8、Gravity-with-Gravitas：A-Solution-to-the-Border-P

AmericanEconomicAssociationGravitywithGravitas:ASolutiontotheBorderPuzzleAuthor(s):JamesE.AndersonandEricvanWincoopSource:TheAmericanEconomicReview,Vol.93,No.1(Mar.,2003),pp.170-192Publishedby:AmericanEconomicAssociationStableURL::10/10/201407:40YouruseoftheJSTORarchiveindicatesyouracceptanceoftheTerms&ConditionsofUse,[e-mail]..AmericanEconomicAssociationiscollaboratingwithJSTORtodigitize,preserveandextendaccesstoTheAmericanEconomicReview.:40:39AMAllusesubjecttoJSTORTermsandConditionsGravitywithGravitas:ASolutiontotheBorderPuzzleByJAMESE.ANDERSONANDERICVANWINCOOP*Gravityequationshavebeenwidelyusedtoinfertradefloweffectsofvariousinstitutionalarrangements.Weshowthatestimatedgravityequationsdonothaveatheoreticalfoundation.Thisimpliesboththatestimationsuffersfromomittedvariablesbiasandthatcomparativestaticsanalysisisunfounded.Wedevelopamethodthat(i)consistentlyandefficientlyestimatesatheoreticalgravityequationand(ii)correctlycalculatesthecomparativestaticsoftradefrictions.WeapplythemethodtosolvethefamousMcCallumborderpuzzle.Applyingourmethod,wefindthatnationalbordersreducetradebetweenindustrializedcountriesbymoderateamountsof20-50percent.(JELF10,F15)Thegravityequationisoneofthemostem-piricallysuccessfulineconomics.Itrelatesbi-lateraltradeflowstoGDP,distance,andotherfactorsthataffecttradebarriers.Ithasbeenwidelyusedtoinfertradefloweffectsofinsti-tutionssuchascustomsunions,exchange-ratemechanisms,ethnicties,linguisticidentity,andinternationalborders.Contrarytowhatisoftenstated,theempiricalgravityequationsdonothaveatheoreticalfoundation.Thetheory,firstdevelopedbyAnderson(1979),tellsusthataftercontrollingforsize,tradebetweentworegionsisdecreasingintheirbilateraltradebar-rierrelativetotheaveragebarrierofthetworegionstotradewithalltheirpartners.Intu-itively,themoreresistanttotradewithallothersaregionis,themoreitispushedtotradewithagivenbilateralpartner.Wewillrefertothetheoreticallyappropriateaveragetradebarrierasmultilateralresistance.Theempiricalgrav-ityliteratureeitherdoesnotincludeanyformofmultilateralresistanceintheanalysisorin-*Anderson:DepartmentofEconomics,BostonCollege,ChestnutHill,MA02467(e-mail:[e-mail]),andNBER;vanWincoop:DepartmentofEconomics,UniversityofVirginia,116RoussHall,Charlottesville,VA22904(e-mail:[e-mail]),andNBER.Wewouldliketothanktworeferees,CarolynEvans,RobertFeenstra,JimHarrigan,JohnHelliwell,RussellHillberry,DavidHummels,AndyRose,andKei-MuYiforhelpfulcomments.WealsothankseminarparticipantsatBostonCollege,BrandeisUniversity,HarvardUniversity,Prince-tonUniversity,TilburgUniversity,theUniversityofCali-forniaatDavis,theUniversityofColorado,theUniversityofNorthCarolina,theUniversityofVirginia,theUniversityofWisconsin,VanderbiltUniversity,andthe2000NBERITIFallmeetingforhelpfulcomments.cludesanatheoreticremotenessvariablere-latedtodistancetoallbilateralpartners.Theremotenessindexdoesnotcaptureanyoftheothertradebarriersthatarethefocusoftheanalysis.Moreover,evenifdistanceweretheonlybilateralbarrier,itsfunctionalformintheremotenessindexisatoddswiththetheory.'Thelackoftheoreticalfoundationofempir-icalgravityequationshastwoimportantimpli-cations.First,estimationresultsarebiasedduetoomittedvariables.Second,andperhapsevenmoreimportant,onecannotconductcom-parativestaticsexercises,eventhoughthisisgenerallythepurposeofestimatinggravityequations.2Inordertoconductacomparativestaticsexercise,suchasaskingwhattheeffectsareofremovingcertaintradebarriers,onehastobeabletosolvethegeneral-equilibriummodelbeforeandaftertheremovaloftradebarriers.Inthispaperwewill(i)developamethodthatconsistentlyandefficientlyesti-matesatheoreticalgravityequation,(ii)usetheestimatedgeneral-equilibriumgravitymodeltoconductcomparativestaticsexercisesoftheef-'JeffreyH.Bergstrand(1985,1989)acknowledgesthemultilateralresistancetermanddealswithitstime-seriesimplications,butisunabletodealwiththecross-sectionaspectswhicharecrucialforpropertreatmentofbilateraltradebarriers.AndersonandDouglasMarcouiller(2002)useaTomqvistapproximationtothemultilateralresistancetermwhichhandlesthecross-sectionvariationofbilateralbarriers.2Recently,someauthors(e.g.,DavidHummels,1999)controlformultilateralresistanceinestimationwithfixedeffects,butcannotconsistentlydocomparativestaticsonthisbasis.170Thiscontentdownloadedfrom123.124.19.3onFri,10Oct201407:40:39AMAllusesubjecttoJSTORTermsandConditionsANDERSONANDVANWINCOOP:GRAVITYWITHGRAVITASfectoftradebarriersontradeflows,and(iii)applythetheoreticalgravitymodeltoresolvetheborderpuzzle.Oneofthemostcelebratedinferencesfromthegravityliter