美國制造商在全球價值鏈中不斷變化的公司和國家邊界 The Changing Firm and Country Boundaries of US Manufacturers in Global Value Chains

NBERWORKINGPAPERSERIES

THECHANGINGFIRMANDCOUNTRYBOUNDARIESOFUSMANUFACTURERSINGLOBALVALUECHAINS

TeresaC.Fort

WorkingPaper31319

/papers/w31319

NATIONALBUREAUOFECONOMICRESEARCH

1050MassachusettsAvenue

Cambridge,MA02138

June2023

AnyviewsexpressedarethoseoftheauthorandnotthoseoftheUSCensusBureau,theBureauofEconomicAnalysis,theNationalBureauofEconomicResearch,ortheCentreforEconomicPolicyResearch.TheCensusBureauandtheBureauofEconomicAnalysishavereviewedthisdataproducttoensureappropriateaccess,use,anddisclosureavoidanceprotectionoftheconfidentialsourcedatausedtoproducethisproduct.ThisresearchwasperformedataFederalStatisticalResearchDataCenterunderFSRDCProjectNumber1975(CBDRB-FY23-P1975-R10185),CESProject1530(releaseon7/15/2019),andCESProject6907751(BEA-FY23-P6907751-R2).JackLiangprovidedexceptionalresearchassistance.IthankPolAntràs,JimFetzer,FarihaKamal,PeterSchott,andChadSyversonforveryhelpfulcomments.Ialsothanktheeditors,ErikHurst,NinaPavcnik,HeidiWilliams,andespeciallyTimTaylorforinvaluableinsights,comments,andsuggestions.

NBERworkingpapersarecirculatedfordiscussionandcommentpurposes.Theyhavenotbeenpeer-reviewedorbeensubjecttothereviewbytheNBERBoardofDirectorsthataccompaniesofficialNBERpublications.

?2023byTeresaC.Fort.Allrightsreserved.Shortsectionsoftext,nottoexceedtwoparagraphs,maybequotedwithoutexplicitpermissionprovidedthatfullcredit,including?notice,isgiventothesource.

TheChangingFirmandCountryBoundariesofUSManufacturersinGlobalValueChains

TeresaC.Fort

NBERWorkingPaperNo.31319

June2023

JELNo.F1,L2,L6

ABSTRACT

ThispaperdocumentshowUSfirmsorganizegoodsproductionacrossfirmandcountryboundaries.MostUSfirmsthatperformphysicaltransformationtasksin-houseusingforeignmanufacturingplantsin2007alsoownUSmanufacturingplants;moreovermanufacturingcomprisestheirmaindomesticactivity.Bycontrast,“factorylessgoodsproducers”outsourceallphysicaltransformationtaskstoarm's-lengthcontractors,focusingtheirin-houseeffortsondesignandmarketing.Thisdistinctfirmtypeismissingfromstandardanalysesofmanufacturing,growinginimportance,andincreasinglyreliantonforeignsuppliers.Physicaltransformation“within-the-firm”thuscoincideswithsubstantialphysicaltransformation“within-the-country,”whereasitsperformance“outside-the-firm”oftenalsoimplies“outside-the-country.”Despitethesedifferences,factorylessgoodsproducersandfirmswithforeignanddomesticmanufacturingplantsbothemployrelativelyhighsharesofUSknowledgeworkers.Thesepatternscallfornewmodelsanddatatocapturethepotentialforforeignproductiontosupportdomesticinnovation,whichUSfirmsleveragearoundtheworld.

TeresaC.Fort

TuckSchoolofBusiness

DartmouthCollege

100TuckHall

Hanover,NH03755

andCEPR

andalsoNBER

teresa.fort@

Adataappendixisavailableat/data-appendix/w31319

1Introduction

TherearetwomainconcernsaboutUSmanufacturingandglobalization:thelossofUSjobstoforeignplacesandthelossofaUSknowledgebaseconnectedtomanufacturing.Inthispaper,IdocumentthefullrangeofUSmanufacturing?rms’domesticandglobaloperations,providingabroadercontextfortheseconcerns.

Myperspectiveemphasizesthatmanufacturinginvolvesthreemajorstages:1)productdesignandinnovation;2)aseriesofphysicaltransformationactivities,suchasmakinginputsandassemblingthem;and3)sales,marketing,anddistribution.Mosttrademodelsimplicitly(orevenexplicitly)includeallthreestages,butgovernmentstatisticsonlyclassifyphysicaltransformationtasksasman-ufacturing.Thismismatchfromtheorytodatawasnotproblematicwhenallthreestageswereperformedinsideaparticular?rmandcountry.However,dramaticimprovementsininformationandcommunicationtechnologyhavemadeitincreasinglypossibletofragmentthesestagesacrossmultiplecountriesand?rms.Thisfragmentationhasmademeasuringthecompleteproductionprocessformanufacturedgoodsdi?cult(orevenimpossible)withtraditionaldatasets.

Figure

1

illustratesa?rm’schoicestofragmentproductionacrosscountriesand?rms,buildingonadiagramintroducedby

Feenstra

(

2010

).Thehorizontalaxiscapturesthe?rmboundarydecisionforphysicaltransformationtasks:the?rmmay“outsource”taskstoother?rms,ormaintainintegratedproduction.Theverticalaxiscapturesthe?rm’slocationchoiceforthesetasks:the?rmmay“o?shore”bylocatingproductioninoneormoreforeigncountries,orproducedomestically.

AUS?rmwithmanufacturingplantsintheUnitedStatesnecessarilyoccupiesquadrant1,sinceitperformsphysicaltransformationtasksin-house.A?rmthatownsforeignmanufacturingplantsoccupiesquadrant3.However,asingle?rmmayoccupymultiplequadrants.Forexample,considerTexasInstruments,aUSsemiconductormanufacturerthatownsandoperatesmultiplewaferfabsintheUnitedStates,alongwith11otherproductionsitesinMexico,Europe,andAsia.TheFordMotorCompanyhas30manufacturingplants,20ofwhichareinforeigncountries.Sincethese?rmsmanufacturein-houseintheUnitedStatesandabroad,theyspanquadrants1and3.BothTexasInstrumentsandFordalsoworkextensivelywitharm’s-lengthpartners,suchthattheyalsospanquadrants2and4,thuscoveringtheentirematrix.

1

Incontrastto?rmsthatperformphysicaltransformationtasksin-house,a“factorylessgoodsproducer”isa?rmthatcontractsforallofitsphysicaltransformationactivities,andassuchoccu-piesonlyquadrants2and/or4,dependingonwhetheritssuppliersaredomestic,foreign,orboth.Forexample,Appleisdeeplyinvolvedinmanufacturingphysicalgoods,butdoessoviathird-partycontractmanufacturingsuppliersprimarilyinforeigncountries.Similarly,Nikereports640manu-facturinglocationsacross38countries,allofwhichinvolveoutsourcedrelationshipswithcontractmanufacturers.Qualcommisoneofmany“fabless”semiconductor?rmsthatdesignchipsandrelyonpredominantlyKoreanandTaiwanesecontractmanufacturersfortheirproduction;indeed,

Ba-

1See/about-ti/company/ti-at-a-glance/manufacturing.htmlforTIplants,https://

/about-ti/suppliers/supplier-portal.html?keyMatch=SUPPLIERSforTIsupplierportal,https://

/operations/locations/global-plants.html

forFordplants,and

https://fsp.portal.

/web/portal

fortheFordSupplierPortal.

1

2

Taskperformedinhome(HQ)country

O

u

t

s

o

u

r

c

i

n

g

Taskperformedinforeigncountry

yardetal

.(

2015

)attribute25percentofglobalsemiconductorsalesin2012tosuchfactorylessgoodsproducers.

2

Itisworthclarifyingthatfragmentingproductiondoesnotjustrefertopurchasinginputs.Forexample,aBelgiancandy-maker’simportsofchocolatearegenerallynotconsideredfragmentedpro-ductioninstudiesonoutsourcingando?shoring.Instead,fragmentedproductionentailsasplittingapartoftheproductionprocessbya?rmthatusedto,orcouldreasonablyhave,producedthefragmentedpart.

Figure1:Firmintegrationandlocationdecisionsforphysicaltransformationtasks

TaskperformedwithinthefirmTaskoutsourcedtoanotherfirm

Domestic

In-houseProduction

(Q1)

FordMotor

Domestic

Outsourcing

(Q2)

Company

Qualcomm

Foreign

Outsourcing

oring”

(Q4)

ForeignDirect

Investment

(Q3)

“Offsh

Source:This?gurebuildsontheframeworkpresentedin

Feenstra

(2010

).Itdepictsthe?rm(horizontalaxis)andcountry(verticalaxis)boundarydecisionsmadeby?rmsinvolvedinthebroadermanufacturingprocessforthestage2,physicaltransformationtasksrequiredtomanufacturegoods.

Firmswithnodomesticmanufacturingplantshavenoactivityinquadrant1.Theytendtobemissingfromresearchonproductionfragmentationando?shoring,becausethereisnoclearwayinstandarddatasetstoidentifytheirdirectinvolvementinmanufacturing.Inthepaper,Ibeginbydescribingthelimitationsofstandarddatasetsinidentifyingsuch?rms.IthenexploittwonovelUSdatasourcestoidentifytwoorganizationalformsmissingfrommanyanalysesonglobalvaluechains:US?rmsthatperformphysicaltransformationtaskswithinthe?rmboundaryusingexclu-sivelyforeignmanufacturingplants,andfactorylessgoodsproducersthatoutsourcealltheirphysicaltransformationtaskstoarm’s-lengthcontractmanufacturers.

ContrarytothefearthatUSmultinationalshaveo?shoredmostoftheirjobs,I?ndthatthevastmajorityofUS?rmsthatownforeignmanufacturingplantsin2007alsomaintaindomesticproduction;moreover,manufacturingcomprisestheirprimarydomesticactivity.Contrarytothefearthatparticipationinglobalvaluechainsentailsalossoftechnologicalskills,I?ndthat?rms

2ForApple,see

/articles/BL-DGB-25630

;forNike,see

https://manufacturingmap

.

/#

;andforQualcomm,see

/company/corporate-responsibility/

acting-responsibly/sustainable-product-design/supply-chain

.

3

withglobalin-housemanufacturingplantsandfactorylessgoodsproducersbothemployrelativelyhighsharesofUS‘knowledge’workers.Indeed,multinationalenterprisesthatmanufacturegoodsaredisproportionatecontributorstoR&Dandpatenting,andfactorylessgoodsproducersarefarmorelikelytodesigngoodsthanother?rmsintheirsector,andhavealsobeenlinkedtogreaterR&D,patenting,andtrademarks(

Kamal

,

2020

).

AcompletepictureofUS?rms’involvementinglobalvaluechainsisnecessarytounderstandthee?ectsofglobalization.Takeforexample

Bermanetal.

(

1994

),whoruledouttradeasanexplanationfortheshifttowardsnon-productionworkersinUSmanufacturingplantsduringthe1980sand1990sbecausethosechangesoccurredinsomeofthesameindustrieswithsurgingimports,notablycomputerandelectronics.TheyreasonedthatatradeexplanationmustentailreallocationacrossindustriesconsistentwithUScomparativeadvantage,whereaswithin-industryadjustmentsdominatedinthedata.Wenowknow,however,thatcomputerandelectronicmanufacturingisoneofthe?rstindustriesinwhichfactorylessgoodproductionarose,assome?rmsfocusedoninnovationandshiftedphysicalproductiontoforeignsuppliers.Itisnotablethatcomputerandelectronicsalsoaccountsforthegreatestgrowthinbreakthroughpatentsoverthelasttwodecades(

Kellyetal

.,

2021

)andthemajorityofrealvalueaddedgrowthinUSmanufacturingfrom1992to2011,evenasimportsofcomputersandelectronicssurged(

Fortetal.

,

2017

).

Iconcludewithadiscussiononhowtradestatisticsandtheoryneedtoexpandtocapturetherealitiesofgoodsproductionacross?rmandcountryboundaries.Theseactivitiesa?ectourunder-standingoftradeandforeigndirectinvestment,aswellasaggregatemeasuresofdomesticvalueaddedandGDP.Thepotentialimplicationsarefar-reaching:increasedspecializationwithintheproductionofaparticularindustryorgoodprovidesadditionalgainsfromtrade(

JonesandKierzkowski

,

2001

).Whensuchspecializationentailsreallocationintoearlyproductionstages,suchasdesignandinno-vation,o?shoringcanevenleadtodynamicgains,asthereturnstoinnovationrise,inducinggrowthinR&Dandideasthatbegetmoreideas(

GrossmanandHelpman

,

1991

;

Rodr′?guez-Clare

,

2010

).

2MeasuringtheRangeofManufacturers’OrganizationalForms

Thetraditionaldataonmanufacturing?rmscollectedbyUSstatisticalagenciesmakeitdi?culttocapture?rmsthatareinvolvedinthemanufacturingprocesseitherbyproducinggoodsexclusivelyoutsidetheUnitedStates,exclusivelyoutsidetheir?rmboundariesviacontracts,orboth.IreviewtheexistingdataonUSestablishmentsand?rms,oninternationaltradeby?rms,andonmultinational?rms,explainingthebene?tsandlimitationsofeachdatasource.

2.1StandardMeasuresofUSEstablishmentsandFirms

TheCensusBureaude?nesan“establishment”asaphysicallocationatwhichemploymentandpayrollrecordsarekept.A?rmcanthushavemultipleestablishments–andtheseestablishmentsneednotbeclassi?edinthesameindustry.

TheUSCensusBureauconstructstheLongitudinalBusinessDatabase,whichisacomprehensive,establishment-leveldatasetofallprivate,non-farmemployerestablishmentsfrom1976to2019.The

4

datasetprovidesemployment,payroll,location,andauniqueindustrycodefortheprimaryactivityofeveryestablishment(see

JarminandMiranda

,

2002

;

Chowetal.

,

2021

,fordetails).Allemployeesofanestablishmentareassignedtoitsindustry.Thedataalsoidentifythe?rmtowhicheachestablishmentbelongs,makingitpossibletomeasurethefullrangeofa?rm’sactivitiesacrossindustriesandsectors.

TheLongitudinalBusinessDatabasecanbemergedtotheEconomicCensuses,whicharecollectedinyearsthatendin2or7.Thesecensusescontaindetailedinformationonestablishmentsales,inputuse,andothersector-speci?cmetrics,suchastechnology.

Anestablishment’sindustryistheprimarymeansthatgovernmentagenciesandresearchersusetoidentifymanufacturingactivity.USstatisticalagenciesusetheNorthAmericanIndustryClassi-?cationSystem,commonlyreferredtoasNAICS(anddescribedat/naics)toclassifyestablishments.TheguidingprincipleofNAICSistoassignanindustrycodetoanes-tablishmentbasedonthemainactivitiesperformedbyitsemployees.Bycontrast,theStandardIndustrialClassi?cationSystem(SIC)classi?edestablishmentsthatprovidedsupportservicesforotherestablishmentsoftheir?rmtothoseestablishments’industry.Forexample,anR&DlabisalwaysinServicesunderNAICS,butwouldhavebeenclassi?edinmanufacturingunderSICifitsR&Dwastosupportthe?rm’smanufacturingplants.USCensusdatatransitionedfromNAICStoSICbetween1997to2002,aperiodthatcoincideswithChina’sentrytotheWorldTradeOrganiza-tion,makingthisissueparticularlyrelevantforresearchonglobalization.ThecurrentLongitudinalBusinessDatabasenowincludesthemostrecentvintageofNAICScodesforeveryestablishmentovertheentireperiodusingthemethodsdevelopedin

FortandKlimek

(

2018

).

Factorylessgoodsproducersarehardtoidentifywiththesedata.Tobeclassi?edinmanufactur-ing,anestablishmentmustperform‘mechanical,physical,orchemicaltransformationofmaterialsorcomponentsintonewproducts’.Asaresult,anestablishmentthatcontractsformanufacturingserviceswillgenerallybeclassi?edinnon-manufacturingsectors,suchaswholesaletrade.

3

Similarly,establishmentsthatperformsupportactivitiesformanufacturing,suchasanR&DlaboranEngineer-ingservicesestablishmentareclassi?edinservices.Giventheongoingfragmentationofdesignandproduction,weneedwaystoidentifyestablishmentsand?rmsinvolvedinthebroadermanufacturingprocess.

2.2MerchandiseTradeDatabyFirm

Startingin1992,theCensus?rmdatacanbemergedto?rm-leveldatafromUSCustomsthatrecordtheuniverseoftradetransactionsabove$2,500ofmerchandisegoodsthatenterorleavetheUnitedStates.ThisLongitudinalFirmTradeTransactionsDatabase(LFTTD)providesdetailedinformationontheproductsshipped,aswellasthesource(forimports)ordestination(forexports).Thesedatawere?rstlinkedby

Bernardetal.

(

2009

);

KamalandOuyang

(

2020

)providedetailsonthelatestlinkinge?orts.

OneuniquefeatureoftheseUStradedataisthattheycontainanidenti?erfortransactions

3ForanexplanationfromtheCensusBureau,see

/naics/?input=31&year=2022&details=

31

.Anexceptionis‘jobbers’incertainapparelmanufacturingindustries.Theseestablishmentsperformthe‘en-trepreneurialfunctionsinvolvedinapparelmanufacturing,’butcontractforthetransformationactivitiesfromother?rms.

5

betweenrelatedparties.Exporttransactionsinwhichoneofthepartieshasatleast10percentownershipoftheotherparty,orimportsbetweenpartieswitha5percentownershipthresholdare?aggedas“related-party”transactions.

4

Thus,arm’s-lengthtradecanbedistinguishedfrom?owsbetweenrelatedparties.However,forthepresentpurposeofstudyingwhetherUS?rmshavein-tegratedmanufacturingplantsinforeigncountries,thesedatahavewell-knownshortcomings;theyhavenoinformationonactivitiesofa?liatesofmultinationalenterprises;theydonotdistinguishUSmultinationalenterprisesfromforeign-owned?rms;andtheyarebasedonverylowownershipthresholds.

2.3MultinationalFirmsintheUnitedStates

TheUSBureauofEconomicAnalysiscarriesouttheAnnualSurveyofUSDirectInvestmentAbroad,knownasBE-11,whichprovidesinformationonallUS-based?rms’outwardforeigna?liateemploy-ment,localsales,salesbacktotheUnitedStates(andwhethertheseareintra-?rm),andsalestothirdmarkets,bythea?liatecountryandindustry.Thissurveythuscapturesoutwardforeigndirectinvestmentinformation.TheBureauofEconomicAnalysisalsocarriesouttheBenchmarkSurveyofForeignDirectInvestmentintheUnitedStates,knownasBE-12,whichisconductedevery?veyearsandprovidesinwardforeigndirectinvestment.Thissurveymakesitpossibletoidentifyforeign-owned?rmsoperatingintheUnitedStates.

For2007,theyearofmyanalysis,thesetwosurveysprovidethemostdetailedandcomprehensiveinformationavailableaboutmultinational?rmsoperatingintheUnitedStates.IncontrasttotheCensusBureau’srelated-partytradedata,thesedataincludeshare-of-ownershipinformation,aswellasforeigna?liates’industries,andtheirlocal,US,andthird-marketsales.

Despitetheiradvantages,thesedataalonearenotsu?cienttostudythefullrangeofUS?rms’manufacturingactivities.First,thesedataonlyincludemultinationalssothereisnoinformationondomesticmanufacturers.Second,theyarereportedatthe?rmlevel,andthereforedonotcontaintheestablishment-levelinformationnecessarytoanalyzethefullrangeof?rms’domesticestablishments.Finally,thedatalackcountryandproduct-levelinformationontheuniverseof?rms’importsandexports.

Inthenextsection,Icombinethe2007USCensusBureauandBureauofEconomicAnalysisdatadescribedheretoidentifyallUS?rmswithintegratedmanufacturingplantsanywhereintheworld.Suchananalysisensurescoverageofall?rmswithanyactivityinquadrants1or3ofFigure

1

.Inthesubsequentsection,Iexploitdetailedquestionsfromthe2017EconomicCensusofWholesaleTradetoidentify?rmsthatareinvolvedinthebroadermanufacturingprocessbycontractingforproductionfromarm-s-lengthsuppliers,thuscapturing?rmsspecializedinquadrants2and/or4.

3CountryBoundariesofIntegratedUSManufacturers

Inthissection,IfocusonUS?rmsthataredirectlyinvolvedinmanufacturingbecausetheyhavemajority-ownershipsharesinmanufacturingplantsintheUnitedStates,inforeigncountries,orboth.

4Related-partyimportsmayincludeotherrelationships,see

KamalandOuyang

(2020

)fordetails.

6

Assuch,these?rmsnecessarilyoccupyquadrants1or3ofFigure

1

.Inthenextsection,IturntofactorylessgoodsproducerssuchasAppleandNike.

3.1NovelDataonallIn-houseManufacturingbyUSFirms

Iusenewdatamergedandanalyzedby

Kamaletal.

(

2022

)and

Antr`asetal.

(

2023

)toprovideacompletepictureofUS?rmsthatperformphysicaltransformationtasksin-houseanywhereintheworld.Akeycontributionofmyanalysisistoinclude?rmswithnodomesticmanufacturingplants,whicharemissingfromstudiesusingtraditionaldatasets.Imeasure?rms’employment,sales,andtradeactivityacrosssectorsbylinkingthe2007LongitudinalBusinessDatabase,EconomicCensuses,andLongitudinalFirmTradeTransactionsDatabase(excludingtradeofminerals,fuels,andoil(HS27)).IidentifymultinationalenterprisesasallUS?rmswithmajority-ownedforeigna?liatesusingthe“outward”foreigndirectinvestmentsurvey.Iusethe“inward”surveytoremoveallUSestablishmentsthataremajority-ownedbyaforeign?rm.

In2007,thereare243,700US?rmsthatownmanufacturingplantssomewhereintheworld(about5.6percentofall?rms),whichaccountfor88percentoftotalUSmanufacturingemployment(foreignmultinationalenterprisesemploytheremainder),20percentoftotalemployment,and29percentoftotalsales.These?rmsmediate42percentofUSimportsand58percentofexports,whichhighlightsthedisproportionateinvolvementofgoods-producing?rmsininternationaltrade.AppendixTableA.1providesamoredetaileddecompositionofthesestatistics.

3.2Domesticversuso?shoredintegratedmanufacturing

AcommonperceptionisthatUSmultinationalenterpriseshaverelocatedthebulkoftheirmanu-facturingplantso?shore.Toevaluatethisclaim,IusethenewdatatocategorizeallUS?rmswithmajority-ownedmanufacturingplantsanywhereintheworldintofourcategories:1)domesticman-ufacturing?rmswithoutanymajority-ownedforeigna?liates;2)USmultinationalenterprisesthathaveonlyUSmanufacturingplants(theirforeigna?liatesareoutsidemanufacturing);3)USmulti-nationalenterprisesthathavebothUSandforeignmanufacturingplants;and4)USmultinationalenterprisesthathaveonlyforeignmanufacturingplants.(Allthese?rmtypesmayalsooutsourcesometasksfromdomesticorforeignsuppliers.)

The?rstrowofTable

1

presentsthenumberofUS?rmsthatmanufacturein-housein2007acrossthesefourcategories.Ofthe243,700USmanufacturing?rms,only1,700havemajority-ownedforeignestablishments(columns2to4).Amongthesemultinationals,1,200?rmsownUSandforeignmanufacturingplants,versus350?rmswithjustdomesticplants,andonly150?rmswithexclusivelyforeignin-housemanufacturing.FirmswithbothdomesticandforeignmanufacturingplantsarethusthemostprevalenttypeofUSmultinationalmanufacturingenterprise.

PanelAofTable

1

presenttotalsalesforthese?rms.The?rstrowcontainsglobalsales,whicharethesumof?rms’USandforeign-establishmentsales,eachofwhichispresentedseparatelyinthenexttworows.Iinclude?rms’totalsaleshere,regardlessofwhethertheyarebookedbymanufacturingornon-manufacturingestablishments.

Thesalesdatadelivertwostarkmessages.First,US?rmswithbothdomesticandforeignmanu-

7

Table1:Sales,employment,andtrade?owsforallUS?rmsthatmanufacturein-house

(1)

(2)

(3)

(4)

FirmType:

Domestic

MNE

MNE

MNE

Majority-OwnedManufacturingPlantsIn:

USOnly

USOnly

US&Foreign

ForeignOnly

Firms

242,000

350

1,200

150

PanelA:Sales($billions)

GlobalSales

2,629

1,695

6,710

345

SalesbyUSEstabs

2,629

1,446

3,853

173

SalesbyForeignEstabs

-

249

2,857

172

PanelB:Employment(thousands)

GlobalEmployment

11,059

5,338

11,883

732

EmploymentinUSEstabs

11,059

4,349

6,556

361

EmploymentinForeignEstabs

-

989

5,327

371

PanelC:USTradeFlows($billions)

Imports

126

39

410

12

Arm’s-Length

89

33

160

6

Related-Party

37

7

250

6

Exports

123

22

437

3

Arm’s-Length

103

16

253

2

Related-Party

19

5

184

1

Source:2007LongitudinalBusinessDatabase,EconomicCensuses,LongitudinalFirmTradeTransactionsDatabase,BEAinwardandoutwardsurveys.Tablepresentstotalnumberof?rmsandtheirglobalsales,globalemployment,andUSmerchandisegoodtrade?owsby?rmtypeandmanufacturingplantlocations.SampleisallUS?rmswithoneormoremajority-ownedmanufacturingplantsanywhereintheworld.

8

facturingplantsdominatebothglobalandUSsales,withglobalsalesof$6.7trillion–morethantheotherthreecategoriescombined–despitethefactthattheyareonly1,200outofthe243,700?rmsinthesample.Second,USmanufacturersthatonlyproducein-houseinforeignplantsaccountforamere3percentUSmanufacturers’globalsales.

Thedominanceof?rmswithbothUSandforeignin-houseproductionisreinforcedby?rms’employmentdi?erences.PanelBshowsthattransnationalmanufacturers–thosethatperformin-housephys