taneously.Thelimitedmarketparticipationresultsbecauseoftradeoffsbetweenincentivesforinvestmentindemandandinsupply:anincreaseintheallocationofprofitforagiventransactiontoagentsthatgeneratedemandraisesdemandgeneratinginvestmentandlowerssupplygeneratinginvestment(andviceversa).Simultaneousincreasesinpriceandvolume

6

canoccurwhenanincreaseinpriceresultsingreatervolumeimpactsthroughsupplygener-atinginvestmentandagentparticipationthanthroughdemandgeneratinginvestmentandthedirecteffectofpriceonvolume.

Apartfromoceanshipping,otherindustriesusevariantsofagent-intermediatedEMs.Inthemotorcarriertransport,DATServices(DAT,1999)offerselectronicfreightmatch-ingtoagentsandcarriers-providingelectronicsupporttoagentsgeneratingdemand,butbypassingagentprovisionofsupply.Indeed,carriersmaycarryouttheagentfunctioningeneratingsupplyinourmodel.Thus,ourresultsgeneralizetoagent-intermediatedEMswherecarriersareagentsonthesupplyside.Inaddition,theNationalTransportationEx-change(NTE,1998)allowsshippersoragentstopostloadsandmembercarrierstoselectloadstofilltheirunusedcapacity.NTEpre-authorizesshippers/agentsandcarriers,andservesstandardizedloadsanddomesticroutes,therebyeliminatingproblemsofqualityas-suranceandownership/possession.Inasimilarindustry,aircargo,effortsbycarrierstolinkdirectlywithcustomers,therebybypassingagents,failedbecausecarriersdidnotaccountfortheagents’specialknowledgeaboutcustomersandothercarriers(ForsterandKing,1995;Alt,ForsterandKing,1997).Anexampleofanagent-intermediatedEMoutsidethefreighttransportationindustryistheMultipleListingService(MLS)intherealestateindustry,amarketthatevolvedinmanualformpriortotheInternet.Thatindustryisdominatedbyagentintermediationratherthandirectbuyer-sellertransactions,andhasagentsrepresent-ingeachsideofatransaction-onethatservesthebuyerandanotherthatlistspropertyfortheseller.

Thepaperisorganizedasfollows.Wefirstexplainthestructureofoursetting,thenota-tion,andassumptions.Nextwecomparethetraditionalmarkettotheagent-intermediatedEMwhenpricesaredeterminedoutsidetheEM.ThenweconsiderwhenpricesareaffectedbydemandandsupplybalancesintheEM,andexploretheeffectoftheEMonwelfare.

7

2StructureoftheSetting

Werepresentagentsbyv󰀁[v,v],wherealargervcorrespondstoanagentwithgreaterpotentialtogeneratedemandandsupply.Anagentcanhavegreaterpotentialifithasalargermarketnicheitcanservewithitsexpertise,orhavemoreindividualcontactswithshippersorcarriers-eitherwouldyieldagreaterreturntoitsinvestment.vhastheuniformdensityh(v)>0overthesupportandiszeroelsewhere.AgentsdecidewhethertoparticipateintheEMandhowmuchtoinvest.Therearetwoinvestmentvariablesforeachagent,dvandsv.dv󰀁[0,dv]isinvestmentingeneratingdemand,forexample,findingcustomerswithorderstoshipless-than-containerloads.sv󰀁[0,sv]isinvestmentingeneratingsupply,forexample,findingless-than-fullcontainersbeingshipped.Wemeasuredvandsvindollarssothatdvandsvalsorepresentthecostsofeachtypeofinvestment,whichareassumedconstantatthemargin.Agentsalsofaceafixedcostofoperations,K.

Inthetraditionalmarketeachagentgeneratesdemandandthenfillsthatdemandwithsupplyitobtainsfromsuppliers,receivingpricep.Werefertothisdemandas”own”demandandrepresentitbyqa(v,dv,sv,p).Thepriceiswhatremainsafteraccountingforthecostofcarriercapacityandtheagent’smarginalcostofcompletingthetransaction,includingadditionalhandlingandadministrativecosts.InitiallyweassumepriceisdeterminedoutsidetheEM,andrelaxthisassumptionlater.

Theagent-intermediatedEMallowsagentstoseparatetheirgenerationofdemandfromtheirprovisionofsupply.Inadditiontothedemandgeneratedandfilledbytheagentinthetraditionalmarket,theEMallowsoneagenttosupplythedemandgeneratedbyanotheragent.Thedemandgeneratedbyagentvandfilledbysupplyfromotheragentsisqb(v,dv,s\\v,z,p),where“\\”means“otherthan”,andz󰀁[0,1]istheproportionofagentsparticipatingintheEM.Demandgeneratedbyotheragentsthatisfilledbysupplyfromvisqc(v,d\\v,sv,z,p).

8

Weassumethatthedemandsarereal-valuedfunctionsandarenon-negativeovertherangeoftheirarguments.Thedemandsarenon-decreasinginalloftheirargumentsexceptprice,marginalreturnsarenon-increasinginowninvestment,andwetakethedemandstobetwicecontinuouslydifferentiablewhereneeded.Significantstructureisimposedonthetypeandsourceofinvestmentthataffectseachdemandfromthenatureofthereal-worldproblem.Specifically,duetonichemarketsandspecializedknowledgewhichmitigatesdirectcompetitionbetweenagents,foreachdemandfacedbyanindividualagent(e.g.,eachqi)therearenocrosseffectsofinvestmentsingeneratingdemandsbyvandagentsotherthanv,andsimilarlyforinvestmentsingeneratingsupply.Ouranalysismakesuseofseveralassumptionsbeyondthoseimplicitinthestructure.

Assumption1(SizeEconomies)Agentswithgreaterpotentialhave(weakly)greaterre-turnstoadditionalinvestment:

∂2qa(v,dv,sv,p)/∂dv∂v,∂2qa(v,dv,sv,p)/∂sv∂v,∂2qb(v,dv,s\\v,z,p)/∂dv∂v,∂2qb(v,dv,s\\v,z,p)/∂s\\v∂v,∂2qc(v,d\\v,sv,z,p)/∂d\\v∂v,∂2qc(v,d\\v,sv,z,p)/∂sv∂v≥0.Aswedetailedearlier,agentshavegreaterpotentialiftheyhavelargermarketnichesorotherwisehavemorecontactswithshippersandcarriers.InAssumption1,foragentswithgreaterpotential,marginalreturnstoinvestmentarelargerduetotheirlargermarketormoreextensivecontacts.

Assumption2(InvestmentComplementarities)Marginalreturntoinvestmentinde-mandis(weakly)greaterwhenthereisgreaterinvestmentinsupply,andviceversa:∂2qa(v,dv,sv,p)/∂dv∂sv,∂2qb(v,dv,s\\v,z,p)/∂dv∂s\\v,∂2qc(v,d\\v,sv,z,p)/∂d\\v∂sv≥0.Assumption2impliesthatwhenthereisgreaterinvestmentinsupplythereismorecapacityavailable(havingbeenfoundfromcarriers).Henceforeveryadditionaldollarspenton

9

demand-thatis,recruitingshippers-thereisahigherchanceofamatchwithavailablesupply.WeuseweakinequalitiesinAssumptions1and2toaccountforshipmentsthataresufficientlyidiosyncratic-forexample,liveanimals-sothatthenextdollarofinvestmentdoesnotyieldamatchwithcapacity,andsimilarlyformatchingbetweenidiosyncraticcapacity-forexample,refrigeration-andshipments.

Assumption3(Separation)FortheagentwiththelowestpotentialparticipatingintheEM,thetotaleffectofagentpotentialonprofitsispositive.

ThisassumptionservestoseparatethecontinuumofagentsintotwosegmentsbasedontheirparticipationintheEM,similartothesingle-crossingconditionintheincentivesliterature.Assumption3isreasonableinoursetting,asweexplainwhenAssumption3isused.Notealsothatiffixedcosts,K,areanincreasingfunctionofagentpotential,thenAssumption3stillsufficesforseparationbuthasagreaterimpactsincethechangesinfixedcostswouldworkagainstprofitsincreasinginagentpotential.

3AgentInvestment

TheTraditionalMarketEachagentchoosesitslevelofinvestmentingeneratingdemandandgeneratingsupply,dvandsvrespectively.Throughoutwemodelagentsasprice-takerswherebyeachagentisnotlargeenoughforitsindividualinvestmentdecisiontoimpactprice.Weexaminetheimpactofapricedeterminedbyaggregatedemandandsupplyconditionslater.Agentprofitsarepricetimesowndemand,lesscostsofinvestmentandlessthecostofoperationsK,

π(v,dv,sv)=pqa(v,dv,sv,p)−dv−sv−K.

10

Thefirst-orderconditionsforagentvtomaximizeprofitsare

p∂qa(v,dv,sv,p)/∂dv−1=0andp∂qa(v,dv,sv,p)/∂sv−1=0.

(1)

FromAssumptions1and2allcross-partialderivativesoftheprofitfunctionare(weakly)positive.Thesolutionto(1)forallagentsyieldsaNashequilibrium.Becauseagentprofitsdonotdependonprofitsbyotheragents,profitsaretriviallysupermodularindifferentagents’investments,andbeingsupermodularsecond-orderconditionsarenotrequiredforanequilibrium.Theequilibriumresultsdirectlyfromindependentoptimizationsforeachagent.Theequilibriuminvestmentsareafunctionofpriceandaredenotedbydv(p)andsv(p).Becauseprofitsaresupermodularinagentpotentialandeachoftheinvestmentvariables,equilibriuminvestmentlevelsareincreasinginv.

TheAgent-IntermediatedElectronicMarketInanagent-intermediatedEMeachagentprofitsasitwouldinthetraditionalmarket,andinadditioncanprofitonitsdemandwhenmatchedwithsupplyfromotheragentsanditssupplywhenmatchedwithdemandfromotheragents.Theagentwiththedemandcollectstheprice,paysatransactionfeetotheEM,andpaysthemargintotheotheragentthatprovidesthesupply.Theagentprovidingthesupplycollectsthemarginfromtheagentwiththedemand.Thedecisionofwhoactuallycollectsandredistributesthepriceisarbitraryanddoesnotaffectthesubstanceofourresults.TheprofitfunctionfortheagentusingtheEMis

ψ(v,d,s,z)=pqa(v,dv,sv,p)+[p−f−m]qb(v,dv,s\\v,z,p)+mqc(v,d\\v,sv,z,p)

−dv−sv−K

wherefisthetransactionfeepaidtotheEM,misthemarginpaidtothesupplyingagent,anddandsarevectorsofdemandandsupplygeneratinginvestmentsforallv.Thefirst-orderconditionsforeachagentchoosingdvandsvtomaximizeprofitsare

p∂qa(v,dv,sv,p)/∂dv+[p−f−m]∂qb(v,dv,s\\v,z,p)/∂dv−1=0

11

andp∂qa(v,dv,sv,p)/∂sv+m∂qc(v,d\\v,sv,z,p)/∂sv−1=0.

(2)

Frominvestmentcomplementarities(Assumption2)thethreenon-zerocross-partialderiva-tivesare(weakly)positive,

∂2ψ(v,d,s,z)∂2qa(v,dv,sv,p)

=p≥0,

∂dv∂sv∂dv∂sv

∂2qb(v,dv,s\\v,z,p)∂2ψ(v,d,s,z)

=[p−f−m]≥0,

∂dv∂s\\v∂dv∂s\\v

and

∂2qc(v,d\\v,sv,z,p)∂2ψ(v,d,s,z)

=m≥0,

∂d\\v∂sv∂d\\v∂sv

sothatprofitsare(strictly)supermodularinthechoicevariables.SupermodularitymeansthatthereisatleastoneNashequilibrium,andifthisequilibriumisnotunique,thentheycanbeorderedfromsmallesttolargest,andweassumethelargestisobtained(seeTopkis1979;MilgromandRoberts,1990).Theequilibriumdemandgeneratinginvestmentsareafunctionoftheprice,transactionfeeandmargin,d(p,f,m).Theequilibriumsupplygener-atinginvestmentsareafunctionofthepriceandmarginonly,s(p,m).Becausethemarginalreturnstoeitherinvestmentareincreasinginv(Assumption1),fromsupermodularityourequilibriuminvestmentsareincreasinginagentpotential.Becausethedemandshavedimin-ishingmarginalreturnstoowninvestment,demandgeneratinginvestmentsaredecreasinginthetransactionfeeandthemargin,andsupplygeneratinginvestmentsareincreasinginthemargin.

InvestmentlevelsinthetraditionalmarketcanbecomparedtothoseobtainedintheEMusingthefirst-orderconditions(1)and(2),andinvestmentcomplementarities.Ifthemargintransferredfromthedemandgeneratingagenttothesupplyingagentispositive(m>0)andthepriceisgreaterthanthetransactionfeeplusthemargin(p>f+m),theninvestmentlevelsarealwayshigherintheEM.Referringtoourexample,oceanshippingisaveryold

12

industrythathistoricallyhassupportedprofitableagentsinanagent-intermediatedinstitu-tionalstructure,thatis,p>m.Recentregulatorychangeshavereinforcedconfidentialityofcontracts,furthermitigatingopencompetitionandenhancingtherolesofagents.Therefore,webelievep>f+missustainable.

4AgentParticipation

EachagenthasthechoiceofparticipatingintheEM,ornotparticipatingintheEMbutbeactiveinthetraditionalmarket,orinactive.Lemma1indicatesthebreakdown.

Lemma1:AllagentsthatareactiveinthetraditionalmarketparticipateintheEM.MoreagentsparticipateintheEMthanareactiveinthetraditionalmarket.

Proof:ComparingagentprofitsintheEMversusthetraditionalmarket,ψ(v,d,s,z)versusπ(v,dv,sv),andnotingthatinvestmentlevelsarealwayshigherintheEM,anagentcouldsetinvestmentstotheequilibriumlevelsfromthetraditionalmarketandbebetteroffintheEM.Definethesmallestactiveagentinthetraditionalmarketasthesmallestagentwithzeroprofits,vˇ=min{v|π(v,dv,sv)=0}.π(v,dv,sv)isincreasinginvsointhetraditionalmarketlargeragents,v>vˇ,areactive.

DefinethesmallestagentthatparticipatesintheEMasvˆ=min{v|ψ(v,d,s,z)=0}.Thismarginalagentmustsatisfytheindifferencecondition

pqa(ˆv,dvv,dvv),p)+ˆ(p,f,m),svˆ(p,m),p)+[p−f−m]qb(ˆˆ(p,f,m),s\\vˆ(p,m),z(ˆmqc(ˆv,d\\vv),p)−dvˆ),ˆ(p,m),z(ˆˆ(p,f,m)−svˆ(p,m)−K=0=µ(p,f,m,vˆ(p,f,m),sv

(3)

whereµ(p,f,m,vˆ)isanimplicitfunctiondefiningvˆ.Takingthepartialderivativewithrespecttovˆ,wecanusethefirst-orderconditionsin(2)toeliminatetheindirecteffectsofvˆthroughinvestmentingeneratingdemandandsupply(oralternatively,theenvelope

13

theorem).Thisyields

∂µ(p,f,m,vˆ)∂qa(·)∂qb(·)∂qb(·)dz(ˆv)∂qc(·)∂qc(·)dz(ˆv)

=p+[p−f−m][+]+m[+],(4)

∂vˆ∂vˆ∂vˆ∂zdvˆ∂vˆ∂zdvˆwheretheargumentsofqa(·),qb(·)andqc(·)includetheequilibriumlevelsofinvestmentasin(3).Usingthesingle-crossingconditionnecessarytoseparatethoseagentsthatparticipatefromthosethatdonot(Assumption3),equation(4)ispositive.Noticethatin(4)thedirecteffectofvˆispositiveandtheindirecteffectthroughzisnegative.However,includedintheindirecteffectisamarginalchangeinaproportion,dz(ˆv)/dvˆ,afraction.Moreover,thedirecteffectofbeinglargerislikelytooutweightheindirecteffectofagreaterproportionofagentsonthenetwork.Hence,Assumption3isreasonableinthiscontext.Asaconsequence,largeragents,v>vˆ,participateintheEM.BecauseoftheadditionalvalueprovidedbytheEM,ψ(ˇv,d,s,z(ˇv))>π(ˇv,dvˆ>vˇmeaningmoreagentsparticipateinˇ,svˇ)=0.ThereforevtheEM.Q.E.D.

Iftherewerefixedcostsofparticipationforagents,forexamplefixedITcosts,thentheEMcouldprovideafixedtransferaspartofitsincentivescheme.Thisimpliesnolossofgeneralityforourresults.MarginalITcostsforanyspecifictransactionwithinanEMareclosetozero.

UsingtheproofofLemma1,theproportionofagentsthatparticipateintheEMcanbewrittenas

z(ˆv)=

󰀂

vvˆ

h(v)dv(5)

whichisdecreasinginvˆ.From(3)theindifferentagentisafunctionoftheprice,thetransactionfeeandthemargin:vˆ(p,f,m).Thus,z(ˆv)≡z(p,f,m).

TodeterminetheoptimalsettingsofthetransactionfeeandthemarginitisnecessarytodeterminetheeffectsofchangesintheseEMchoicevariablesontheproportionofagentsthatparticipatein,andonthevolumeprocessedby,theEM.Toshortenournotationwesubsti-14

tutevˆ(·)forvˆ(p,f,m).Similarlyfortheproportionofagentsthatparticipate,z(p,f,m),weusez(·).Lemma2determinestheimpactofthetransactionfeeontheproportionofagentsthatparticipateintheEM.

Lemma2:TheproportionofagentsthatparticipateintheEMisdecreasinginthetrans-actionfee.

Proof:From(5)thechangesintheproportionofagentsthatparticipateintheEMasaresultofchangesinthetransactionfeeisdz(·)/df=−h(ˆv(·))[∂vˆ(·)/∂f].Theimpactofanincreaseinthetransactionfeethroughvˆ(·)canbedeterminedthroughtheimplicitfunctionruleusingtheindifferenceconditionµ(p,f,m,vˆ)=0from(3).Thatis,fromtheimplicit∂µ(p,f,m,vˆ)/∂vˆ

.

∂µ(p,f,m,vˆ)/∂f

Thesignofthenumeratorispositivefrom(4).Thedenominatoris

∂vˆ(·)/∂f=−

∂µ(p,f,m,vˆ)∂qc(·)∂d\\vˆ(·)

=−qb(·)+m<0

∂f∂d\\v∂fˆ

wheretheargumentsareasin(3).Thisresultsin∂vˆ(·)/∂f>0.Substitutingbackinindz(·)/dfresultsindz(·)/df<0.Q.E.D.

ChangesintheproportionofagentsthatparticipateintheEMasaresultofchangesinthemarginisdz(·)/dm=−h(ˆv(·))[∂vˆ(·)/∂m].Usingtheimplicitfunctiontheoremwith(3)fromtheproofofLemma1,wefind

∂vˆ(·)/∂m=−

∂µ(p,f,m,vˆ)/∂vˆ

,

∂µ(p,f,m,vˆ)/∂m

functiontheorm

andthereforethesignofdz(·)/dmfollowsthesignof

∂µ(p,f,m,vˆ)∂qb(·)∂s\\v∂qc(·)∂d\\vˆ(·)ˆ(·)

=−qb(·)+qc(·)+[p−f−m]+m

∂m∂s\\v∂m∂d\\v∂mˆˆ

(6)

through∂vˆ(·)/∂m.qb(·)andqc(·)arethedemandsoftheagentthatisindifferentbetweenparticipatingintheEMornot,andtheargumentsincludetheoptimalvaluefunctions

15

ofdemandandsupplygeneratinginvestmentasin(3),andtheproportionofagentsthatparticipateasin(5).

In(6)thesignofthefirsttwotermstogetherdependonwhethersmalleragentsprovidemoredemandormoresupply.Ifthesearebalancedorifdemandandsupplyaresensitivetoinvestment,thentheeffectofdemandandsupplygeneratinginvestmentsinthethirdandfourthtermsdeterminetheeffectofthemarginontheproportionofagentsthatparticipateintheEM.Inthethirdandfourthtermsdemandsareincreasinginbothinvestments.Inthethirdterm,supplygeneratinginvestmentisincreasinginthemarginfromtheanalysisfollowing(2)-thatis,anincreaseinthemargintransfersprofitstotheagentgeneratingthesupply.Inthefourthterm,demandgeneratinginvestmentisdecreasinginthemarginalsofromtheanalysisfollowing(2)-inotherwords,anincreaseinthemargintransfersprofitsawayfromtheagentgeneratingthedemand.

Iftheeffectsofchangesinthemarginoninvestmentsanddemandsaresmooth,thenstartingfromamarginofzero,m=0,(6)ispositivefromthethirdterm,andthereforedz(·)/dmispositive.So,atlowvaluesofthemargin,increasesinthemarginincreasetheproportionofagentsthatparticipateintheEM.However,asthemarginbecomeslarger,theimpactthroughinvestmentingeneratingdemand-thefourthtermin(6)becomeslarger,andthereforedz(·)/dmbecomesnegative.Consequently,athighvaluesofthemargin,increasesinthemargindecreasetheproportionofagentsthatparticipateintheEM.

Becausetheyareequivalent,eitherbetween-agentdemandcanbeusedtodefineEMvolume.TotalEMvolumecanbewrittenas

Qb(ˆv(·),d(p,f,m),s(p,m),p)=

󰀂

vvˆ(·)

qb(v,dv(p,f,m),s\\v(p,m),z(ˆv(·)),p)h(v)dv.

WesimplifythenotationbyusingQb(·)inplaceofQb(ˆv(·),d(p,f,m),s(p,m),p).Inaddition,weusedv(·)inplaceofdv(p,f,m),andsv(·)inplaceofsv(p,m).Lemma3providesournextresult.

16

Lemma3:EMvolumeisdecreasinginthetransactionfee.Proof:TheeffectofachangeinthetransactionfeeonEMvolumeis

∂Qb(·)∂Qb(·)∂Qb(·)dz(ˆv(·))∂vˆ(·)∂Qb(·)∂dv(·)

=++.∂f∂vˆ∂zdvˆ∂f∂dv(·)∂f

FromLeibnitz’srulethefirstterminsquarebracketsisnegative.Fromthedefinitionoftotalvolume,thefirsttermintheproductinsquarebracketsispositive,andfrom(5)thesecondtermintheproductisnegative.Thus,thecontentofthesquarebracketisnegative.FromtheproofofLemma2,∂vˆ(·)/∂f>0sotheterminvolvingsquarebracketsisnegative.Usingthedefinitionoftotalvolumeandthefactthatdemandgeneratinginvestmentisdecreasinginf,thelastproductisalsonegative.Q.E.D.

IncreasesinthetransactionfeetransfersprofitsfromagentstotheEM.Thismakeseachparticipatingagentlessprofitableandreducestheagents’investmentincentives.Theeffectofachangeinthemarginis

∂Qb(·)∂Qb(·)dz(ˆv(·))∂vˆ(·)∂Qb(·)∂s\\v(·)∂Qb(·)∂dv(·)∂Qb(·)

=[+]++.∂m∂vˆ∂zdvˆ∂m∂s\\v(·)∂m∂dv(·)∂m

󰀃

󰀁

󰀃

󰀁

(7)

Theterminsidethelargesquarebracketstakesthesignofdz(·)/dm,theeffectofthemarginontheproportionofagentsthatparticipateintheEM.Thus,thistermcapturestheeffectsontotalvolumeofchangesinthemarginthroughtheproportionofagentsthatparticipateintheEM.Referringtotheanalysisfollowing(6),atlowvaluesofthemargin,increasesinthemarginincreasetheproportionofagentsthatparticipateintheEM,andathighvaluesofthemargin,increasesinthemargindecreasetheproportionofagentsthatparticipateintheEM.

Similarto(6),theimpactofthemarginthroughsupplygeneratinginvestments,thenexttermin(7),ispositive,andtheimpactofthemarginthroughdemandgeneratinginvestment,thethirdtermin(7),isnegative.Thetwotermscapturetheeffectsontotal

17

volumeofchangesinthemarginthoughdemandandsupplygeneratinginvestment.Toexploretherelativemagnitudesofthesetwoimpactswecanusetheimplicitfunctionruleoneachequationin(2)togetherwithconcavityoftheagent’sprofitfunctionineachinvestmenttowriteouttheeffectofachangeinthemarginondemandandsupplygeneratinginvestment:

∂qb(v,dv,s\\v,z,p)/∂dv∂dv(p,f,m)

=∂2q(v,d,s,p)<0∂2qb(v,dv,s\\v,z,p)avv∂mp+[p−f−m]∂d2∂d2v

v

−∂qc(v,d\\v,sv,z,p)/∂sv∂sv(p,m)

=∂2q(v,d,s,p)>0.∂2qc(v,d\\v,sv,z,p)avv∂mp+m∂s2∂s2v

v

Focusingonthedenominatorsofthetwoeffectsabove,atlowvaluesofthemargin,thepositiveeffectofachangeinthemarginonsupplygeneratinginvestmentislargerelativetothenegativeeffectondemandgeneratinginvestment,andviceversaathighvaluesofthemargin.

FortheEM,maximizingprofitmeansmaximizingtheproductofthetransactionfeeandtotalvolumechoosingthelevelsofthetransactionfeeandthemargin:

maxΓ(f,m)=max{fQb(·)}.

f,m

f,m

Thefirst-orderconditionsformaximizingprofitare

∂Qb(·)∂Γ(f,m)∂Γ(f,m)∂Qb(·)

=Qb(·)+f=0and=f=0.∂f∂f∂m∂m

(8)

Aninteriorsolutionrequires∂Qb(·)/∂mbeincreasingatlowvaluesofthemarginandbedecreasingathighvaluesofthemargin.Thisislikelyfromthetwoseparateeffectsin(7)describedabove:atlowvaluesofthemargin,achangeinthemarginincreasestheeffectontotalvolumethroughincreasedparticipationintheEMandthroughagreaterrelativeimpactonsupplygeneratinginvestmentcomparedtodemandgeneratinginvestment.AthighlevelsofthemarginachangeinthemargindecreasestheeffectontotalvolumethroughreducedparticipationintheEMandthroughgreaterdemandgeneratinginvestmentrelativetosupply

18

generatinginvestment.Thus,EMprofitsaremaximizedbylimitingagentparticipation.Thiskeyresultisourfirsttheorem,statedwithoutproof.

Theorem1:(LimitedMarketParticipation)Iftotalvolumeisincreasinginthemarginatlowvaluesofthemarginandisdecreasinginthemarginathighvaluesofthemargin,thenEMprofitsaremaximizedbylimitingagentparticipation.

TheEM’soptimizationproblemdoesnotmaximizetheproportionofagentsthatpartici-pate.Thisisbecausechangesinthemarginhaveoppositeeffectsonequilibriuminvestmentsingeneratingdemandandingeneratingsupply.Anincreaseinthemarginreducesthein-centivestoinvestingeneratingdemandandincreasesincentivesingeneratingsupply.ThisimpactsnotonlythedirecteffectsofdemandandsupplygeneratinginvestmentonEMvolume,butalsotheindirecteffectoftheproportionofagentsthatparticipateintheEM.Therefore,theEMimprovesservicenotonlybyprovidingamatchingmechanismforsupplyanddemand,butalsobyprovidingtherightcombinationofinvestmentincentivestobal-ancethegenerationofsupplyanddemand.LimitingagentparticipationiscriticalfortheEM’sprofitmaximizingwhendemandinvestmentsareimportant.Inthiscase,themarginissetlowtoincentinvestmentsfromlargeragents,andsmalleragentsmaynotmakeenoughthroughtheEMtocovertheiroperatingcosts.

OurtheoremisrelatedtoresultsobtainedbyBankeret.al.(2000)andbyBakosandBrynjolfsson(1993).Theirresults,wherebyinabuyer-supplierITframeworkitcouldbeoptimalforbuyerstolimitparticipationbysuppliers,derivefromtheneedtoprovidesupplierswithanadequateincentivetoinvest.Intheirworktheincentivewasnecessarytoinducesupplierstoinvestinnon-contractibleITassets.InourmodeltheEMmustprovideinvestmentincentivesforagentstogeneratedemandandsupplybecauseeachagent’sexpertiseisdifferentandunobservabletotheEM.

19

5

5.1

EndogenousPriceandWelfare

ImpactonthePrice

SofarwehaveconsideredthepriceasfixedbyconditionsoutsideoftheEMsuchthatbothindividualagentsandtheindustryareprice-takers.Althoughagentsmaynotbeabletoinfluencethepriceindividually,theiraggregateinvestmentsingeneratingdemandandsupplymayimpactthepricebecausetheymaychangethebalancebetweendemandandsupply.Endogenizingpricemeansincorporatingthedifferencebetweeninvestmentsingeneratingdemandandinvestmentsingeneratingsupply.

Letthebalancebetweendemandandsupplygeneratinginvestmentberepresentedbythevectorofdifferencesininvestment:󰀗δ(f,m)=d(p,f,m)−s(p,m).Recallthatinvest-mentsingeneratingdemandaredecreasinginthetransactionfeeandthemargin,fandm,andinvestmentsingeneratingsupplyareincreasinginthemargin,m.Directlyfromthepropertiesoftheseinvestmentfunctions,eachvectorelementof󰀗δ(f,m)isdecreasinginthetransactionfeeandthemargin.

Letρ(f,m)bethepricethataggregatestheeffectsofthetransactionfeeandthemarginfrom󰀗δ(f,m).Thepriceρ(f,m)capturestheeffectsofthetransactionfeeandthemarginonthebalancebetweendemandandsupply.Thepricemusthavethesamepropertiesas󰀗δ(f,m),namelyρ(f,m)isdecreasinginthetransactionfeebecausetheincreasesinthetransactionfeereducesdemandgeneratinginvestment.Similarly,thepriceρ(f,m)isdecreasinginthemarginbecauseincreasesinthemarginreducedemandgeneratinginvestmentandincreasesupplygeneratinginvestment.

Usingthisformofpricemakespriceendogenous.Thatis,pricedependsontheEM’schoicevariablesfandm.Wecanrewritetotalvolumebyreplacingpwithρ(f,m).Tosimplythenotationweletthedemandgeneratinginvestmentvectord(ρ(f,m),f,m)be

20

writtenasd(·),thesupplygeneratinginvestmentvectors(ρ(f,m),m)bewrittenass(·),agentv’sdemandgeneratinginvestmentdv(ρ(f,m),f,m)bewrittenasdv(·),andv’ssupplygeneratinginvestments\\v(ρ(f,m),m)bewrittenass\\v(·).Totalvolumewithendogenizedpriceisthen

˜b(ˆQv(·),d(·),s(·),ρ(f,m))=

󰀂

vvˆ(·)

qb(v,dv(·),s\\v(·),z(ˆv(·)),ρ(f,m))h(v)dv.

Withtotalvolumesodefined,wecanprovideourlasttheoremfollowedbyacorollarythatdescribestheresultingeffectsonprice.

Theorem2:(EndogenousPrice)Ifthetotaleffectofanincreaseinthepriceontotalvolumeispositive,thentheoptimaltransactionfeeandmarginarelower.Otherwise,theoptimaltransactionfeeandmarginarehigher.

Corollary:Ifthetotaleffectofanincreaseinpriceontotalvolumeispositive,thenpriceishigher.Otherwise,priceislower.

˜b(·)inplaceofQ˜b(ˆProofofTheorem2:WesimplifythenotationusingQv(·),d(·),s(·),ρ(f,m)).Itisnotpossibletocomparetotalvolumewiththeendogenouspriceandwiththepricede-˜b(·)andQb(·),withoutafixedpointwheretheendogenouspriceterminedoutsidetheEM,Q

ρ(f0,m0)=pforthepair(f0,m0).However,theinterioroptimalsolutionformaxf,mΓ(f,m),profitswhenthepriceisdeterminedoutsidetheEM,providessuchafixedpoint.Weusethistocomparetheoptimaltransactionfeeandmargin.

Letf∗andm∗betheinterioroptimalsolutiontomaxf,mΓ(f,m)=fQb(·),theEMprofitswhenpriceisdeterminedoutsidetheEM.Usethisoptimumtofixthepointp=ρ(f∗,m∗)wherethepricesareequal,soatthatpointEMvolumeinthetwocasesisequal˜b(·)at(f∗,m∗).Thenecessaryfirst-orderconditionsfortheinterioroptimumwithQb(·)=Q

thedeterminedoutsidetheEMpriceareasin(8),yielding(f∗,m∗).

21

˜f,m).Evaluatedat(f∗,m∗)wehaveDenoteprofitswhenpriceisendogenousasΓ(

˜f,m)˜b(·)˜b(·)∂ρ(f∗,m∗)∂Γ(dQ∂Q(·)dQb∗∗˜b(·)+f˜b(·)+f[=Q=Q+]∗∗∂fdf∂f∗dρ∂f∗

and

˜f,m)˜˜b(·)∂ρ(f∗,m∗)∂Γ(dQ∗dQb(·)∗∂Qb(·)=f=f[+],

∂m∗dm∗∂m∗dρ∂m∗

where∂Qb(·)/∂f∗and∂Qb(·)/∂m∗representthedirecteffectsofchangesinthetransactionfeeandthemarginlikethosein(8).Theremainingtermsaretheindirecteffectsofthetransactionfeeandthemarginthroughprice.Ifthetotaleffectofanincreaseinpriceon

˜b(·)/dρ>0,thenboth∂Γ(˜f,m)/∂f∗and∂Γ(˜f,m)/∂m∗aretotalvolumeispositive,dQnegative.

Thesecond-orderconditionsforaninterioroptimumrequireconcavityineachvariable.

˜f,m)/∂f∗and∂Γ(˜f,m)/∂m∗arenegative,thenthedecisionvariablesThus,ifboth∂Γ(

havebeensettoohighandoptimizingprofitswhenpriceisendogenouslydetermined(asrepresentedbythefirst-orderconditions)requiresalowerf∗andm∗.Proofoftheconverseisstraightforward.Q.E.D.

Theorem2anditscorollaryembodyoursecondkeyresult:theagent-intermediatedEMcancausebothvolumeandpricetoincreasesimultaneously.Theorem2anditscorollarydependontheimpactofthepriceontotalvolume,andthusonindividualagentdemands.

˜b(·),achangeinpricehasfourseparateeffects.ThefirstisanReferringtotheargumentsofQ

indeterminateindirecteffectthroughagentparticipation-similartobutmorecomplexthantheonein(6).Thesecondisthenegativeindirecteffectthroughinvestmentsingeneratingdemand.Thethirdisthepositiveindirecteffectthroughinvestmentsingeneratingsupply.Andfinally,thereisanegativedirecteffectofpriceonvolume.

Anincreasedpriceisnormallyassociatedwithdecreasedvolume.However,hereanin-creasedpricehasapositiveindirecteffectthroughsupply-generatinginvestmentandpossibly

22

apositiveindirecteffectthroughparticipation,whichcansimultaneouslyresultinincreasedvolume.Thatis,ifthedemand-supplybalanceissuchthatidentifyingandbringingaddi-tionalsupplytothemarketcanincreasethenumberoftransactions,thenanincreaseinpricecanmotivateagentstobringthatadditionalsupplytomarket.Duetotheconflictingsetofdirecteffectsandindirecteffectsthroughdemandandsupplygeneratinginvestments,itisnotpossibletodetermineingeneralifendogenizingpriceincreasesordecreasesprofitsfortheEMandfortheindividualagents.

5.2EffectsonWelfare

WhenpriceisdeterminedbyconditionsoutsidetheEM,theeffectsoftheEMonwelfarearepositive.Fourgroupsofstakeholdersmustbeaccountedfor:buyers,sellers,agents,andtheEM-whereinoceanshippingthefirsttwowouldbeshippersandcarriersrespectively.IfthefixedcostsofagentsandtheEMarecovered,Qb(·)>0impliesthateachofthegroupscompletetransactionstheyotherwisewouldnot-atthesameprice-andarethereforebetteroff.Thus,theEMisnotonlywelfareenhancing,itisParetooptimalforallstakeholders.AlthoughthisshouldbetruegiventhattheEMeffectivelyrelaxesaconstraintintheopti-mizationprogramforthemarket,whatisimportantisthewayinwhicheachstakeholderbenefits.

Whenpriceisendogenous-affectedbyagentinvestments-theeffectsoftheEMonwelfarearelessclear.Inadditiontothedirecteffectsofpriceondemands,changesinpricealsohaveinvestmentimpactsonvolumethroughchangesinthetransactionfeeandthemargin.However,aswesawinTheorem2,thedirectionofthesechanges-andhenceincentivesforinvestment-dependonthesignofthedirecteffectofachangeinpriceontotalvolume.

Evenignoringtheinvestmenteffects,theoutcomeisequivocal.Consider,forexample,

23

whenthepresenceoftheEMincreasesprice.BuyersbenefitfromcompletingtransactionsthroughtheEMthatwouldnothavebeencompletedotherwise.Nonetheless,thosetransac-tionsthatwouldhavebeencompletedwithouttheEMnowfaceahigherprice.Hence,theaggregateimpactonthewelfareofbuyersismixed.Sellersfaceasimilartradeoff.VolumeisincreasedbythosetransactionsthatoccurthroughtheEM.Unfortunately,transactionsnotrequiringtheEMfaceanincreasedprice,reducingtheoff-EMvolumeforsellerscomparedtothetraditionalmarket.ThisreductionmaymorethanoffsetthenewvolumefromtheEM.Similarly,agentsthatparticipateintheEMgainfromthetransactionsontheEM,butloseoff-EMvolumecomparedtothetraditionalmarket.ThoseagentsthatdonotparticipateintheEMsimplylosevolumeandthelowerprofitsthatresultmaynotbeoffsetbytheincreasedprice.

Alternatively,whenpresenceoftheEMdecreasesprice,buyersgainfromcompletingtransactionsthroughtheEMthattheywouldnotcompleteotherwise,andalsogainfromlowerpricesonthosetransactionscompletedwithouttheEM.SellersalsobenefitfromtheEMtransactionsandhighervolumesascomparedtothetraditionalmarketfromthelowerprice.Forbothbuyersandsellers,however,thesegainsmaybenegatedbyareducedincentivefordemandandsupplygeneratinginvestment.Theimpactonagentsismixed:thosethatparticipateintheEMmaygainfromtransactionsovertheEM,butthosethatdonotparticipatemaybelessprofitablebecauseofthereducedprice-evenathighervolumes.Solongasitsfixedcostsarecovered,theagent-intermediatedEMisprofitable.MakingtheEMprofitableiscriticaltomotivatetheneededinvestmentinITinfrastructuretosup-portthedevelopmentofthecentralizedrepositoryfordemandandsupplyinformation.Inparticular,thisinvestmentcanbecostlysincethespecialinformationrequirementsforthismarket-thespecificityandperishabilityofdemandandsupply-aremoreextensivethanforfinancialorcommoditymarkets.

24

6Conclusion

Wehaveshownthatwithpricedeterminedbythegeneralfreighttransportationmarketplace,anagent-intermediatedEMincreasesagentparticipationandinvestment.TheimportanceofagentinvestmentishighlightedbytheEM’sallocationofprofitsbetweendemandandsupplyprovidingagentsonagiventransactionwherebytheEMbalancesparticipationandinvestmenteffectsinitsallocation,consequentlyrestrictingagentparticipationinthemarket.Whenpriceisendogenous-influencedbythedemandandsupplybalanceintheEM,wefoundthatifpriceincreasessupplygeneratinginvestmentandparticipation,thenpriceandvolumeontheEMcanincreasesimultaneously.Finally,sinceanendogenouspricecanbehigherorlowerwithanagent-intermediatedEM,theEMmaynotbewelfareenhancingforallthestakeholders.

Oneimportantmanagerialimplicationforagentsisthattheagent-intermediatedEMchangesthereturnstodifferentinvestments.Inthetraditionalmarketanagentmustinvestinmatchingitsowndemandandsupply-forexample,onceashipment(demand)issecured,theagentmustinvestinfindingcapacity(supply)forthatshipmentinordertoobtainareturnoninvestment.InanEManagentcanobtainanimmediatereturnoninvestmentfromgeneratingasinglesideofthetransaction-forexample,securingashipmentwhosecapacityisprovidedbyanotheragentthroughtheEM.OneconsequenceofthechangeinreturnstoinvestmentbroughtaboutbytheEMisthatagentsarerewardedforspecialization,wherebyagentsmayfocusonaparticularsideofthetransactionandaspecificnicheinthatsideofthetransaction.Forexample,anagentusinganEMmayfocusitsinvestmentonshippersthatshipautomobilestotheUSfromEuroperatherthanhavingtosplititsinvestmentbetweenshippersandcarriers.Thisincreasesthedifferentiationamongagents,providingthemwithhigherprofits.Competitionamongagentsultimatelyreappearsthroughprice-thepricethatisdeterminedviachangesinthedemandandsupplybalancesintheEM.

25

Withanagent-intermediatedEMindividualagentswilleventuallybeforcedoutofus-ingthetraditionalmarketexclusivelybecausethechangeinthestructureofthereturnstoagentinvestmentsbroughtaboutbytheEMwillmakeagentsthatjointheEMmoreprofitablethroughanincreaseintransactionvolume.ThiswillincreasetheliquidityofthefreighttransportationmarketplaceforthoseagentsontheEM.Indeed,theEMmayincreasepriceandtransactionvolumesimultaneously.Thus,participatingintheEMwillbecomeastrategicnecessity.

FortheEMthecriticalissueisthebalancebetweenprovidingincentivesforagentin-vestmentandopeningthemarkettobroadagentparticipation.TheEMeffectivelybalancesprofitsfrominvestmentsbylargeragentswithincreasedvolumefromgreaterparticipation-thesettingofthemargindeterminesinvestmentsbyagentsandwhichagentsareprofitablebyparticipatingintheEM.Therefore,calibratingthelevelofthemargindividedbetweenagentsisfundamentaltotheperformanceoftheEM.

7Acknowledgements

HelpfulsuggestionshavebeenprovidedbyseminarparticipantsatPennStateUniversity,MichiganStateUniversity,TheOhioStateUniversity,UniversityofSouthernCalifornia,UniversityofWisconsin-Madison,UniversityofOregon,UniversityofPittsburgh,andpar-ticipantsatWISE.WeespeciallythankCliveWrigleyandJohnKingfordiscussionsaboutfreighttransportation.WealsothanktheNationalScienceFoundation,theNaturalSci-enceandEngineeringResearchCouncilofCanada,andtheSocialScienceandHumanitiesResearchCouncilofCanadaforgeneroussupport.

26

8References

Alt,R.,P.Forster,andJ.L.King.1997.TheGreatReversal:InformationandTrans-portationInfrastructureintheIntermodalVision.ProceedingsoftheNationalConferenceonSettinganIntermodalTransportationResearchFramework,NationalAcademyPress,WashingtonD.C.,31-53.

Ba,S.,Whinston,A.B.,andH.Zhang.1999.BuildingTrustintheElectronicMarketThroughanEconomicIncentiveMechanism.ProceedingsoftheTwentiethInternationalConferenceonInformationSystems,De,P.andDeGross,J.I.(eds.),Charlotte,NorthCar-olina,December.

Bailey,J.P.,andY.Bakos.1997.AnExploratoryStudyoftheEmergingRoleofElectronicIntermediaries.InternationalJournalofElectronicCommerce,1,3(Spring),7-20.Bakos,J.Y.1998.TheEmergingRoleofElectronicMarketplacesontheInternet.Commu-nicationsoftheACM,41,8(August),35-42.

Bakos,J.Y.,andE.Brynjolfsson.1993.FromVendorstoPartners:InformationTechnol-ogyandIncompleteContractsinBuyer-SupplierRelationships.JournalofOrganizationalComputing,3,3,301-328.

Banker,R.D.,Kalvenes,J.,andR.A.Patterson.2000.InformationTechnology,ContractCompleteness,andBuyer-SupplierRelationships,ProceedingsoftheTwenty-FirstInterna-tionalConferenceonInformationSystems,De,P.andDeGross,J.I.(eds.),Brisbane,Aus-tralia,December,218-228.

Chircu,A.M.,Davis,G.B.,andR.J.Kauffman.2000.TheRoleofTrustandExpertiseintheAdoptionofElectronicCommerceIntermediaries.MISRCWorkingPaper,March2000.Clemons,E.K.,andB.W.Weber.1996.AlternativeSecuritiesTradingSystems:Testsand

27

RegulatoryImplicationsoftheAdoptionofTechnology.InformationSystemsResearch,7,2(June),163-188.

DATServices.1999.http://www.dat.com.

Forster,P.,andJ.L.King.1995.InformationInfrastructureStandardsinHeterogeneousSectors:LessonsfromtheWorldwideAirCargoCommunity.InStandardsPolicyforInfor-mationInfrastructure(B.KahinandJ.Abbateeds.).MITPress,Cambridge,MA.Gellman,R.1996.DisintermediationandtheInternet.GovernmentInformationQuarterly,13,1,1-8.

GTNexus.2001.http://www.tradient.com/about.

Halper,M.1997.MeettheNewMiddlemen.Computerworld,31,17(April28),C10-C14.Hayek,F.A.1945.TheUseofKnowledgeinSociety.AmericanEconomicReview,35,4(September),519-530.

Lee,H.G.,1998.DoElectronicMarketplacesLowerthePriceofGoods?CommunicationsoftheACM,41,1(January),73-80.

Hoffman,T.1995.NoMoreMiddlemen.Computerworld,July17,55.

Lewis,I.,andD.B.Vellenga.2000.TheOceanShippingReformActof1998.TransportationJournal,39,4(Summer),27-34.

Lynch,K.2000.The7ImmutableLawsofCollaborativeLogistics,WorldTrade,13,10(October),86-90.

Milgrom,P.andJ.Roberts.1990.Rationalizability,Learning,andEquilibriuminGameswithStrategicComplementarities.Econometrica,58,6(November),1255-1277.NTE.1998.http://www.nte.net/PressRelease/28.html.

28

Riggins,F.J.,Kriebel,C.H.,andT.Mukhopadhyay.1994.TheGrowthofInterorgani-zationalSystemsinthePresenceofNetworkExternalities.ManagementScience,40,8(August),984-998.

Shashikumar,N.,andG.L.Schatz.2000.TheImpactofU.S.RegulatoryChangesonInternationalIntermodalMovements.TransportationJournal,40,1(Fall),5-14.

Sherwood,C.S.,andR.Bruns.1992.SolvingInternationalTransportationProblems.Re-viewofBusiness,14,1(Summer/Fall),25-30.

Topkis,D.1979.EquilibriumPointsinNonzero-Sumn-PersonSubmodularGames.SiamJournalofControlandOptimization,17,6(November),773-787.

Wang,E.T.G.,andA.Seidmann.1995.ElectronicDataInterchange:CompetitiveExternal-itiesandStrategicImplementationPolicies.ManagementScience,41,3(March),401-418.Wigand,R.,andR.Benjamin.1996.ElectronicCommerce:Definition,TheoryandContext.TheInformationSociety,13,1,1-16.

Williamson,O.E.1981.TheEconomicsofOrganization:TheTransactionCostApproach.AmericanJournalofSociology,87,3,548-577.

Williamson,O.E.1985.TheEconomicInstitutionsofCapitalism::Firms,Markets,Rela-tionalContracting.NewYork:FreePress.

29