HeterogeneityandtheFDIversusExportDecision

HeterogeneityandtheFDIversusExportDecisionofJapaneseManufacturers∗KeithHead†JohnRies‡October28,2003Weinvestigatewhetherproductivitydifferencesexplainwhysomemanufacturerssellonlytothedomesticmarketwhileothersserveforeignmarketsthroughexportsand/orFDI.Whenoverseasproductionoffersnocostadvantages,ourmodelpredictsthatinvestorsshouldbemoreproductivethanexporters.Anextensionallowingforlow-costforeignproductioncanreversethisprediction.Datafor1070largeJapanesefirmsrevealthatfirmsthatinvestabroadandexportaremoreproductivethanfirmsthatjustexport.Amongoverseasinvestors,moreproductivefirmsspanawiderrangeofhost-countryincomelevels.JELclassification:F120,F230Keywords:FDI,exports,productivity,heterogeneousfirms∗WethankKyojiFukao,JotaIshikawa,ananonymousreferee,andparticipantsatthe15thAnnualTRIOConferenceheldinTokyo,December10–11,2002fortheirhelpfulcomments.†SauderSchoolofBusiness,UniversityofBritishColumbia,2053MainMall,Vancouver,BC,V6T1Z2,Canada.Tel:[phone],Fax:[phone],Email:[e-mail]‡SauderSchoolofBusiness,UniversityofBritishColumbia,2053MainMall,Vancouver,BC,V6T1Z2,Canada.Tel:[phone]Email:[e-mail]1IntroductionTheoreticalandempiricalresearchhasrevealedthefactorsthatunderliethedecisionoffirmstoserveforeignconsumersthroughforeigndirectinvestment(FDI)orexporting.EvidenceshowsthatfirmspreferFDItoexportingwhentradecostsarehighandplant-levelscaleeconomiesarelow.Inanygivenindustry,however,weobservethatsomefirmsengageinFDI,someexport,andsomedoneither.Thereislimitedunderstandingofheterogeneousactionsoffirmsfacingthesametradeandinvestmentopportunities.ThispaperbuildsonresearchofHelpman,Melitz,andYeaple(forthcoming),here-afterHMY,examiningtheroleofdifferencesintheproductivitylevelsoffirmsinanindustryasexplanationsforthepresenceofpurelydomesticfirms,exportingfirms,andinvestingfirms.WedevelopanalternativemodelthatyieldsthepredictionsfoundbyHMYandthentestthesepredictions.Wealsoextendthemodelandshowthattheorder-ingpredictedbyHMYcanbereversediftheforeigncountryisthelow-costproductionsite.Wecategorize1070publiclytradedJapanesefirmsin1989intogroupsaccordingtowhethertheyexportandhaveforeigndirectinvestment.Thenweusevariousmeasuresoftheproductivityofthesefirmstoinvestigateproductivitydifferencesacrossthesegroups.OuranalysiscontributestotheliteratureinvestigatingtheexportversusFDIdecisionoffirms.Brainard(1997)showshowtradecosts,marketsize,andplant-leveleconomiesofscaleinteracttoexplaintheexportandFDIdecisionoffirmsproducingdifferentiatedproducts.Sheshowsthattheshareoffirmsthatexportinanindustryisanincreasingfunctionofscaleeconomiesandadecreasingfunctionoftradecostsandforeignmarketsize.UsingU.S.exportsandforeignaffiliatesalesdatafor63industriesin27countries,1Brainardfindsthatexportsharesdecreasewithtradeandfreightcostsandincreasewithplant-levelscaleeconomies.Brainardidentifiesapossibleequilibriumwheresomefirmsexportandothersdoforeigninvestment.Hertheory,however,simplyspecifiestheshareoffirmsdoingeachactivityanddoesnotpredictwhichfirmsdowhichactivity.HMYextendthislineofresearchbyidentifyingaroleforproductivitydifferencestoexplainthechoice.HMYdevelopafreeentrymodelwherefirmshaveheterogeneousproductivitylevelsandeachproduceadifferentiatedproduct.ConsumershaveCESpreferencesandgoodsaresubjectto“iceberg”transportationcosts.Theyfindthattheleastproductivefirmsdonotservetheforeignmarket,themostproductivefirmsengageinFDI,andthemid-rangefirmsexport.TheirpaperextendsMelitz(forthcoming)thatshowsthatonlythemostproductivefirmsinanindustryexport.Bernard,Eaton,Jenson,andKortum(2003)useaRicardianmodeloftradeindifferentiatedproductsanddemonstratethatthemostproductivefirmsinanindustryexport.Neitherofthelatterpapersconsidertheoptiontoreachforeignconsumersthroughdirectinvestment.BernardandJenson(1999)measurethetotalfactorproductivity(TFP)ofU.S.man-ufacturingplantsovertheperiod1984–1992.TheyfindthatexportingplantsexhibithigherTFPthannon-exporters(aftercontrollingforindustryandstatefixedeffects)andproductiveplantsaremostlikelytobecomeexporters.BernardandJensonalsoinvestigatewhetherexporting“causes”greaterTFPbyexaminingproductivitygrowthofexportersandnon-exporters.TheirresultsshowingthatexportershavesignificantlyslowerTFPgrowththannon-exportersdonotsupportthepropositionthatexportinggeneratesTFPgrowth.HMYanalyzetherelationshipbetweentheexporttoFDIratio2of4-digitU.S.manufacturingindustriesandthedispersionofsaleslevelsintheseindus-triesandfindsupportfortheirmodel.GirmaandKneller(2003)testthepredictionsofHMYusingdataon5,332firmsintheUnitedKingdom.Basedontestsofstochasticdominance,theyfindsupportforthepropositionthattheproductivitydistributionofmultinationalsdominatesthatofexportfirmswhichdominatesthatofnon-exporters.Weextendboththetheoreticalandempiricalliterature.ThenextsectiondevelopsasimplelineardemandmodelthatgeneratestheHMYpredictionsundertheassumptionthatfactorspricesareequalandthatfirmsserveforeignconsumerseitherbyexportingoropeningupanewmanufacturingfacilityintheforeigncountry.Section3explainshowweclassifythe1070Japanesemanufacturingfirmsintofourgroupsandidentifiesthedif