日本汽車出口交易水平分析-2024-12-市場(chǎng)解讀

InvoiceCurrencyChoiceinIntra-FirmTrade:

ATransaction-LevelAnalysisofJapaneseAutomobileExports

FacultyofEconomics,ChuoUniversity,Japan/PolicyResearchInstitute,MinistryofFinance,Japan

TaiyoYOSHIMI

PolicyResearchInstitute,MinistryofFinance,Japan

UrakuYOSHIMOTO

FacultyofInternationalSocialSciences,YokohamaNationalUniversity,

Japan/PolicyResearchInstitute,MinistryofFinance,Japan

KiyotakaSATO

SchoolofInternationalRelationsandPublicPolicy,Columbia

University,TheUnitedStates/NationalGraduateInstituteforPolicy

Studies,Japan

TakatoshiITO

FacultyofEconomics,GakushuinUniversity,Japan/PolicyResearch

Institute,MinistryofFinance,Japan

JunkoSHIMIZU

FacultyofEconomics,ShigaUniversity,Japan/PolicyResearch

Institute,MinistryofFinance,Japan

YushiYOSHIDA

PRIDiscussionPaperSeries(No.23A-03)

Theviewsexpressedinthispaperarethoseofthe

authorsandnotthoseoftheMinistryofFinanceor

thePolicyResearchInstitute.

ResearchDepartmentPolicyResearchInstitute,MOF

3-1-1Kasumigaseki,Chiyoda-ku,Tokyo100-8940,Japan

TEL03-3581-4111

1

InvoiceCurrencyChoiceinIntra-FirmTrade:

ATransaction-LevelAnalysisofJapaneseAutomobileExports*1,*2

TaiyoYOSHIMI*3UrakuYOSHIMOTO*4 KiyotakaSATO*5 TakatoshiITO*6 JunkoSHIMIZU*7YushiYOSHIDA*8

Abstract

Thisstudyempiricallyinvestigatestheextenttowhichthechoiceofinvoicecurrencydiffersbetweenintra-firmandarm’slengthexports.Wealsoexaminewhetherotherfirm-andproduct-levelfactorsaffectthechoiceofinvoicecurrency.Thisstudyisthefirsttobegrantedaccesstohighlydisaggregatedtransaction-leveldataonJapaneseautomobileexportstoFrance.Byconductingpanellogitestimation,wedemonstratethatimporters’currencytendstobechoseninintra-firmexportinvoicing,whichhasnotbeenrigorouslyshowninpreviousliterature.Ourempiricalfindingsremainrobustwhenintroducingdifferenttypesofintra-firmexportvariablesandotherconventionalexplanatoryvariablessuchasfirm-andproductmarketshare,exchangeratevolatility,adummyforintermediategoodsexports,euro-invoicedimports,laborproductivity,andresearchanddevelopmentintensity.Amidgrowingintra-firmtradeandexpandingglobalvaluechains,Japaneseparentfirmstendtoinvoiceusingtheimporter’scurrency,assumingtheforeignexchangeriskthatarisesfromintra-firmtrade;thus,exchangerateriskmanagementisasignificantconsiderationforJapaneseparentfirms.

Keywords:Invoicecurrency;Intra-firmtrade;JapanCustomsdata;Marketshare;Exportcompetitiveness

JELclassification:F14;F31

*1ThisstudyistheresultofresearchconductedjointlywiththePolicyResearchInstitute(PRI)oftheJapaneseMinistryofFinance(MOF),whichsubmittedanoffertotheMOFbasedonthe“GuidelineontheutilizationofCustoms’importandexportdeclarationdatainajointresearchwiththePolicyResearchInstitute,”andreceivedapprovalinFebruary2022.Theviewsexpressedinthisresearcharetheauthors’personalresponsibilityanddonotrepresenttheofficialviewsoftheMOForPRI.WewouldliketothankFumiharuIto,KentaAndo,ShintaroNegishi,andtheseminarparticipantsatPRI.WeacknowledgefinancialsupportfromtheJSPSKAKENHIGrant(JP20H01518and20KK0289toYoshimi,JP21K20172toYoshimoto,JP19H01504,23H00836,and23K17550toSato,JP18K01698toShimizu,andJP22K18527toYoshida).Allsubsequenterrorsareourown.

*2Correspondingauthor:TaiyoYoshimi;Address:FacultyofEconomics,ChuoUniversity,742-1Higashinakano,Hachioji-shi,Tokyo,192-0393,Japan.Tel:+81-42-674-3424;Fax:+81-42-674-3425;E-mail:yoshimi@tamacc.chuo-u.ac.jp.

*3FacultyofEconomics,ChuoUniversity,Japan/PolicyResearchInstitute,MinistryofFinance,Japan*4PolicyResearchInstitute,MinistryofFinance,Japan

*5FacultyofInternationalSocialSciences,YokohamaNationalUniversity,Japan/PolicyResearchInstitute,MinistryofFinance,Japan*6SchoolofInternationalRelationsandPublicPolicy,ColumbiaUniversity,TheUnitedStates/

NationalGraduateInstituteforPolicyStudies,Japan

*7FacultyofEconomics,GakushuinUniversity,Japan/PolicyResearchInstitute,MinistryofFinance,Japan*8FacultyofEconomics,ShigaUniversity,Japan/PolicyResearchInstitute,MinistryofFinance,Japan

2

1.Introduction

Thechoiceofinvoicecurrencyhasattractedconsiderableattentioninthefieldofinternationaleconomics.AsdemonstratedbyObstfeldandRogoff(2000),whenconsideringtheimpactofmacroeconomicpolicyacrosscountries,thechoiceofinvoicecurrency(i.e.,whetherexporterschoosetheproducer’scurrencypricing(PCP),thelocalcurrencypricing(LCP),orthird-currencypricing)determinestheextentofinternationalshocktransmission.Asaresult,optimalmonetaryandexchangeratepoliciessignificantlydependonfirms’choseninvoicecurrencyforexportsandimports(e.g.,DevereuxandEngel,2003).

Numerousresearcheffortshavebeendevotedtoassessingthedeterminantsofinvoicecurrencychoice.Manyrecentstudieshaveempiricallyanalyzedhowthechoiceofinvoicecurrencyisaffectedbymicroandmacroeconomicfactorsusinghighlydisaggregatedtransaction-levelcustomsdatatodetermineinvoicecurrencyineachexport/importtransaction.Forinstance,usingunpublishedmicrodataonCanadianimports,GoldbergandTille(2016)revealedacorrelationbetweenthesizeofindividualtransactionsandinvoicecurrencyinadditiontotheimpactofconventionalmacroeconomicfactorssuchasexchangeratevolatilityandexchangerateregime.Devereuxetal.(2017)alsousedCanadiandatatoinvestigatetheimpactofexporter’sandimporter’smarketshareonthechoseninvoicecurrency.1ThisstudyisthefirsttoemployJapanCustoms’transaction-leveldata,whichhavenotbeenpreviouslydisclosedtoresearchers.Inearly2022,theJapaneseMinistryofFinance(MOF)startedallowingtheuseoflarge-scalecustomstransaction-leveldataforempiricalresearch,includingdetailedinformationonJapaneseexportandimporttransactions,ontheconditionofmaintainingtheanonymityoffirmsthatmadecustomsdeclarationsinexportsand/orimports.Ourresearchgroupwasapprovedforthisstudyandgranteduseofthetransaction-leveldataofexportsandimportsforthe2014–2020period.

Thenoveltyofthisstudyisitsrigorousinvestigationofwhetherinvoicecurrencychoicediffersbetweenintra-firmandarm’s-lengthtrade.WeuseapaneldatasetconstructedemployingtheJapanesetransaction-levelcustomsdata.Althoughglobalvaluechains(GVCs)havebecomeanimportantconsiderationintheempiricsofinternationaleconomics(e.g.,Timmeretal.,2014),alimitednumberofstudieshaveinvestigatedtheinvoicecurrencychoicealongGVCs.Forinstance,motivatedbyrecentresearchonGVCs,Chung(2016)andAmitietal.(2022)consideredtheeffectofimportedintermediateinputsonexportingfirms’invoicecurrencychoice.2Thesestudiesdemonstratedthatwhenmoreintermediateinputsareimportedfromabroadandinvoicedinforeigncurrencies,itismorelikelyforexportingfirmstouseforeigncurrenciesintheirexportsofproductiongoods.Suchtradetransactionsalongproductionchainsareusuallyexpectedtobeconductedviaintra-firmtradebetweengroupcompanies;however,thesestudiesdidnotclearlydistinguishintra-firmtradefromarm’s-lengthtradewhenanalyzingtheeffectoftheinvoicecurrencychoice.AnotherexampleisItoetal.(2018),whoemployedaquestionnairesurveyofJapaneseexportfirms,revealingthatinvoicecurrencychoicesignificantlydifferedbetweenintra-firmandarm’s-lengthexports.WhiletheauthorspresentednotablefindingsonJapanesefirms’invoicecurrencychoice,only227firms’validresponsestothequestionnairesurveywereincluded,andthe

1Manyotherexamplescanbecited,includingGopinath,Itskhoki,andRigobonetal.(2010)fortheUS;Casasetal.(2017)andGopinathetal.(2020)forColombia;Hayakawaetal.(2019b)forThailand;Chenetal.(2022)andCorsettietal.(2022)fortheUK;Montfauconetal.(2021)forMalawi;Amitietal.(2022)forBelgium;andSayg?l?(2022)forTurkey.

2Wealsoidentifyseveralstudiesintheliteratureofexchangeratepass-through(ERPT)thatconsideredGVCs,whichiscloselyrelatedtoinvoicecurrencychoice.Neiman(2010)empiricallyexaminedpossibledifferencesinERPTbetweenintra-firmandarm’s-lengthtradeusingatransaction-leveldatasetobtainedfromtheUSBureauofLaborStatistics,revealingthatthedegreeofERPTbetweenintra-firmandarm’s-lengthimporttransactionsdifferssignificantly.RondeauandYoshida(2023)investigatedtheeffectofvalue-addedcontributionsofexportersandimportersonthedegreeofERPTtoJapaneseimports,demonstratingthatthedegreeofERPTincreasesforindustrieswithhighercontributionsfromexportingcountries’value-addedand/orlowercontributionsfromimportingcountries’value-added.

3

empiricalresultsarebasedonpooledcross-sectiondatarepresentingalimitednumberofobservations.3

WefocusonJapaneseautomobileexportstoFrancetoinvestigatehowinvoicecurrencychoicediffersbetweenintra-firmandarm’s-lengthtrade.TheautomobilesectorisoneofJapan’slargestexportsectors,andJapaneseautomobilefirmshaveestablishedaconsiderablenumberofoverseassubsidiariesacrosstheworld,activelyexportingproductstothesesubsidiaries,andintra-firmtradehasasignificantroleinsuchtransactions.WechooseFranceasthedestinationcountrybecausetheproportionofarm’s-lengthexportsaccountforaround20%–30%ofthetradeflow,whichenablesustoconductarigorousempiricalexaminationofpossibledifferencesininvoicecurrencychoicebetweenintra-firmandarm’s-lengthexports.4

Theidentificationofintra-firmtradeisthekeyissueinthisresearch.Invoicecurrencychoicehasnotbeenrigorouslyexaminedbecauseitisdifficultwithoutknowingdetailedinformationonexportersandimporterssuchastheirnamesandaddressestoidentifywhetherimportfirmsareoverseassubsidiariesorindependentfirmswithoutcapitalties,evenwhenweusetransaction-levelcustomsdata.WematchedtheJapanCustoms’datawithexternalsourcesofoverseassubsidiariesandcapitalrelationships.FocusingonJapaneseautomobileexportstoFrance,wecouldpreciselyidentifythecapitaltiesbetweenexportersinJapanandimportersinFrance.Intheempiricalanalysis,weusealternativedefinitionsofintra-firmexportsconsideringthedegreeofcapitalties.WepresentthedetailsofthisapproachinSection2.

Ourempiricalexaminationdemonstratesthattheeuroismorelikelytobechosenforintra-firmexportsthanarm’s-lengthexports,implyingthatJapaneseautomobilefirmstendtostabilizetheeuro-basedexportpriceinintra-firmexportstoFrance,assumingallexchangeraterisksthatarisefromeuro-invoicedtransactions.Inarm’s-lengthexportstoFrance,Japanesefirmspreferinvoicinginyen,passingtheexchangeraterisktoFrenchimporters.Thisresultremainsrobusttousingalternativedefinitionsofintra-firmtrade.Wealsofindthatexporterstendtousetheeurowhentheproportionoftheeuroinimportsishigh,thefirms’marketshareinFranceishighorlow,orwhentheyenislessvolatileagainsttheeuro.ThesefindingsareconsistentwiththoseofpreviousstudiessuchasDevereuxetal.(2017).Wealsofindthattheeuroisusedmoreforexportsoffinalgoodsthanforintermediategoods.Weinterpretthisresultasmeaningthatthechoseninvoicecurrencytendstomatchthedestinationcountry’scurrencywhenexportedproductsarefinallyconsumed.Furthermore,regardingfirmcharacteristics,wefindthatlaborproductivityandresearchanddevelopment(R&D)investment,whichcanbeconsiderednon-pricesourcesofexportcompetitivenessthatnegativelyaffectsthelikelihoodofeuro-invoicedtransactions.Weinterpretthisresultasmeaningthatexporterscanoffertheuseofthehomecurrencyininvoicingwhentheyholdsignificantbargainingpowerthatisgeneratedbycompetitivenessinthedestinationcountry.

Theremainderofthispaperisorganizedasfollows.Section2describestheJapanesecustomsdataandthetransaction-levelpaneldatasetconstructedforthisstudy.Section3presentstheempiricalframeworkandourbaselineresults.Section4checkstherobustnessoftheempiricalfindings.Finally,Section5concludesthispaper.

2.Data

2.1.MatchingMethodology

WeuseJapanesecustomsdatathatwasnewlydisclosedtoconductthisresearch.Thedatacontainthefollowinginformationforeachexportandimporttransaction:(i)tradevalue(inyen),(ii)correspondingquantity

3FribergandWilander(2008)alsoconductedaquestionnairesurveyofSwedishexporters,whichprovideduswithusefulinformationonfirm-levelSwedishexportinvoicingdecisions;however,thestudydidnotidentifysignificantdifferencesbetweenintra-firmandarm’s-lengthtradeintheinvoicecurrencychoice.Itoetal.(2018)conductedtwoquestionnairesurveysofJapaneselistedfirmsin2009and2013fortheirempiricalanalysis;however,asnotedpreviously,panelestimationwasnotconductedsincealimitednumberoffirmsrespondedtothesurveystwice.

4Infact,theproportionofarm’s-lengthtradeinthissectorisextremelylowforsomeimportingcountries.

4

(numbersand/orweightsuchastonsorkilograms),(iii)invoicecurrency,(iv)productinformation(HarmonizedSystem(HS)nine-digitclassification),(v)identificationinformationforJapaneseexportersandimporters(IDnumber,name,address,andtelephonenumber),(vi)destinationorsourcecountry,(vii)tradepartners’information(nameandaddress),and(viii)transactiondate,time,andotherrelevantdetails.Thesampleperiodspansfromthefirstdayof2014tothelastdayof2020.

Althoughthedataenableaccesstoalltransactionsofallproductsforallpartnercountries,wefocusonautomobileexportstoFrancetopreciselyidentifythedegreeofcapitaltiesbetweenexportersandimporters.Becausewecannotdirectlyobtaininformationoncapitaltiestodifferentiateintra-firmandarm’s-lengthtradefromtheJapanesecustomsdata,weidentifyimporterswithfinancialtiestoJapaneseexportersbyconductingmachinematchingusingthePythonsoftware.Themachinematchingwasconductedbasedonfirm-levelcapitalrelationshipdataprovidedbytheTeikokuDataBank(theoriginalsourceofwhichistheBureauvanDijk’sOrbisdatabase),inadditiontothenameandaddressofexportersandimportersobtainedfromthecustomsdata.Frenchimporters’namesandaddressesaresubmittedbyJapaneseexportersorcustomsbrokers,andunfortunately,thisinformationisnotrecordedinaconsistentmanner,andtypographicalerrorsareobserved.Thisissuecouldreducetheaccuracyofthemachinematchingprocedure;therefore,forvalidationandaccuracy,wecomplementarilymatchimportersandexportersmanually.

Furthermore,theOrbisdatabasedoesnotnecessarilycoveralloverseassubsidiariesofJapaneseautoexporters;therefore,weprepareadditionallistsofoverseassubsidiariesfromtwootherdatabases,includingthedatabaseforJapanesecompaniesexpandingbusinessabroad(KaigaiShinshutsuKigyoSoran),whichispublishedbyToyoKeizaiInc.andtheMarkLinesInformationPlatform,whichincludesextremelydetailedandcomprehensivedataonautomobileproductionfromcountriesaroundtheworld.5Thesedatabasesarealsoincludedinourmanualmatchingprocedure,expandingtheshareofmatchesinthemanualprocedure,whichislargerthanthatofthemachineprocedure.Aslongasmanualmatchinghasanimportantroletocompensateformachinematching,dealingwithallsamplecountriesandindustriesisnotareasonableoption.6

WechoosetofocusonJapaneseautomobileexportersbecauseautomobilesareoneofJapan’slargestexportproductsandhaveamacroeconomicimpactontheJapaneseeconomy.Inaddition,Japaneseautomobilefirmshaveestablishedaconsiderablenumberofoverseassubsidiariesacrosstheworld,whichindicatesthesignificanceofintra-firmtradeinthesector.ToensureasufficientnumberofJapaneseexportfirms,wechose12parentfirmsthatproducecompletedautomobilesand/ormotorcycles.WealsoincludefivedomesticsubsidiariesinJapanthataredirectlyownedbytheaboveparentfirmswithatleasta50%ownershipstake.Therefore,weexaminedatafor17JapaneseautofirmsthatexportcompletedautomobilesandmotorcyclesandrelatedpartsandcomponentstoFrance.

WechooseFranceasthesampledestinationcountryfortworeasons.First,theyenandtheeuro(i.e.,thelocalcurrency)arepredominantlyusedinthistradeflow,andtheproportionofothercurrenciesisnegligible.Doingsomayrendertheinvoicecurrencychoiceinourtargettradeflowmoreimmunetothedominantcurrencyparadigm(DCP)proposedbyGopinathetal.(2020).WeareabletofocusonthebinarychoiceofinvoicecurrencybetweenPCPandLCP,whichistheoreticallysimplerthanincludingathirddominantcurrencysuchas

theUSD.Asacomparison,panels(a)and(b)inFigure1illustratetheshareofyenandUSDinJapan’sexportsandimportsto/fromtheworldforallindustries,respectively,revealingasignificantshareofUSDincontrastto

5Inadditiontotheabovethreedatasources,whennecessary,wedouble-checkedthewebsitesofimportfirmstoconfirmourfindings.

6Forinstance,inApril2014,morethan1.6millionexporttransactionstotheworldwereconductedbymorethan50thousandfirms.Coveringallthesesamplesisnotreasonabletoconductmanualmatchingtocompensatemachinematching.

5

oursampletradeflow.OnemaythinkthatwecanextendthesampletootherEUcountriessincetheyenandthe

euroaresupposedtobedominantforotherEUcountries.Panels(c)and(d)inFigure1showtheshareofyen,USD,andEUcountries’currencies,(includingtheeuro)inJapan’sexportsandimportsto/fromEUcountriesforallindustries,respectively.Comparedwiththecasefortheworld,theproportionofUSDhasbeensmallinrecentyears,hoveringaround12%–14%forbothexportsandimports;however,theUSDstillhasanotableshare,implyingthattheUSDcouldbeoneoptionconsideredbyJapaneseexportersandimportersinthesetradeflows.WesubsequentlyfocusedonJapaneseautomobileexportstoFrance,asonlytheyenandeuroareusedinmosttransactions.Oursecondreasonforchoosingthistradeflowisthattheshareofarm’s-lengthexportsissignificant,whichallowsustoconductareasonablecomparisonbetweenintra-firmandarm’s-lengthtradewithasufficientnumberofsamplesforbothtypesoftrade.Asadditionalinformation,Figure2illustratesthesalesofJapaneseautomakersinmajordestinationcountriesobtainedfromtheMarkLinesdatabase,7indicatingthatFranceisthefifthlargestoverseasmarketforJapaneseautomakers.Furthermore,Figure3presentsforeignautomaker’ssalesinFrance,showingthatJapaneseautomakershavethesecondlargestshareoftheFrenchmarket,indicatingastrongpresenceofJapaneseautomakersinFrance.

===Figure1===

===Figure2===

===Figure3===

2.2.DefinitionsofIntra-firmExports

Onechallengeinidentifyingintra-firmandarm’s-lengthexportsisdeterminingadefinitionofoverseassubsidiariesbasedonthedegreeofcapitaltiesbetweenJapaneseexportersandFrenchimporters.Weconsiderthefollowingfivetypesofoverseassubsidiaries,thataregraphicallyillustratedinFigure4,forinclusioninintra-firmexports.

Sub-1firm:OverseasfirmsthataredirectlyownedbyaparentfirminJapanwithatleasta10%ownershipstake.

Sub-2firm:OverseasfirmsthataredirectlyownedbyaSub-1firmwithatleasta50%ownershipstake.Sub-3firm:OverseasfirmsthataredirectlyownedbyaSub-2firmwithatleasta50%ownershipstake.

Sub-Sub-1firm:Overseasfirmsthataredirectlyownedbydomesticsubsidiaries(i.e.,theparentfirm’ssubsidiariesinJapan)withatleasta10%ownershipstake.

Sub-Sub-2firm:OverseasfirmsthataredirectlyownedbyaSub-Sub-1firmwithatleasta50%ownershipstake.

===Figure4===

Aspreviouslynoted,weconsidertheexportsof17Japanesefirms,including12parentfirmsandfivedomesticsubsidiaries.Wesetupthefourdefinitionsofintra-firmexportsasfollows.

Intra-firm1:SeventeenJapanesefirms’exportstoSub-1andSub-2firms.

Intra-firm2:SeventeenJapanesefirms’exportstoSub-1,Sub-2,andSub-3firms.

7Figures2and3representthevalueofsalessincetheMarkLinesdatabaseonlyprovidesthisinformation,whereasourempiricalanalysisisconductedforeachtransaction(notthetransactionvalue).

6

Intra-firm3:SeventeenJapanesefirms’exportstoSub-1,Sub-2,Sub-Sub1,andSub-Sub2firms.

Intra-firm4:SeventeenJapanesefirms’exportstoSub-1,Sub-2,Sub-3,Sub-Sub1,andSub-Sub2firms.

Thedefinitionofintra-firmexportsisnarrowestinintra-firm1andbroadestinintra-firm4.Intra-firm2addsSub-3firmstointra-firm1,whileintra-firm3addsSub-Sub1andSub-Sub2firmstointra-firm1withoutincludingSub-3firms.Inthissense,intra-firm2andintra-firm3complementoneanother.Weobtain114,335exporttransactionsforthewholesampleperiod(2014–2020).Amongthesetransactions,theshareofintra-firmexportsrangesfrom73.4%to81.2%,dependingonthedefinitionofintra-firmtrade.Table1presentstheproportionofintra-firmtradeforeachdefinition,andFigure5showstheshareofintra-firmandarm’s-lengthtradeeachyear.

===Table1======Figure5===

Table2presentsthenumberandshareoftransactionsusingtheyenandeuro.Weexcludedtransactionswithothercurrenciesbecausetheyaccountedforlessthan0.1%ofthetotalnumberoftransactionsandwereomittedfromourempiricalinvestigation.Theupperpanelshowsthenumberandsharesinthefullsample.Oursample’stotalnumberoftransactionsrangesfrom14.8thousandto18.7thousand,dependingontheyear.Theeurosharewaslargerthanthatoftheyeninallsampleyears.

Nevertheless,theyenhadasignificantshareeachyear,andtheproportionsofthetwocurrenciesarecomparableinsomesampleyears(i.e.,in2017and2020).Thelowerpanelshowsthenumberandsharebytradetype.Wecountedthenumberandcalculatedthesharebasedonthefouralternativeclassificationsofintra-firmtradeintroducedinTable1.Aclearcontrastbetweenintra-firmandarm’slengthtradeisevident,revealingthattheeurohadadominantshareinintra-firmtrade,whereastheyenhadadominantshareinarm’slengthtrade,whichholdsforallfourclassificationsofintra-firmtrade.Notably,thelocalcurrency(i.e.,theeuro)isusedmoreinintra-firmtradethaninarm’slengthtrade.

===Table2===

3.EmpiricalAnalysis

3.1.EmpiricalFramework

Toexaminethechoiceofinvoicecurrencyinintra-firmtrade,weestimatethefollowingempiricalequationforJapanesefirmi’sexporttransactionkofaproductdefinedattheHSnine-digitlevelinyearttoFrance:

yi,k,HS9,t=α+βIntrai,k,HS9,t+YX+εi,k,HS9,t(1),

yi,k,HS9,tisadummyvariablethattakesavalueofonefortransactionsinvoicedintheeuroandzeroforthoseinvoicedintheyen.Asnotedabove,sincemosttransactionsareconductedintermsofeithertheeurooryen,weexcludedsampleswithothertypesofinvoicecurrency.Intrai,k,HS9,tisadummyvariablethattakesavalueofoneforintra-firmexporttransactionsandzerootherwise.Wehypothesizethattheeuroismorelikelytobechoseninintra-firmtrade,implyingapositivesignofβ.Weinterchangeablyusethealternativedefinitionsofintra-firmtradeshowninTable1tovalidatetherobustnessoftheresults.Xisthevectorofotherexplanatoryvariables.Inthebaselineanalysis,weemployfourvariablesinadditiontotheintra-firmdummy.First,weincludeanintermediategoodsdummyIntermediateHS6,whichtakesavalueofoneiftheproductinthetransaction

7

representsintermediategoodsandzeroforfinalgoods.8Weexpectthattheeuro(i.e.,thelocalcurrency)ismorelikelytobeusedforfinalgoodsexports,asfinalgoodsaresoldtoFrenchcustomersintheeurowhenimportedfromJapan;thus,importershadsignificantmotivationtoinvoicethesefinalgoodsintheeuro,indicatinganegativecoefficientforIntermediateHS6.Itisalsoexpectedthatthisvariablecouldcorrelatewiththeintra-firmdummy,asintermediategoodsmightbeexportedtoamanufacturingcenterofthesamecompanygroupinthelocalmarket.Nevertheless,ourbaselineresultsremainunchangedwhenweintroduceintra-firmandtheintermediategoodsdummiesseparately.9

Second,weconsidertheshareofinvoicecurrencyinimports,capturingthisvariable(IIsi,t)byusingthenaturallogoftheshareofthenumberofimporttransactionsinvoicedintheeurointhetotalnumberofimporttransactionsconductedbythefirmeachyear.WearealsoallowedtouseJapan’stransaction-levelcustomsimportdata,althoughourfocusisonfirms’exportbehavior.Weusethisimportinformationtoexaminethepossibleeffectofinvoicecurrencyshareinintermediateinputimportsontheexporters’invoicecurrencydecision,whichhasbeenwelldocumentedbyChung(2016).10Firmstendtochoosethesamecurrencyforexportsandimportsofintermediateinputstooffsettheexchangerateriskarisingfromexportandimporttransactions;therefore,weexpectapositivecoefficientforthisvariable.

Third,weincludethenaturallogoftheexchangeratevolatilityoftheyenvis-à-vistheeuroineachyear(volt).Wedefinethisvolatilityasthestandarddeviationofthedailyaverageexchangerate.AsnotedbyDevereuxetal.(2004),risk-averseexportersprefertheexporter’scurrencywhenthevalueofthecurrencyisvolatileagainsttheimporter’scurrency.Inourcontext,weexpectthattheeuroislesslikelytobechosenwhentheyenismorevolatileagainsttheeuro.Inthisregard,weexpectanegativecoefficientforvolt.TheexchangeratedataareobtainedfromthePacificExchangeRateServicefromtheUniversityofBritish