www.elsevier.com/locate/econbase
Doesforeigndirectinvestmenttransfertechnologyacrossborders?
Newevidence☆JürgenBitzera,⁎,MonikaKerekesbaUniversityofAppliedSciencesOOWatEmden,Germany
bFreeUniversityBerlin,Germany
Received6December2006;receivedinrevisedform29January2008;accepted26February2008
Availableonline5March2008
Abstract
Basedonindustry-leveldataofseventeenOECDcountriesweexamineFDIasapotentialchannelforknowledgediffusion.WefindthatFDI-receivingcountriesbenefitstronglyfromFDI-relatedknowledgespillovers.Wedonotfindevidenceforpositiveoutward-FDI-relatedtechnologysourcingeffects.
©2008ElsevierB.V.Allrightsreserved.
Keywords:ForeignDirectinvestment;InternationalknowledgespilloversJELclassification:F21;O33;O47;E23
1.Introduction
Theanalysisofpotentialchannelsforinternationalknowl-edgediffusionhasalonghistory.Sincetheearlynineties,anincreasingnumberofempiricalstudieshaveattemptedtodeter-minetheextenttowhichtradeisasourceforknowledgespill-overs(e.g.CoeandHelpman,1995;Keller,1999;BitzerandGeishecker,2006,tomentiononlyafew).ButthequestionofwhetherFDIisachannelforknowledgediffusionhasre-ceivedcomparativelylittleattention.Researchonthistopichasremainedlimitedtocountrystudies(e.g.Blomström,1986;Kokko,1994)andmicro-econometricanalyses(e.g.AitkenandHarrison,1999;Javorcik,2004;GörgandStrobl,2003).1Tothebestofourknowledge,theonlypanelcountrystudyanalysingtheroleofFDIininternationalknowledgediffusionisthatof
☆WearegratefultoIngoGeishecker,HolgerGörg,PhilippJ.H.SchröderandtheparticipantsoftheAnnualConferenceoftheEuropeanTradeStudyGroup(ETSG)inDublinforvaluablecomments.Theusualdisclaimerapplies.
⁎Correspondingauthor:DepartmentofBusinessStudies,Constantiaplatz4,26723Emden,Germany.Tel.:+49498071207;fax:+49498071228.
E-mailaddress:jbitzer@fh-oow.de(J.Bitzer).1Forreviewsoftheliterature,seeGörgandGreenaway(2004),Keller(2004).0165-1765/$-seefrontmatter©2008ElsevierB.V.Allrightsreserved.doi:10.1016/j.econlet.2008.02.029
vanPottelsberghedelaPotterieandLichtenberg(2001)here-afterreferredtoasPL2001.TheirmoststrikingresultisthatinwardFDIdoesnotaffecthostcountryproductivityviainter-nationalR&Dspillovers,whileoutwardFDIintoR&D-intensivecountries,ontheotherhand,createssignificanttech-nologyspillovers.Thus,basedontheirdata,theyconcludethatFDItransfersknowledgeinonlyonedirection.
UsinganewlycompileddatasetontenmanufacturingsectorsinseventeenOECDcountriesduringtheperiod1973–2000,weprovidenewevidenceonthequestionofwhetherFDItransfersknowledgeacrossboarders.Furthermore,weexpandtheapproachofPL2001intwodirections.First,toaccountforFDI-relatedspilloverswhichmightoccuronlyinthemediumorlongrun,weuseFDIstocksinsteadofFDIflowsforconstructingourFDI-weightedforeignR&Dcapitalstocks,assuggestedbyPL2001.2Second,weexpandthefocusofPL2001beyondbilateralFDI-relatedspilloverstoaccountforthird-countryeffects,capturing,forexample,knowledgeflowsfromcountryCintocountryB,andthenviacountryB'sFDI
2SeePL2001,p.492.
356J.Bitzer,M.Kerekes/EconomicsLetters100(2008)355–358
intocountryA.3OurmainfindingisthatitisinwardFDIthatactsasachannelforknowledgespillovers.IncontrasttoPL2001,wedonotfindevidenceforpositiveFDI-relatedtechnologysourcingeffects.Inaddition,weareabletoshowthatfornon-G7countries,theoutputelasticityofoutward-FDI-weightedforeignR&Dcapitalisnegative.
2.EmpiricalimplementationandestimationtechniqueInordertoanswerthequestionofwhetherFDItransferstechnologyacrossborders,weuseindustrydataonseventeenOECDcountriesinthetimeperiod1973–2000.4WeapplythefollowingstandardCobb–Douglasproductionfunctionapproach:
lnQcit¼aiþb1lnSdctþb2lnSfmctþb3lnSffctþb4lnSft
ct
ð1Þ
þb5lnKcitþb6lnLcitþb7lnMcitþatDtþmcit;
whereQisgrossoutput,SdisthedomesticR&Dcapitalstockofthemanufacturingsector,SFistheexternalforeignR&Dcapitalstock,Lislabour,Kisphysicalcapital,Mismaterialandinter-mediateinputsandDtisafullsettimedummies.Subscriptsi,c,andtdenotesectors,countries,andyearsrespectively.
Thephysical,domesticR&DandFDIcapitalstocksarecon-structedusingtheperpetualinventorymethodwithanassumeddepreciationrateof10%.TheinitialstocksareestimatedusingtheproceduredescribedinGotoandSuzuki(1989).
Theimport-weightedR&DcapitalstockSfmisobtainedusingthefollowingformula:
SfmmctXct
¼dQSctlt;ð2Þlpcwheremctdenotesimports.
WeuseFDIcapitalstocksratherthanflowsintheconstructionoftheFDI-weightedforeignR&Dcapitalstocksbecausestockscapturemediumandlong-runeffectsofknowledgediffusion.Wetakethird-countryeffectsintospecialconsideration.Takeforexampleasituationwhereamultinational'sheadquartersisincountryB,butitsR&DdepartmentisatasubsidiaryincountryC.EvenifcountryCdoesnotinvestincountryA,itsknowledgemaybetransferredtocountryAviaFDIfromcountryB.5Weaccountforthiseffectbyweightingtheworld-wideforeignR&Dcapitalstockwiththeoverallinvestmentactivityofforeignersinthereceivingcountry.TheoverallinvestmentactivityofforeignersmeasuredastheratioofthetotalinwardFDIcapitalstock(finis
ct)tothetotaldomesticphysicalcapitalstock(kct).Formally,the
3AsimilarpointismadebyLumenga-Nesoetal.(2005)regardingtrade.4Adetaileddescriptionofthedataisgiveninanappendixavailableuponrequest.5Considerthefollowingreal-worldexample.Opel,aformerGermancarmanufacturerandtodayasubsidiaryofGM,hasitsEuropeanheadquartersinSwitzerland,wherenoR&Discarriedout.ThemajorityofOpel'sR&DiscarriedoutinGermany.IfOpelinvestsinEasternEuropetheFDIwouldbeaccountedascomingfromSwitzerland.However,noR&Discarriedoutattheheadquarters.Indeed,anypositiveknowledgespilloverviatheFDIwouldbeduetotheR&DcarriedoutbyOpelinGermany.ThisindirectimpactwouldnotshowupinaspecificationusingbilateralFDIshares.
inward-FDI-weightedforeignR&Dcapitalstockisconstructedas
follows:
Sff¼finct
Xct
kSdctlt:ð3Þ
lpc
Besidesthird-countryeffects,inter-andintrasectoralspilloversfromFDIarecaptured.6Theoutward-FDI-weightedforeignR&DcapitalstocksareconstructedanalogouslytotheinwardFDI-weightedforeignR&Dcapitalstockssoastocapturethird-countryeffectsinknowledgeabsorption.Wemeasureacountry'sinvestmentacti-vityabroadasitstotaloutwardFDIcapitalstock(foutct)inrelationtoitstotaldomesticphysicalcapitalstock(kct)andusethisrelationtoweighttheforeignR&Dcapitalstock:
Sft¼foutctXct
d
kSctlt:ð4Þ
lpc
TheempiricalassessmentisbasedonthemodelgivenbyEq.(1).TheestimationshavebeencarriedoutasFGLS,withacorrectionforpanel-specificautocorrelationoftheformAR(1)andpanelheteroscedasticity,fixedeffectsandasetoftimedummies.Unitroottestsindicatedthatalltimeseriesaretrend-stationary.Furthermore,endogeneitytestscouldnotrejectthehypothesisthatinputsareexogenous.73.Results
WeestimateEq.(1)usingdifferentspecifications(seeTable1).InColumnI,thesoletransmissionchannelisassumedtobeimports.InColumnIIinwardFDIisadded,andinColumnIIIinwardandoutwardFDIareaddedasfurthertransmissionchan-nels.FurthermoreincolumnsIVtoVI,wecontrolforthepossi-bilitythatG7countriesfacedifferentoutputelasticitiesthanothercountries,asproposedbyCoeandHelpman(1995).Tothisend,alltransmissionmechanismsareinteractedwithadummyvaria-ble,whichtakesonthevalueofoneifthecountryisoneoftheG7countries.
Theoutputelasticitiesofthetraditionalinputfactorsphysicalcapital(K),labour(L),andmaterialandintermediateinputs(M)havetheexpectedsignsandarehighlysignificantinallspeci-fications.ThesameistruefordomesticR&Dcapitalstock(Sd),butcolumnsIVtoVIshowthatnon-G7countriesbenefitmorefromthedomesticR&DcapitalstockthanG7countries.Onepossibleinterpretationofthisresultisthatknowledgeexhibitsdecreasingreturnstoscale.Inaccordancewithotherstudies(e.g.CoeandHelpman,1995;BitzerandGeishecker,2006),
6TheuseofalternativeweightingschemesasproposedforexamplebyLichtenbergandvanPottelsberghedelaPotterie(1998)andinGuellecandvanPottelsberghedelaPotterie(2004)cannotbeimplemented,becausebilateralindustry-levelFDIdataarenotavailableforthetimeperiodandaggregationlevelused.7Thedetailsoftheunitrootandendogeneitytestsperformed,aswellasthetestresults,aredescribedinanappendixavailableuponrequest.
J.Bitzer,M.Kerekes/EconomicsLetters100(2008)355–358
Table1
FGLSestimationresultsforlevelsIndependentvariable
I
II
III
IV
V
VI
357
DependentvariableislnQ
DomesticR&DlnSdlnSd⁎G7
ForeignR&DlnSfm(imports)lnSfm⁎G7(imports)lnSff(inwardFDI)lnSff⁎G7(inwardFDI)lnSft(outwardFDI)lnSft⁎G7(outwardFDI)lnKlnLlnM
Waldχ2(df)p-valueWaldχ2Obs.
0.0496⁎⁎⁎(0.0044)0.0350⁎⁎⁎(0.0043)0.0449⁎⁎⁎(0.0045)
0.0593⁎⁎⁎(0.0050)0.0406⁎⁎⁎(0.0056)0.0578⁎⁎⁎(0.0057)−0.0368⁎⁎⁎(0.0063)−0.0261⁎⁎⁎(0.0070)−0.0489⁎⁎⁎(0.0074)
−0.0092⁎(0.0052)0.0316⁎⁎⁎(0.0046)0.0153⁎⁎⁎(0.0022)−0.0054⁎⁎(0.0023)0.0351⁎⁎⁎(0.0030)0.1456⁎⁎⁎(0.0057)0.7880⁎⁎⁎(0.0037)2.10e+09(207)0.0003220
0.0205⁎⁎⁎(0.0042)0.0138⁎⁎⁎(0.0041)0.0088⁎⁎(0.0042)
0.0114⁎⁎⁎(0.0011)0.0120⁎⁎⁎(0.0011)
−0.0106⁎⁎⁎(0.0017)
0.0323⁎⁎⁎(0.0032)0.1458⁎⁎⁎(0.0056)0.7889⁎⁎⁎(0.0037)1.24e+09(203)0.0003220
0.0328⁎⁎⁎(0.0031)0.1488⁎⁎⁎(0.0054)0.7869⁎⁎⁎(0.0036)2.18e+09(204)0.0003220
0.0259⁎⁎⁎(0.0031)0.1494⁎⁎⁎(0.0055)0.7899⁎⁎⁎(0.0037)2.28e+09(205)0.0003220
0.0009(0.0054)0.0262⁎⁎⁎(0.0044)
0.0344⁎⁎⁎(0.0032)0.1468⁎⁎⁎(0.0059)0.7878⁎⁎⁎(0.0038)1.06e+09(205)0.0003220
−0.0024(0.0055)0.0204⁎⁎⁎(0.0049)0.0210⁎⁎⁎(0.0021)−0.0113⁎⁎⁎(0.0024)−0.0199⁎⁎⁎(0.0022)0.0205⁎⁎⁎(0.0029)0.0260⁎⁎⁎(0.0027)0.1536⁎⁎⁎(0.0053)0.7879⁎⁎⁎(0.0036)9.15e+08(209)0.0003220
Remarks:Thedependantvariableislogoutput.Fixedandtime-specificeffectsareincludedandgroupwisesignificantatthe1%level.Robuststandarderrorsareinparentheses.⁎⁎⁎,⁎⁎,⁎indicateasignificanceatthe1%,5%and10%levels,respectively.
averagespilloversviaimportsarepositive(seecolumnsItoIII).However,theresultsincolumnsIVtoVIindicatethatthiseffectrelatessolelytoG7countries.
AsamajordeparturefromPL2001,wefindthatinwardFDIgeneratesstatisticallyhighlysignificantpositiveknowledgespilloversinallmodelspecifications.8Non-G7countriesbenefitmorefrominwardFDIthanG7countries(seecolumnsVandVI),aresultwhichagainhighlightstheimportanceofdomesticknowledgestocks.
UnlikePL2001,wecannotidentifypositivespilloversgene-ratedbyoutwardFDI(columnsIIIandVI).Rather,wefindthatoutward-FDI-weightedforeignR&DcapitalhasanegativeimpactontheoutputoftheFDIsourcecountry.9However,controllingfordifferenteffectsforG7andnon-G7countries,itturnsoutthatonlynon-G7countriesfaceahighlysignificantnegativeoutputelasticity,whileG7countriesneithersuffernorbenefitfromoutward-FDI-weightedforeignR&Dcapital.104.Conclusions
Inthispaper,weprovidenewevidenceonthequestionofwhetherFDItransferstechnologyacrossborders.Usinganewlycompileddatasetontenmanufacturingsectorsinseventeen
OECDcountriesduringtheperiod1973–2000,weareabletoaccountforthird-countryeffectsandthelong-termnatureofFDIintransferringknowledge.
ApplyingastandardCobb–Douglasproductionfunctionapproach,wefindpositiveandstatisticallyhighlysignificantknowledgespilloversstemmingfrominwardFDI.Incontrasttoformerstudies,wedonotfindpositiveinternationalknowledgespilloverstransferredviaoutwardFDI.Rather,outward-FDI-weightedR&Dcapitalhasastatisticallysignificantnegativeimpactontheproductionofnon-G7countriesinvestingabroad.References
Aitken,B.J.,Harrison,A.E.,1999.Dodomesticfirmsbenefitfromdirect
foreigninvestment?EvidencefromVenezuela.AmericanEconomicReview89,605–618.
Bitzer,J.,Geishecker,I.,2006.Whatdrivestrade-relatedR&Dspillovers?
Decomposingknowledge-diffusingtradeflows.EconomicsLetters93,52–57.
Blomström,M.,1986.Foreigninvestmentandproductiveefficiency:thecaseof
Mexico.JournalofIndustrialEconomics35,97–112.
Coe,D.T.,Helpman,E.,1995.InternationalR&Dspillovers.EuropeanEconomic
Review39,859–887.
Görg,H.,Greenaway,D.,2004.Muchadoaboutnothing?Dodomesticfirms
reallybenefitfromforeigndirectinvestment?WorldBankResearchObserver19,171–197.
Görg,H.,Strobl,E.,2003.Multinationalcompanies,technologyspilloversand
plantsurvival.ScandinavianJournalofEconomics105,581–595.
Goto,A.,Suzuki,K.,1989.R&Dcapital,rateofreturnonR&Dinvestmentand
spilloverofR&DinJapanesemanufacturingindustries.ReviewofEco-nomicsandStatistics71,555–564.
Guellec,D.,vanPottelsberghedelaPotterie,B.,2004.FromR&Dtoproduc-tivitygrowth:dotheinstitutionalsettingsandthesourceoffundsofR&Dmatter?OxfordBulletinofEconomicsandStatistics66,353–378.
Javorcik,B.S.,2004.Doesforeigndirectinvestmentincreasetheproductivityof
domesticfirms?Insearchofspilloversthroughbackwardlinkages.AmericanEconomicReview94,605–627.
Keller,W.,1999.Howtradepatternsandtechnologyflowsaffectproductivity
growth.NBERWorkingPaper,No.6990.
Itwouldhavebeendesirabletochecktherobustnessoftheresultsusingdifferentweightingschemes,howeverduetomissingbilateralindustry-levelFDIdataforthetimeperiodandaggregationlevelusedthishastobeleftforfutureresearch.9Theseresultsholdeveniftheoutward-FDI-weightedforeignR&DcapitalstockvariableistheonlyforeignR&Dcapitalstockweightedvariableintheregression.Theestimationsarereportedinaseparateappendixavailableuponrequest.10AWaldtestshowsthatthesumofthecoefficientsfornon-G7andG7-countriesarenotsignificantlydifferentfromzero.
8358J.Bitzer,M.Kerekes/EconomicsLetters100(2008)355–358
Lumenga-Neso,O.,Olarreaga,M.,Schiff,M.,2005.On‘indirect’trade-relatedR&Dspillovers.EuropeanEconomicReview49,1785–1798.
vanPottelsberghedelaPotterie,B.,Lichtenberg,F.,2001.Doesforeigndirectinvestmenttransfertechnologyacrossborders?ReviewofEconomicsandStatistics83,490–497.
Keller,W.,2004.Internationaltechnologydiffusion.JournalofEconomic
Literature42,752–782.
Kokko,A.,1994.Technology,MarketCharacteristics,andSpillovers.Journalof
DevelopmentEconomics43,279–293.
Lichtenberg,F.,vanPottelsberghedelaPotterie,B.,1998.InternationalR&D
spillovers:acomment.EuropeanEconomicReview42,1483–1491.
因篇幅问题不能全部显示，请点此查看更多更全内容