The Impact of Outsourcing and High-Technology Capi

THEIMPACTOFOUTSOURCINGANDHIGH-TECHNOLOGYCAPITALONWAGES:ESTIMATESFORTHEUNITEDSTATES,1979-1990*Revised,September1998RobertC.FeenstraandGordonH.HansonDepartmentofEconomics,DepartmentofEconomicsandUniversityofCalifornia,Davis,SchoolofBusinessAdministrationHaasSchoolofBusiness,UniversityofMichiganUniv.ofCalifornia,Berkeley,andNBERandNBER*TheauthorsthankJamesAnderson,RobertBaldwin,AlanDeardorff,JamesHarrigan,LawrenceKatz,EdwardLeamer,DavidRichardson,andMatthewSlaughterforveryusefulcomments.FinancialsupportfromtheNationalScienceFoundationisgratefullyacknowledged.TheImpactofOutsourcingandHigh-TechnologyCapitalonWages:EstimatesfortheUnitedStates,1979-1990AbstractWeestimatetherelativeinfluenceoftradeversustechnologyonwagesina“largecountry”setting,wheretechnologicalchangeaffectsproductprices.Tradeismeasuredbytheforeignoutsourcingofintermediateinputs,whiletechnologicalchangeismeasuredbyexpendituresonhigh-technologycapitalsuchascomputers.Inourinitialspecification,wefindthatcomputersexplainabout35percentoftheincreaseintherelativewageofnonproductionworkers,whileoutsourcingexplainsatmost15percent.Inanalternativespecification,outsourcingexplainsabout40percentoftheincreaseintherelativenonproductionwage,whereascomputerexpenditurescanexplain75percentofthisincrease.1I.IntroductionTherecenteconomicperformanceofless-skilledworkersinindustrialcountriesisanimportantpolicytopicandthesubjectofintenseacademicattention.Duringthe1980sand1990s,thewagesoflow-skilledworkershavefallenbothinrealtermsandrelativetothoseofhigh-skilledworkers.Thetwomostwidely-citedexplanationsfortheriseinwageinequalityareskill-biasedtechnicalchangeandtradewithlow-wagecountries.Ofthesetwo,technicalchangeduetotheuseofcomputersisoftenbelievedtobethedominantexplanation.Thegoalofthispaperistodevelopanewmethodologyandestimatetheimpactoftradeandtechnologyonwages,fortheUnitedStatesovertheperiod1979-1990.Wewillmeasuretradebytheforeignoutsourcingofintermediateinputs,1whilewewillmeasurepotentialtechnicalchangebytheshifttowardshigh-technologycapitalsuchascomputers.Thestartingpointforouranalysisisapopularmethodtopredictwagechangesunderzero-profits:aregressionofthechangeinindustrypricesontheleveloffactorcost-sharesinthatindustry,wheretheestimatedcoefficientsareinterpretedasthepredictedchangeinfactor-pricesthatareconsistentwiththemovementinproductprices.This“priceregression”wasfirstusedbyBaldwinandHilton[1984],andmorerecentlybyLeamer[1994,1998],BaldwinandCain[1997],andKrueger[1997].Incontrasttoexistingliterature,wearguethatwhenfullyspecified,thisregressionbecomesanidentityandcannotofferanypredictionoftheimpliedchangesinfactorprices,otherthanthatwhichactuallyoccurred.1ForeignoutsourcingwasfirstconsideredbyLawrenceandSlaughter[1993]andmorerecentlybyFeenstraandHanson[1996a,1996b].LawrenceandSlaughter[1993]andBerman,BoundandGriliches[1994]arguethattheamountofoutsourcingfromtheU.S.istoosmalltoexplainthechangeinwages,butthiswasduetothenarrowmeasureofoutsourcingthattheyused[seeFeenstraandHanson,1996a,pp.106-107].WewillbeusingameasureofoutsourcingconstructedasinFeenstraandHanson[1996b],whichisestimatedimportsofintermediateinputsintoeachindustry.Thismeasuremayalsomissaspectsofoutsourcing,suchastheuseofcomputerprogrammersinIndiaforproductsotherwisemanufacturedintheU.S.Leamer[1998]introducesthebroaderterm“delocalization”toindicatethemanywaysthatpiecesoftheresearch/production/marketingprocessescanbemovedoffshore.2Tomovebeyondthisstalemate,weshallmodifytheconventionalpriceregressionusingatwo-stageestimationprocedure.First,weexaminehowchangesinstructuralvariables,suchasforeignoutsourcingandhigh-technologycapital,affectindustrypricesandproductivity.Bytreatingindustryprices(andproductivity)asendogenous,weallowforalarge-open-economysetting.Fromthesefirst-stageresults,wedecomposepriceandproductivitychangesintoportionsthatareattributabletoeachstructuralvariable.Second,usingamodifiedversionofthepriceregression,weusethedecomposedpriceandproductivitychangesfromthefirststagetoestimatethechangeinprimaryfactorpricesthatisattributabletoeachstructuralvariableseparately.Theresultsindicatehowmuchoftheobservedriseinwageinequalityisattributabletoforeignoutsourcingorhigh-technologycapital.Whilewefocusonthesetwoexplanations,themethodologywedevelopisquitegeneralandcouldbeusedtoexaminetherelationshipbetweenfactorpricesandmanytypesofchangesinproductiontechniques.Ourapproachmayhelpresolveanapparentconflictintheliteratureoverwhetheritisthefactorbiasorthesectorbiasoftechnologicalchangesthatmattersforwages.2Krugman[1995]andLeamer[1998]havedebatedthispoint,withKrugmanarguingthatfactorbiasisimportantinaclosedorlargeopeneconomy,andLeamerarguingthatsectorbiasisallthatmattersinasmallopeneconomy(orevenwithlog-linearpass-throughfromproductivitytoprices).Toresolvethis,weneedtohaveaindicationofwhichsettingisempiricallyrelevant.Thiswillturnouttobeabyproductofouranalysis,sinceourfirst-stageregressioncandistinguishbetweensector-biasedandfactor-biasedtechnologicalchanges:bothofthesechangesaffectindustryprices,but(withCobb-Dougla