The Effects of Foreign DirectInvestment on Local Communities

doi:10.1006rjuec.2000.2170,availableonlineathttp:rrwww.idealibrary.comon

TheEffectsofForeignDirectInvestmenton

LocalCommunitiesU

DavidN.Figlio

DepartmentofEconomics,Uni¨ersityofFlorida,Gaines¨ille,Florida32611-7140

E-mail:figliodn@dale.cba.ufl.edu

and

BruceA.Blonigen

DepartmentofEconomics,Uni¨ersityofOregon,Eugene,Oregon97403-1285

E-mail:bruceb@oregon.uoregon.edu

ReceivedFebruary1,1999;revisedSeptember17,1999

LittleevidenceexistsontheeffectsofforeigndirectinvestmentŽFDI.onlocalcommunitiesintheUnitedStates,despiteevidencethatU.S.communitiesactivelybidagainsteachotherforFDI.Weusedetailedcounty-levelpaneldatafromSouthCarolinaacross5-yearintervalsfrom1980through1995toinvestigatetheeffectofforeignmanufacturingplantsonlocallabormarketsandonthelevelanddistributionoflocalgovernmentbudgets.Wefindthatforeigninvestmentraiseslocalrealwagesmuchmorethandoesdomesticinvestment,butlowerspercapitacounty-governmentexpendituresandredistributesmoniesawayfrompublicschoolexpenditures.ᮊ2000AcademicPress

KeyWords:foreigndirectinvestment;localpublicfinance;labormarkets.

1.INTRODUCTION

ThecompetitionfornewfirminvestmentbystateandlocalgovernmentsintheUnitedStatesseemstobeeverincreasing.Theamountandvarietyofstateandlocalincentivestoattractfirmshaveprogressedtoincludelocalpropertytaxrelief,freeland,jobtaxcredits,‘‘enterprisezones’’whichgivefirmsgreaterbenefitsforlocatingineconomicallydepressedareas,

*WethankJanBrueckner,EckhardJaneba,JohnList,Kei-MuYi,twoanonymousreferees,andsessionparticipantsatthe1998WinterEconometricSocietymeetingsforhelpfulcomments.WealsothankthestaffoftheSouthCarolinaStateLibraryandClariceGrayfortheirexcellentresearchassistance.WegratefullyacknowledgefinancialsupportfromtheCollegeofArtsandSciencesoftheUniversityofOregon.Anyerrorsoromissionsarethesoleresponsibilityoftheauthors.

338

0094-1190r00$35.00

Copyrightᮊ2000byAcademicPress

Allrightsofreproductioninanyformreserved.LOCALEFFECTSOFFOREIGNDIRECTINVESTMENT

339

andmajorinfrastructureimprovements.Beyondthesenow‘‘standard’’incentives,competingstatesalsospendsignificantresourcestailoringspe-cializedincentivepackagesforpotentiallylargeinvestments.

Thiscompetitionhasledbothpolicymakersandeconomiststoquestionwhetherthecompetitivebiddingforinvestmentbylocalcommunitiesisactuallyharmful.Themainconcernisthatvariouslocalitiesmayendupinabiddingwarthatresultsina‘‘prisoner’sdilemma’’thatbenefitsthefirmattheexpenseofthewinningcommunityandthewelfareoftheentirecountry.Infact,totheextentthatcommunitieshaveacommonvaluationoftheplantlocatedintheirarea,eventhelocalcommunitythatreceivestheinvestmentmaysuffera‘‘winner’scurse,’’becauseitbidtoomuch.ThemostnotablerecentincidentmaybetheincentivesofferedbytheStateofAlabamatoattractMercedes-BenzAGintheearly1990s.AsdetailedbyaNovember24,1993articleintheWallStreetJournal,Alabamaendeduppromisingover$300millioninincentivestoMercedes,includingfreeland,employeesalariesthefirstyearofoperation,propertytaxrelief,payrolltaxcredits,statespendingonMercedesautomobiles,etc.ThisledmanytobelieveAlabamamayhavepaidtoohighaprice,asexpressedbyGeorgeAutryw28x,headofaneconomicdevelopmentgroupbasedinNorthCarolina:‘‘They’rewAlabamaxlosingmoneytoinvestintheirpeople,theirroads,theirstateingeneral.ForastatelikeAlabama,whichneedsmoneyforeducation,that’saproblem.’’Asthisexampleshows,evensomepolicymakersunderstandthatthesecompetitionsinvolvesignificanttrade-offs.

ThelargewaveofforeigndirectinvestmentŽFDI.flowsintotheUnitedStatesinthepasttwodecadesaddsanotherdimensiontothecompetitionforinvestment.1Beyondthepotentialadversewelfareeffectsdescribedabovefromstateandlocalcompetition,foreignfirms’gainsfromtheincentivesaccruetocapitalownersthatlikelyresideprimarilyoutsidetheUnitedStates.Inaddition,foreignplantsmaybelessinvolvedinthelocalcommunityŽe.g.,throughcharitablegiving.thandomesticones,whichcouldlessenlocalbenefitsfromtheinvestment.2Theseissues,alongwith

ResearchonFDIintotheUnitedStateshasmainlyexaminedthepossiblereasonsforthelargeinflowinthelate1980sandearly1990sandlocationdecisions,includingagglomerationissuesaddressedinHead,Ries,andSwensonw18x.GrahamandKrugmanw17xprovidesasurveyofpossibleexplanationsforthewaveofinwardFDIintheU.S.,andrelatedliterature.MuchlesshasbeendonetoexaminetheimpactsoftheforeignfirmsandplantsontheUnitedStates,particularlyatalocallevel.Therearesomeexceptions,besidesthepaperswediscussbelow,thatexaminenationwideandindustrywageeffectsofforeigninvestment.TheseincludeGrahamandKrugmanw17xandBlomstrom¨w3x,whichdiscusseconomy-wideimpactsofFDIfortheUnitedStates,Cavesw5xwhichexaminesproductivityeffectsofFDI,andBlonigenandFigliow4xwhichfindslocalFDIaffectslegislators’decisionsontradepolicy.2

Inthispaper,theterms‘‘foreign’’and‘‘domestic’’plantsreferstoownershiplocation,notgeographiclocationoftheplant.

1

340

FIGLIOANDBLONIGEN

thepotential‘‘prisoner’sdilemma’’problemwithstateincentives,haveevenledsometorecommendaU.S.governmentbanonstateincentivestoforeigninvestorsŽGlickmanandWoodwardw15xandGrahamandKrug-manw17x..

Yet,statesseemtobeparticularlycompetitiveintryingtoattractforeignplantstotheirarea.Thebiddingforforeignautomobileplantshasbeenwelldocumentedinthepopularpress,butthereisalsoindicationthatstatesmaybemoregenerallyinterestedininvestmentbyforeignplants.Woodwardw27xpointsoutthatmanyU.S.stateshaveoverseastradeanddevelopmentofficeswhichareintendedtobothpromotethestate’sexportsandattractforeigninvestment,anddatabytheNationalAssociationofStateDevelopmentAgenciesshowsthatover75%oftheseexpendituresareoneffortstoattractforeigninvestment.Thisissignificantsincetherearenocomparableexpenditurestoattractinvestmentbydomesticplants.

Whiletheevidenceabovesuggeststhatlocalcommunitiesmayviewforeigninvestmentdifferently,itmaynotbeclearatfirstglancewhyonewouldexpectforeignplantstohavedifferentimpactsoncommunitiesthandodomesticplants.Studiesexaminingtheeffectofnewmanufacturingjobsonlocalcommunities,andparticularlytheirimpactonlocallabormarkets,3maybesufficientforgaugingthelocaleffectofmanufacturingFDI.However,manyrecentstudieshavefoundevidencethattheremaybesignificanteconomicdifferencesbetweenforeignanddomesticestablish-ments.HowensteinandZeilew19xuseplant-leveldatafromtheAnnualSurveyofManufacturesfor1989and1990andfindthatforeignaffiliatesintheUnitedStatesarelarger,morecapitalintensive,andpayhigherwagesthandomesticplants.Globerman,Ries,andVertinskyw16xfindqualitativelyidenticalresultstothoseofHowensteinandZeilew19xusingdataonforeignaffiliatesanddomesticplantsinCanada.DomsandJensenw8xexaminemanufacturingplant-leveldataintheU.S.andfindthathigherwagesandproductivityarenotjustparticulartoforeignaffiliates,buttoanyplantsconnectedwithamultinationalfirm,foreign-orU.S.-owned.Aitken,Harrison,andLipseyw1xfindthatwagedifferentialsbetweendomesticandforeignenterprisesaremoresubstantialindataonMexicanandVenezuelanenterprisesthanforU.S.enterprisedata,particularly

TheseincludeMarstonw21x,Topelw26x,Crihfieldw7x,Bartikw2x,TerklaandDoeringerw25x,andGlaeseretal.w13x.Bartikw2xgivesperhapsthemostcomprehensiveoverviewoftheeffectsoflocaljobgrowth.Bothhisextensivereviewoftheliteratureandhisownanalysisfindthatlocaljobgrowthhasapositiveandsignificantlong-runimpactonrealearningsinthecommunity.Healsofindsevidencethatlong-rununemploymentratesdecreaseandhousingpricesincreasewithlocaljobgrowth.

3

LOCALEFFECTSOFFOREIGNDIRECTINVESTMENT

341

whencontrollingforindustry,size,andcapitalintensitydifferences.4Finally,FelicianoandLipseyw10xfindevidencethatforeignownershipleadstohigherwagesinU.S.manufacturingestablishmentsaftercontrol-lingforindustryandstatecharacteristics.Thus,thesestudiesprovidesomeevidencethatforeignplantspayhigherwages,whichcanexplainwhystatesmaybecompetingmoreintenselyforforeignplantsthancompara-bledomesticones.Inaddition,ouranalysisbelow,whichemployscounty-levelrealwagedataforthefirsttimetoaddressthisissue,providesfurthersupportforsubstantialdifferencesbetweenforeignanddomesticinvest-ment.

Whilehigherrealwagesandemploymentseemtobethemotivationforstates’competitionforinvestment,thereisalsotheexpectationthatthetradeoffforthesebenefitsisadverseimpactstogovernmentbudgets.Aswithwagesthough,onemightnotexpectanydifferentialimpactonbudgetsdependingonwhethertheplantisforeignordomestic:stateincentivesandbiddingwarsoccurwithlargeinvestmentsbydomesticplantsaswell.However,asdescribedabove,stateandlocalcompetitionappearsstrongerforforeigninvestments,andthiswouldthenpresumablyleadtolargerimpactsonthebiddingcommunities.Inaddition,anecdotalevidencesuggeststhatforeignplantsmaybeinterestedindifferenttypesofincentivesthanaredomesticones,whichmayalsoaffectthecomposi-tionoflocalbudgetsaccordingly.AnumberofWallStreetJournalarticlesonforeigninvestmentintheUnitedStateshavedetailedtheextensivefundingfortrainingandeducationprogramsthatforeignplantsŽespeciallyGermanplantsintheCarolinas.haverequestedandreceived.5Foreignplantsmayalsodifferfromdomesticonesintermsofroadsandotherinfrastructuretheyasklocalgovernmentstoprovide.

Thispaperexplorestheseissuesandprovidesnewevidencethatlocalcommunitiesfacedifferenttrade-offswhentryingtoattractforeignplantsratherthandomesticones.WeexaminetheeffectsofFDIonlocalcommunitiesusingdetailedcounty-levelpaneldatafromSouthCarolinaacross5-yearintervalsfrom1980through1995.WefocusonSouthCarolinaforseveralreasons.First,SouthCarolinahasasubstantiallevelofforeign-plantmanufacturingjobsrelativetototalmanufacturingem-ployment.Asdetailedbelow,FDIinSouthCarolinavariessignificantly

ArelatedpaperbyFeenstraandHansonw9xexaminestheimpactofFDIontherelativewagesofskilledandunskilledworkersinMexico.TheyfindthatFDI,andthelikelymoreskilled-intensiveproductionprocessesconnectedwithit,accountsforoverhalftheincreaseinskilledlaborwagesharethatoccurredinMexicointhelate1980s.5

Forexample,aWallStreetJournalarticle,May4,1993,reportsastate-funded212yeartrainingprogramforaRobertBoschGmbhplantinCharleston,SC.

4

342

FIGLIOANDBLONIGEN

acrossits46countieswiththepercentageofforeign-plantjobsineachcountyrangingfromzerotooverfiftypercent,makingSouthCarolinaanexcellentfocusforouranalysis.Inaddition,becauseofrelativelyhighlevelsofFDIhistorically,SouthCarolinahasexceptionallydetailedinfor-mationonforeignplantpresenceintheirstateoveralongtimeperiod.Second,asubstantialportionofincentivesofferedbySouthCarolinatoplantscomesfromlocalpropertytaxrelief,asopposedtoincentivesfundedatthestatelevel.Raineyw24xreportsthat54%ofallrevenuelossesfrombusinessincentiveprogramsinSouthCarolinaduring1996᎐1997Ž20.1outof$37.3million.werefromlocalpropertytaxreductions.Thismeansthatthebudgetimpactsfromtheinvestmentshouldsubstantiallyaffecttheunitofouranalysisᎏthelocalcounty.

Wefindsubstantialevidencethatforeigninvestmentraiseslocalrealwagesmuchmorethandomesticinvestment,butthiscomesatacostoflowerpercapitaexpendituresbythelocalcountyandbudgetaryredistribu-tionawayfrompublicschoolexpenditures.Inparticular,wefindthattheadditionofanaverage-sizednewforeignmanufacturingplantŽ190employ-eesinoursample.isassociatedwithmorethana2.3%increaseinrealwagesforallworkers,thoseofbothforeignanddomesticplants,inthatindustryinthecounty.Incontrast,theestimatedwageincreaseassociatedwithanequal-sizednewdomesticplantisjust0.3%.Thisresultisstatisti-callysignificantandconsistentwithpreviousstudiesthatfoundwagedifferencesacrossallU.S.plantswhennotcontrollingforlocallabormarketconditions.Italsoexplainswhylocalcommunitiesmaybemoreinterestedinattractingforeignplants.

Unlikepreviousanalysis,weprovideformalevidenceofthecostsassoci-atedwithattractingaforeignplantforalocalcommunity.Ourestimatesfindthatanaverage-sizednewforeignplantisassociatedwitha1.2%reductioninrealpercapitarevenuesatthecountylevelinSouthCarolinaanda1.8%reductioninrealpercapitaexpenditures,whiletherelevantcomparisonfiguresfordomesticplantsare0.1and0.2%reductions,respectively.Wealsofindthatnotonlylevels,butalsothecompositionofcountybudgetschange.Specifically,foreignplantpresenceisassociatedwithlowerperpupilexpendituresbycountygovernmentsŽthemainsourceofschooldistrictfinancinginSouthCarolina.,buthigherexpendituresontransportationandpublicsafety.Noneofthesecompositionaleffectsonbudgetsoccurswithdomesticmanufacturingemployment.

WhileourdataandanalysisfocusspecificallyonSouthCarolina,thereisreasontobelievetheseissuesaffectmanyotherstatesaswell.First,whilethepercentageofforeignmanufacturingwassubstantialforSouthCarolinaduringourdatabase’stimeperiod,thispercentageisnotdifferentfromtheU.S.nationalaverage.In1980,6.0%ofSouthCarolina’smanu-

LOCALEFFECTSOFFOREIGNDIRECTINVESTMENT

343

facturingwasforeign,whilethenationalaveragewas6.4%.By1994r1995,theSouthCarolinaandnationalaverageswere12.2and13.9%,respec-tively.Second,SouthCarolinaisnotuniqueinhavingmanybusinessincentivesofferedbythelocallevelŽcitiesandcounties.,asopposedtothestatelevel.FisherandPetersw12xcontendthatthemostimportantU.S.economicdevelopmentspending,suchasabatementsandspecializedin-frastructure,arefinancedoutoflocalpropertytaxes,whileO’Connorw23xgivesadetailedlistofthemanyandvariedincentivesofferedbycitiesandcountiesacrosstheUnitedStates.6

2.TRENDSINFDI,MANUFACTURINGWAGES,AND

COUNTYBUDGETSINSOUTHCAROLINAAsjustnoted,SouthCarolinahashadsignificantlevelsofemploymentbyforeignplantsasapercentageoftotalstateemployment,aswellasahighlevelofgrowthinthispercentageduringthelasttwodecades.ThissectionprovidesmoredetailonthisforeignplantpresenceinSouthCarolina.Throughoutthissectionandmostofthepaper,wedefineforeignplantsasonlythosethatwereestablishedbytheforeignparentasnewŽorgreenfield.investments.Inlateryearsofoursample,thereweresignificantnumbersofforeignacquisitionsofexistingdomesticplantsinSouthCarolina.Inourstatisticalanalysisdescribedbelow,wefindthattheseacquiredplantsaremuchmoresimilartodomesticoperationsthanforeignonesintheirimpactonlocalcommunitiesandthus,weclassifythemasdomesticplants.However,ifoneincludestheseforeign-acquiredplants,employeesinforeignplantsgrewtoover18%oftotalSouthCarolinamanufacturingemploymentby1995.Thus,regardlessofhowonedefinesFDI,employmentduetoforeignaffiliatesinSouthCarolinaissubstantial,bothintermsoflevelsandgrowth.

ThesetrendshavenotbeenuniformacrossSouthCarolina’s46countiesbyanymeans.Table1givesabreakdownoflevelsindomestic-plantmanufacturingemploymentandforeign-plantmanufacturingemploymentfrom1980through1995inSouthCarolina’sMetropolitanStatisticalAreasŽMSAs..OurdataontotalmanufacturingemploymentcomesfromtheCountyBusinessPatternsdatabase,whileforeign-plantemploymentnum-bersbycountycomefromappropriateannualissuesoftheSouthCarolina

Forexample,O’Connorw23xfindsthat83%ofsurveyedcitiesandcountiesofferrevenuebondfinancing,40%offeredpubliclyownedindustrialparkcites,and45%offeredloansforbuildingandconstructioncosts.

6

344

TABLE1SouthCarolinaDomestic-andForeign-PlantManufacturingEmployment,byMetropolitanStatisticalAreaandYear1985Foreign2,6313,7271,07816,40713.3513.508.8812.8217,83322,02511,028107,9593,7993,9391,69821,86917.5615.1713.3416.84%ForeignDomesticForeign%Foreign1990Domestic17,82223,76211,075104,7551995Foreign4,0854,1531,88324,367%Foreign18.6514.8814.5318.871980RegionDomesticForeign%ForeignDomesticFIGLIOANDBLONIGEN

Charleston1Columbia2FlorenceGreenville-Spartanburg-Anderson3OthercountiesStatetotal5,66229,5053.208.09176,488335,3338,43239,7374.5610.5917,81025,67612,406128,3041,4542,93920014,2277.5510.271.599.9817,07823,87511,055111,523193,737377,9335,37724,1972.706.02171,459334,990175,470332,88411,58646,0746.1912.1612Berkeley,Charleston,andDorchestercounties.LexingtonandRichlandcounties.3Anderson,Cherokee,Greenville,Pickens,andSpartanburgcounties.LOCALEFFECTSOFFOREIGNDIRECTINVESTMENT

345

IndustrialDirectory.7Intermsoflevels,theGreenville᎐Spartanburg᎐An-dersonMSAhashadaproportionallyhighshareofSouthCarolina’stotalandforeign-plantmanufacturingemploymentacrossallyears.However,intermsofgrowthinthepercentageofforeign-plantmanufacturingemploy-ment,theFlorenceandCharlestonMSAshaveseenthegreatestincrease.ItisclearfromTable1thatMSAsinSouthCarolinagenerallyhavealargerpercentageofmanufacturingjobsinforeignplantsthantheaverage‘‘othercounty’’inSouthCarolina.8Notableothercountieswithahighpercentageofforeign-plantmanufacturingjobsareChesterfieldandGeorgetowncountieswith22.3and26.9%ofmanufacturingjobsbyforeignplantsin1995,respectively.Figure1givesacomprehensivelookat

FIG.1.ForeigndirectinvestmentinSouthCarolina,1995.

TheSouthCarolinaIndustrialDirectory,anannualpublication,detailsinformationonmanufacturingplantsinSouthCarolina,includinglocation,totalemployees,StandardIndus-trialClassificationŽSIC.codes,yearofestablishment,andparentcompany.Thelatterinformationallowedustoestablishwhichplantsweresubsidiariesofforeigncompanies.Thesedatawerelistedconsistentlyintheannualpublicationbackto1980.Onefeatureofthedatawaschangesinplantownershipfromforeigntodomestic,ordomestictoforeign.Thesechangeswereoftenobservedintheindustrialdirectoriesbyobservingchangesinthelistedparentcompany,butwherequestionsarosewecalledplantsdirectlytoverifyinformation.Inabout8%offoreignplant-yearobservations,wehadmissingdataonemployeenumbers.Inallcaseswehadsomeinformationtohelpestimatethemissingdata,suchaspreviousorsubsequentperiodemployeelevels,butthisofcourse,leadstosomemeasurementerror.8

DistributionofFDIbysourcecountryisquitevariedacrossSouthCarolinaaswell.ThetopsourcecountriesintermsofemployeenumbersinSouthCarolinaareGermanyŽ27%oftotalforeign-plantemploymentinSouthCarolina.,FranceŽ19%.,JapanŽ17%.,andtheUnitedKingdomŽ10%..

7

346

FIGLIOANDBLONIGEN

foreignpresenceinallcountiesinSouthCarolina,presentingamapwithcountiesshadedaccordingtotheforeignplantpercentageoftotalmanu-facturingemployment.Inadditiontovariationacrosscounties,thereisalsosignificantvariationinthepercentageofforeign-plantemploymentthatexistsacrossthe2-digitindustriesineachcounty.AlltheSouthCarolinaMSAsseethispercentagerangefromnoforeign-plantemploy-mentinsomeindustriestoessentiallyalltheindustryemploymentduetoforeignplantsinothers.Ourwageregressionsbelowexploitthisvariationacrossindustries,aswellascounties.

Table1alsoshowsthatgrowthinforeign-plantmanufacturingemploy-mentduringthisperiodwasnotmatchedbysimilargrowthindomestic-plantmanufacturinginSouthCarolina.Infact,exceptfortheCharlestonMSA,therewassignificantdeclineindomestic-plantmanufacturingnum-bersfrom1980to1985,withfairlyconstantnumbersafter1985.ThiswilleaseouridentificationofthepotentiallydifferenteffectsofforeignvsdomesticinvestmentonlocalcountiesinSouthCarolina.

AsafirstlookattherelationshipbetweenFDIandlocalcommunityeffects,weexaminechangesinrealwagesandbudgetsacrossSouthCarolinacountiesfrom1980to1994Žor1995,dependingonthevariable.,andthenbreakthesechangesdownbyhowmuchgrowthinFDIacountyreceivedoverthesameperiod.Colunms1᎐3ofTable2indicatehowaveragerealwagesandpercapitabudgetschangedinSouthCarolinafrom1980through1994r5.Realwagesgrew16.9%overthistimeperiodfromanannualrealwageof$15,600toover$18,000,bothexpressedin1982dollars,whilepercapitarealcountyrevenuesandexpendituresbothgrewslightlyduringthisperiod.Columns4and5examinewhetherthesechangesvarywiththedegreeofFDIcountiesreceivedduringthissameperiod.Infact,countiesthatexperiencedrelativelyhighlevelsofFDIgrowthalsoexperiencedgreatergrowthinrealwagesŽ25.9%comparedto12.4%.,whilethesesamecountieshadlessgrowthinrealpercapitarevenuesanddeclinesinrealpercapitaexpenditures.Theseresultsareonlysuggestivebecausewehavenotcontrolledforotherfactorsandadditionally,thedifferencesbetweencolumns4and5inallinstancesarenotstatisticallysignificantatconventionallevels.However,Table2doesindicatethepossibilitythatFDImayhaveapositiveimpactonwagesattheexpenseofcountybudgets,andtherefore,wenextturntomoreformalestimationoftheserelationships.

3.TESTINGFORDIFFERENTIALEFFECTSOFFOREIGN

ANDDOMESTICINVESTMENTONWAGESGivenfixedlaborsupply,newplantinvestmentintoaregionwillincreaserealwages,atleastforaparticularindustryandrorlabor-skilllevel,becauseitincreaseslabordemand.Ofcourse,laborsupplyislikely

TABLE2Over-TimeChangesinWagesandCountyBudgetsinSouthCarolina:BrokenDownbyChangesinForeignDirectInvestmentSharesVariable18.241.020.992.16.316.912.411.68.5Meanrealvaluein1980Žthousandsof1982dollars.Percentagechangefrom1980to1994r5Meanrealvaluein1994r5Žthousandsof1982dollars.Percentagechangefrom1980to1994r5Žcountieswithbelow-medianchangeinFDI.Percentagechangefrom1980to1994r5Žcountieswithabove-medianchangeinFDI.25.92.1y3.015.600.96LOCALEFFECTSOFFOREIGNDIRECTINVESTMENT

RealwageinindustryincountyRealpercapitarevenuesRealpercapitaexpenditures0.97347

348

FIGLIOANDBLONIGEN

notfixed,andmigrationofworkerstotheregionwillmitigaterealwageincreases,atleasttosomeextent.Ifplantsareidenticalinproductivity,theeffectofnewplantinvestmentonrealwageswillbeidenticalregardlessofwhichplantisresponsiblefortheincreaseinlabordemand.However,plantsarenotidentical,andpreviousstudieshavefoundevidenceofsignificantdifferencesbetweenforeignanddomesticplantsŽinparticular,domesticplantsthatarenotmultinational..FortheUnitedStates,DomsandJensenw8xfindthatforeignaffiliatesaremoreproductiveandpayhigherwagesthantheaverageU.S.-ownedplant.SimpleaveragesfromtheirdatashowthatproductionworkersinU.S.domesticplantsaveraged$18,760in1987,whilethoseinforeignplantsaveraged$22,290;i.e.,approximately19%higher.Wagedifferencespersistevenaftercontrollingforindustry,plantsize,plantage,andlocation,thoughtheyarereducedtoaboutaquartertheeffectwithoutcontrols.DomsandJensenw8xandHowensteinandZeilew19xofferavarietyofpotentialreasonswhythismaybetrue,includingdifferentlaborskillmixesforforeignaffiliatesandrorwagepremiumstodeterunionization.Whatevertheunderlyingfactorsforthedifferences,theyseemtobespecifictomultinationalityoftheplant,sinceDomsandJensenw8xfindnowagedifferencesbetweenforeignaffiliateplantsandplantsofU.S.-basedmultinationals.

Ouranalysisexaminescounty-andindustry-levelrealwagedatatodeterminewhethertherearesignificantdifferencesbetweentheeffectsofforeignvsdomesticinvestmentonwages,asindicatedbytheU.S.plant-levelstudies.BecausewecannotidentifywhetherU.S.-ownedplantsaremultinationalornot,ourrelevantcomparisonisbetweenforeignanddomesticinvestment.Thus,giventhepreviousplant-levelstudies,thehypothesiswetestiswhetherinvestmentbyforeignŽmultinational.plantsincreasesrealwagesrelativetoallU.S.domesticplants.

Itisnotclearthattheresultsofplant-levelstudieswilltranslateintosimilareffectsinourdatabaseoflocalrealwages,becauseanalysisoflocalwagesfactorinvaryingpercentagesofforeign-plantjobsinthelocalcommunityandindirecteffectsofFDIonlocallaborsupplyanddomesticdemandconditions.Inaddition,thereissuggestiveevidencethatstateslikeSouthCarolinaareattractivetoforeignplantsbecauseoflowlaborcosts.AWallStreetJournalarticleoninvestmentbyGermanplantsintheCarolinasreportsthatlowlaborcostsinthesestatesmorethancompen-satesfortheincreasedtransportationcoststheseplantsfaceŽWallStreetJournal,May4,1993,p.A1;1..Finally,studiesbyGlickmanandWood-wardw14xandCoughlin,Terza,andArromodeew6xfindthathighwagesinastatediscouragedinvestmentbyforeignplants.Onemightexpectthewagepremiumgivenbyaforeignplanttobemitigatedoreliminatedwhenthatplantmaybeattractedtotheareapreciselybecauseofitsrelatively

LOCALEFFECTSOFFOREIGNDIRECTINVESTMENT

349

lowwages.Thismakesthedifferentialeffectofforeigninvestmentvsdomesticinvestmentonrealwagesanimportantempiricalquestion.3.1.EmpiricalAnalysisofRealWageEffects

Toexaminetherelationshipbetweenforeignanddomesticinvestmentinacountyandthewagelevelsinthecounty,weusedatafromCountyBusinessPatternsonindustry-specificŽ2-digitSIC.county-specificrealwagesin1980,1985,1990,and1994Žthelastyearforwhichwehavedata.toestimatethemodel

wikts␣eiktq␤fiktq␥ktq␦i

forcountyiintimetforeachtwo-digitmanufacturingindustryk.Here,wreflectstheaverageannualwage,deflatedbytheconsumerpriceindextobeexpressedin1984dollars,inindustrykincountyiduringtimet.9ThevariableeiktistotalmanufacturingemploymentforcountyiinSICindustrykintimet,whilefiktisthelevelofemployeesinforeigngreenfieldestablishmentsforthespecificcounty,industry,andtime.Thesevariablescomefromthesourcesnotedinthesectionabove.Thecoeffi-cientsonthevariableseandfareourkeyparametersofinterest.Theparameter␣representsthemarginalrelationshipbetweenwagesandmanufacturingemploymentintheindustryinthecounty.Thisissimilartowhathasbeenestimatedbypreviousstudiesusingcomparablespecifica-tionsŽseeBartikw2x..Unlikepreviousstudies,weestimatethedifferentialmarginaleffectofforeignmanufacturingemploymentintheindustryinthecounty,representedbyparameter␤.Tocaptureunobservedcounty-specificdifferencesinwages,wecontrolforcounty-specificfixedeffects␦,whiletocontrolfortime-varyingindustry-specificcommoneffectsweincludeindustry-time-specificfixedeffects␥.

Whilethisspecificationisextremelyparsimonious,wecontendthatitcapturessufficientlythedifferentialrelationshipbetweenforeignanddomesticinvestmentandthewagesinacommunity.10Inparticular,wenotethatwhileitiseasytothinkoftime-varying,countyandindustry-Ashortcomingofthesedataisthatwecannotdistinguishfull-timefrompart-timeworkers,andwemustpooltogetheralloccupationswithinanindustry.Theidealdatasetwouldhaveindividual-levelobservationsonspecificoccupationandhoursworkedtomorefullycontrolforthesepotentialdifferences.However,weknowofnodatasetsthatwouldhavesufficientindividual-levelobservationsinanygivengeographicareatoaddressthisissue.Wenote,however,thatourapproachiscomparabletothatusedbypreviousstudies.10

Wetriedalternativeregressorspecificationstoestimatethedifferentialimpactofforeign-plantemploymentonwagesandbudgets.Theseincludedsubstitutingtheregressors,fande,withfrebyitselfandwithfreande.Thesegenerallygavequalitativelysimilarresultswithlessprecision,butalsoraisedextraissues,suchascollinearityproblemsinthecaseofthelatter.

9

350

FIGLIOANDBLONIGEN

specificfactorsthatarecorrelatedwithchangesinbothforeignanddomesticinvestmentinacounty,itisdifficulttoconceiveofavariablethatshouldbeassociatedwiththeshareoftotalemploymentheldbyforeignplantsandalsowithwagespaidinthecommunity.Therefore,whiletheestimatedparameters␣maynotreflectthetruerelationshipbetweenemploymentandwagesasawhole,theparameters␤shouldbeaccuratereflectorsofthedifferentialeffectofnewforeignvsdomestic-plantem-ployment.

WereporttheresultsofthisestimationinthefirstrowofTable3.First,weobservethattherelationshipbetweendomesticmanufacturingemploy-mentandwagesinacountyissignificantlypositive.Wefind,forinstance,thateachadditionaldomestic-plantmanufacturingworkerinanindustryinacountyisassociatedwithabouta25-centincreaseŽin1984dollars.inannualwagesforallworkersinthatindustry.Contrastthisfindingwiththeestimatedrelationshipbetweenforeign-plantemploymentandcountywages:eachadditionalforeign-plantmanufacturingworkerinanindustryinacountyisassociatedwithabouta$1.75increaseinannualwagesforallworkersinthatindustry.Hence,wefindthatthemarginalnewforeign-plantmanufacturingjobhasaboutseventimestheeffectonwagesasdoesthemarginalnewdomestic-plantmanufacturingjob.Thisdifferenceisstatisticallysignificantatanyreasonablethreshold.Allstandarderrorsareadjustedtocorrectforheteroskedasticityandwithin-county-timecorrela-tionsoferrors.Specifically,weusethestandarderrorcorrectionsuggestedbyMoultonw22x,sothatweconstructourstandarderrorsasifwehave184effectiveobservationsŽ46counties=4yearsofdata.whilestillmak-inguseofthevariationthatexistsacrossindustriesinacounty.Weem-ploythecovariancematrixCs␴2ŽXXX.y1wIq␳ŽNyI.x,whereNsXXZZXXŽXXX.y1,Xisthematrixofexplanatoryvariables,Zisamatrixofcounty-timeinteractions,and␳istheintra-county-timecorrelationofthedisturbances.11

Howlargearetheseeffects?Atfirstglance,thoughstronglystatisticallysignificant,thesenumbersappearquitesmall.Butconsidertheestimatedeffectsonwagesofaddingasingleaverage-sizedplant.Inourdata,theaverage-sizednewforeignmanufacturingplanthasabout190employees;ourresultswouldsuggest,therefore,thataddingasingleforeignplanttoacountyisassociatedwithmorethana2.3%increaseinrealwagesforall

Asanalternative,wealsoestimateourmodelwithonlyoneobservationpercountypertimeperiod,inwhichweaggregateourindustriesuptothecounty-timelevel.Inthiscase,theestimateddifferencebetweenanotherforeign-plantjobandanotherdomestic-plantjobis3.0ratherthan1.5,andissignificantattheps0.047level.Therefore,ourresultsareclearlynotbeingdrivensubstantivelybyourchoiceofdisaggregatingthedatabyindustry;infact,ifanything,disaggregatingthedataaswedoappearstoreducetheestimateddifferentialeffectofforeigninvestment.

11

TABLE3DifferentialEffectsofDomestic-andForeign-PlantManufacturingEmploymentonIndustry-Specific,County-SpecificRealWagesandRealCounty-LevelperCapitaBudgetsandBudgetItemsDependentvariableEffectofanadditionaldomesticmanufacturingjob0.649Žps0.078.y0.056Žps0.196.y0.035Žps0.203.y0.046Žps0.016.y0.677Žps0.053.nra4nra4EffectofanadditionalforeignmanufacturingjobDifferencebetweenforeignanddomesticmanufacturingjobsDifferenceinthecaseinwhichdependentvariablemeasuredinchanges1RealannualwageŽ1984dollars.21.751Žps0.001.y0.061Žps0.103.y0.090Žps0.061.y0.032Žps0.045.y0.585Žps0.018.0.035Žps0.052.0.058Žps0.190.1.502Žps0.001.y0.056Žps0.099.y0.079Žps0.054.y0.045Žps0.004.y0.716Žps0.002.0.029Žps0.089.0.058Žps0.164.RealpercapitarevenuesŽ1984dollars.3RealpercapitaexpendituresŽ1984dollars.3RealperpupilschoolexpendituresŽ1984dollars.3FractionofK᎐12studentsinpublicschoolsŽ=100.3FractionoflocalexpendituresgoingtotransportationŽ=100.3FractionoflocalexpendituresgoingtopublicsafetyŽ=100.30.249Žps0.013.y0.005Žps0.481.y0.011Žps0.233.0.013Žps0.110.0.131Žps0.308.0.006Žps0.174.0.000Žps0.959.1LOCALEFFECTSOFFOREIGNDIRECTINVESTMENT

Modelscontrolforcounty-specifictimetrendsaswellasdependent-variable-specificfixedeffects,notedbelow.Modelcontrolsforcounty-specificfixedeffectsandindustry-year-specificfixedeffects.Standarderrorsareheteroskedasticity-robustandcorrectforwithin-county-timecorrelationintheerrors.pvaluesareinparentheses.3Modelcontrolsforcounty-specificfixedeffectsandyeareffects.Standarderrorsareheteroskedasticity-robustŽthereisnowithin-county-timevariation..pvaluesareinparentheses.4Cannotestimatemodelduetolackofobservations.2351

352

FIGLIOANDBLONIGEN

workers,inforeignanddomesticplants,inthatindustryinthecounty.Theestimatedwageincreaseassociatedwithanequal-sizednewdomesticplantisjust0.3%.12,13Asasensitivitycheck,wealsoestimateourmodelindifferences,inwhichwecannowcontrolforcounty-specifictrends,ratherthanjustlevelfixedeffects.Whiletheestimateddifferencebetweentheestimatedeffectsofforeign-anddomestic-plantmanufacturingjobsisconsiderablysmallerthanbefore,itremainsstatisticallysignificantatconventionallevels.SincesomeofthedifferencebetweenourlevelsanddifferencesresultsareduetoattenuationbiasresultingfrommeasurementerrorŽwhichisexacerbatedusingfirst-differences.,oursuspicionisthatthe‘‘true’’differentialeffectsofforeignplantsliessomewherebetweenthetwosetsofresults.

Ofcourse,itisalwayspossiblethatourresultscouldsufferfromendogeneitybias.Ifforsomereasonforeignplantsareattractedtohigh-wageareas,thenwemightoverstatethedifferencebetweenforeignanddomesticplants.WhilethisexplanationseemsunlikelyŽasdiscussedabove,theevidencesuggeststhat,ifanything,foreignplantsareattractedtolowwageareas.itisstillplausible.Wecouldnotfindaninstrumentthatexplainsasignificantportionofthevarianceinwithin-county,within-in-dustrychangesinforeigninvestmentsharesovertime,whilealsopassingHausman-typeinstrumentexogeneitytests.14However,weproposean

OnefocusofAitken,Harrison,andLipseyw1xwasmeasuringspillovereffectsinwagesfromforeignplantstodomesticones.Becauseourindustrywagedataisatthecountylevel,ratherthanthecountrylevelasinAitken,HarrisonandLipseyw1x,wedonothaveseparatewagedataforforeignanddomesticplantstoidentifydirectvsspillovereffectsfromforeign-plantemployment.However,theremaybesomeevidenceforspilloversinourestimates.Ifweassumethatdomesticplantsdonotrespondtohigherforeign-plantemploy-mentinthesectorŽi.e.,nospillovers.,thena14%paydifferentialbetweenforeignanddomesticplantsisnecessaryinourdatatoexplainthelargerincreaseinindustryrealwagesfromanadditionalforeignplantvsanadditionaldomesticone.HowensteinandZeilew19xfinda16%paydifferentialŽ$38,300inforeignplantscomparedto$33,000.usingBEAplant-leveldataforallU.S.plantsin1990.However,theyfindthatonly30%ofthispaydifferentialisduetowithin-industrydifferencesratherthanindustry-mixeffects.Thissuggestsawithin-industrydifferentialofonly5%,whichissignificantlylowerthanthe14%differentialnecessarytoprecludespilloversinourestimates.13

Asasensitivitycheck,wealsoestimateourmodelwiththedependentvariableexpressedinlogs,ratherthanlevels.Inthisspecification,asbefore,boththecoefficientsonmanufac-turingemploymentandforeign-plantemploymentarestronglystatisticallysignificant.Thecoefficientsimplythatanewdomesticplantwith190employeeswouldincreasewagesby0.3%whileanewforeignplantofthesamesizewouldincreasewagesby1.9%,roughlythesamemagnitudesaswhenthedependentvariableisexpressedinlevels.14

Examplesofinstrumentcandidatesthatwetriedincludemeasuringthehistoricalstockofforeignmanufacturinginthecounty,andmeasuresreflectingthequalityoftransportationinfrastructureintheregion.

12

LOCALEFFECTSOFFOREIGNDIRECTINVESTMENT

353

alternativewayofgaugingthedegreetowhichthispossiblesimultaneitymaybedrivingourresults.Asnotedearlier,tothispointwehavetreatedforeign-acquiredplantsasiftheyweredomestic.Onerationaleforthisisthatacquireddomesticplantshaveexistingcapitalandworkers,whichmakesitmuchlesslikelythatwewouldseethesenowforeign-ownedplants‘‘looking’’substantiallydifferentfromotherdomesticplants.Infact,mostofthegivenreasonswhyforeignplantswouldpayhigherwagesconcernadifferentplant-leveltechnologyŽe.g.,foreignplantsusebettertechnologyandworkersaremoreproductive,orforeignplantsuserela-tivelymoreskilledlabor.ordifferenthiringpracticesorstrategiesŽe.g.,foreignplantspayhigherwagestoattractbetterworkersinalabormarketwithwhichtheyareunfamiliar..Thesearefeaturesthataredifficulttochangewithexistingcapitalandlabor.However,ifthereissomethingspecialandunobservableaboutacountythatwouldattractadispropor-tionateamountofforeigninvestmentŽratherthandomesticinvestment.andthatisdrivingourresults,wewouldexpectthatforeign,butacquired,plantswouldhavethesameeffectastheonewefindregardingnewforeigninvestment.Toexplorewhetherthisisthecase,weestimateourmodelwiththreecategoriesofplants:domesticplants,foreignacquisi-tions,andnewforeigninvestment.WefindthatthereisnodiscernibledifferenceŽeitherinmagnitudeorstatisticalsignificance.betweendomes-ticplantsandforeign-acquiredonesintherelationshipbetweenemploy-mentandwages,butbotharesubstantiallyŽandsignificantly.differentfromnewforeigninvestment.Thissuggeststhatourresultsarenotlikelydrivenbyendogeneityofforeigninvestmentandwages.

Ouridentificationstrategytakesalinearform,inthatweaskwhatistheeffectofanadditionalforeign-plantjob,conditionalonthesizeoftheindustryinagivencounty.However,itisreasonabletosuspectthatthiseffectmayvarysystematicallydependingonthesizeoftheindustryinthecounty.Forinstance,itispossiblethatoneadditionaljobmayhavealargerimpactwhenthemarketissmaller,althoughitisequallypossiblethattheimpactofanadditionalforeign-plantjobincreaseswithindustrysize.Togaugethedegreetowhichtheseeffectsarenonlinear,weestimateseveralalternativemodelsinwhichweinteractforeignandtotalemploy-mentwithsometime-varyingcommunitycharacteristicsŽthenumberofemployeesintheindustryinthecounty;theindustry’semploymentshare,asafractionoftotalmanufacturingemploymentinthecounty;orcountypopulation.aswellasincludingthecharacteristicitselfasacontrolvariable.WereporttheresultsoftheseexercisesinTable4.Weobservethatthedifferencebetweentheeffectofaforeign-plantjobandadomestic-plantjobdecreasessomewhatŽthoughnotstatisticallysignifi-cantlyatconventionallevels.withthesizeoftheindustry’semploymentinthecounty,aswellaswiththeshareofindustryemploymentasafraction

354

FIGLIOANDBLONIGEN

nsiw.eo3cr4n13en9.re10.ee0fsfw.itpsrDebŽalloDc,4if8ic9nciseetŽ1pesbeoSew..-gtmj56yaeog91rbdnitrsW8091edu5.4.ulct4060dann..acnu1eas1sInrnfppnneguŽŽfinoAfieatlnaDromfeemRysoIlpelmbEanbirgojganie..nargniV37foi00rtofr0un101lu7..ttatcec5060ad.encfn2foas8.2suefifEtppiu4npdnŽŽaedaEMDamLt:sBneAalcTPn-antsgciibemtrusojocrne..Fiamgn0000dCfoi300ntodru1.2.anetltcac101.-rceena1fs20.1sifoftfppsfiŽŽeEituinmDdadoramDedfnousstecgea..ftefEWnetlnlileitetlaaemnmneiyeytnRoccrorlecpeleriefimpppfcehmhtetfieDpns5s5’’SoyŽ7yŽ2-iryttertlltasnuugsaruutdamoinldsCSisniseiihyeythttntnfoufuoooececziezeshishltltatnatnoioiTT.65042.40.0spŽ..71102.6009.s40.1sppŽŽ..10100.7005.0s0.2sppŽŽ..90100.1006.s50.0sppŽŽggnnir.,i.truelnueltictetitanfemcanucyfenroucrlaepnaempmhmpytehyttn5st5’uŽ7ynŽ2oruetlcgsolucalraadlmtlnatsosiostietifot,nhf,ttoneeferoerammhyeaysozhlisoapslplasmasmaetoaeT..4se99s21eh.7.0t0ysnepraŽpnieraseula..v8402p0015.3.0.s3060r..o1s0srrppeŽŽehtninoitalerro..10c03e6022.71.mi6010t..-1s1sytppnŽŽuoc-nihtiwroftc..21e00rr5060o9.1.c2050.d0s.0snppaŽŽtsubor-yticitsadeksoreelltgaerhamlsessriiannsooriiottraarllu.u.epelpeldoitoitrpnpnadyeyetctcnnrnratueueSopopchh.tctelta5la5otoŽ7toŽ2NTT1.3901.0760.314Totalsizeoftheindustry’semployment,LOCALEFFECTSOFFOREIGNDIRECTINVESTMENT

355

oftotalcountymanufacturingemployment.Inaddition,wefindthatdifferencebetweentheeffectofaforeign-plantjobandadomestic-plantjobincreasessomewhatŽthoughalsonotstatisticallysignificantlyatcon-ventionallevels.withthepopulationofthecounty.ThelackofstatisticalsignificanceorsubstantialmagnitudesofthesedifferencessuggeststhattheresultsthatwepresentinTable3areunlikelytobehighlynonlinear,atleastintermsofcommunityorindustrysize.

4.TESTINGFORDIFFERENTIALEFFECTSON

COUNTRYBUDGETS

Whiletherehasbeensomeresearchontheeffectofinvestmentonrealwages,therehasbeenlittleformalanalysisofinvestmentonlocalgovern-mentbudgets,muchlesswhetherforeigninvestmenthasanimpactdiffer-entfromthatofdomesticinvestmentontheselocalbudgets.Newplantinvestmentobviouslybringsinadditionaltaxrevenue,butalsobringsincreaseddemandforothergovernmentservices,suchasinfrastructureandpubliceducation,fromanincreaseinpopulation.Aslocalgovern-mentsoffertaxreliefasincentivesfornewplantinvestment,thepotentialcosttothelocalcommunityislowerlevelsofgovernmentservicepercapita.Asmentionedintheintroduction,statesandlocalcommunitiesarguablyspendmoretoattractforeigninvestment,andthereisanecdotalevidencethatcommunitiesareespeciallyinterestedinattractingforeigninvestment.15Providedeveryoneknowsthesepreferencestosomeextent,onemayexpectlocalcommunitiestooffergreatertaxrelieftoforeignplants,whichleadstolowerpercapitagovernmentserviceswithforeigninvestment.Belowweproxypercapitalocalgovernmentserviceswithmeasuresofpercapitarevenuesandpercapitaexpendituresandtestthehypothesisthatforeigninvestmentleadstodecreasesinthesepercapitameasuresthatarestatisticallydifferentfromchangesindomesticinvest-ment.

Additionally,foreignplantsmaybeinterestedinincentivepackagesfromlocalcommunitiesthataredifferentfromdomesticones.Thepossi-bleincentivesofferedbystatesandlocalcommunitiesarenumerousandoftenindividuallytailoredtoaparticularnewplant’sneeds.Obviously,differentincentivescanaffectnotonlythelevel,butalsothecompositionoflocalgovernmentbudgets.Ourdataallowustofocusonlocalgovern-mentspendingforpubliceducation,transportation,andpublicsafety,andweexaminebelowwhetherforeignplantinvestmentaffectsthesecompo-nentsofthelocalbudgetdifferentlythandomesticinvestment.

Forexample,Alabamastateliteratureonbusinessincentivesindicatesthestateisespeciallyinterestedinattractingforeignplantinvestment.

15

356

FIGLIOANDBLONIGEN

4.1.EmpiricalAnalysisofLocalBudgetLe¨elEffects

TotestthesehypothesesconcerningFDIandlocalbudgets,weestimatethedifferentialrelationsbetweendomesticandforeigninvestmentandtwomeasuresoflocalgovernmentbudgets:realpercapitarevenuesandrealpercapitaexpenditures.16ŽAsbefore,weexpressthesevariablesintermsof1982dollars..OurbudgetdatacomefromtheSouthCarolinaDepart-mentofRevenueŽfortheyears1990and1995.andtheCityandCountyDataBooksŽfortheyears1980and1985..Weestimatevariantsoftheequation:

bits␣eitq␤fitq␥tq␦i,

wherebrepresentstherealpercapitabudgetsofalllocalgovernmentsincountyiduringtimet,eandfaretotalmanufacturingemploymentandforeignmanufacturingemploymenttakenfromthesamesourcesasbefore,and␥tand␦iareyearandcountyfixedeffects,respectively.Asbefore,the␤coefficientwillestimatethedifferentialimpactofforeign-plantemploymentrelativetodomestic-plantemployment.

WereporttheresultsofthisexerciseinthesecondandthirdrowsofTable3.Weobservethatinthecasesofbothrevenuesandexpenditures,newforeign-plantemploymentapparentlyleadstosignificantlylowerlev-elsofpercapitabudgetsthandoesacomparableamountofnewdomestic-plantemployment.Specifically,anewforeignplantisassociatedwith12timestherevenuereductionand8timestheexpenditurereductionofanewdomesticplantofthesamemagnitude.Forinstance,whileanaverage-sizednewforeignplantisassociatedwitha1.2%reductioninrealpercapitarevenuesanda1.8%reductioninrealpercapitaexpenditures,therelevantcomparisonfiguresfornewdomesticplantsare0.1and0.2%,respectively.17

Weobservethatforeigninvestmentapparentlyleadstolowerbudgetlevelsthanbefore.Doesitseemtosystematicallychangethecompositionoflocalbudgetsaswell?Toexplorethispossibility,weinvestigatethedifferentialrelationshipsbetweenforeign-anddomestic-plantemploymentandseveralimportantlocalspendingcategoriesforwhichwehavedata.

HerewemeasurebudgetsasthesumofalllocalgovernmentŽcountyormunicipal.revenueorexpendituresinacounty.17

TheoreticalworkbyJanebaw20xsuggeststhateventhoughstatebudgetsmaybeadverselyaffectedintheshort-runfromtaxincentives,statesmaybeabletoextractgreaterrevenueoncetheplantisrelocated,particularlywhenplantsareimmobile.Unfortunately,thenecessarydatatotestJaneba’spropositiondonotcurrentlyexist.

16

LOCALEFFECTSOFFOREIGNDIRECTINVESTMENT

357

4.2.EmpiricalAnalysisofEducationSpendingEffects

Two-thirdsofalllocalgovernmentexpendituresinSouthCarolinasupportpubliceducation.Aswithlocalbudgets,itisimpossibletosignexantetheexpectedrelationshipbetweenforeignplantsandsupportforpubliceducation.Ontheonehand,theanecdotalevidencesuggeststhatforeignmanufacturersvalueahighlyskilledandwell-educatedworkforceandsomightbeexpectedtopushforhighereducationalspendingᎏfor-eignplantsoftenrequesteducationandtrainingexpendituresfromlocalcommunitiesaspartoflocationincentivepackages.Ontheotherhand,ifemployeesofforeignplantstendtodisproportionatelyenrolltheirchildreninprivateschools,onemightexpectsupportforpubliceducationtoatrophyincommunitieswithincreasinglyimportantforeign-plantemploy-mentshares.18SinceschooldistrictsinSouthCarolinaaredependentoncountygovernments,countygovernmentshaveconsiderablymorelatitudeinshiftingresourcestoorawayfromschoolsinSouthCarolinathaninstatesŽprincipallyinthenorthandwest.whereschooldistrictsareinde-pendent.

Toinvestigatetherelationshipbetweenforeigninvestmentandsupportforpubliceducation,weestimateanequationidenticaltothoseusedforthebudgetspecificationsabove,exceptthatnowthedependentvariableisrealperpupilexpendituresonK᎐12education.TheresultsofthisanalysisarereportedinthefourthrowofTable3.Theresultssuggestthatwhilecommunitieswithincreasinglevelsofdomestic-plantemploymenttendtoincreasetheirsupportforpubliceducationŽthoughthisrelationshipisinsignificant.,thosewithincreasingsharesofforeign-plantemploymentapparentlytendtodecreasetheirsupportforpublicschools.Whiletheeffectsofanyoneplantarequitemodestᎏforinstance,anadditionalnewforeignplantisassociatedwithlessthanhalfapercentdecreaseinperpupilschoolexpendituresᎏaggregatinguptoanumberofnewforeignplantsinacountycouldleadtomoresubstantialchanges.Forinstance,a1-SDincreaseinforeign-plantemploymentinacountyisassociatedwithalmosta2%reductioninrealperpupilschoolexpenditures.

Whatcouldleadtothischange?InthefifthrowofTable3weexplorethedifferentialeffectsofforeignvsdomesticinvestmentonthefractionofK᎐12studentsinthecountywhoattendpublicschools.Weobservethatforeign-plantemploymentissignificantlyrelatedtothefractionofstudentsattendingpublicschoolsinacounty.Anewforeignplantofaveragesizeisassociatedwithabout0.11percentagepointsfewerstudentsattending

GermannationalslocatedintheCarolinasapparentlyavoidsendingtheirchildrentopublicschoolsinfavorofprivateones,accordingtoaMay4,1993WallStreetJournalarticle.Also,trainingandeducationprogramsrequestedbyforeignplantsareoftenforapprentice-shipprogramsinareavocationalschoolsthatwilldirectlyleadtoemploymentintheirplant.

18

358

FIGLIOANDBLONIGEN

publicschools.Sincejustover6%ofstudentsattendprivateschoolsinSouthCarolina,onaverage,thissuggeststhataddingasinglenewaverage-sizedforeignplantisassociatedwitha1.8%largerprivateschoolsectorinthecounty.AnewdomesticplantisŽinsignificantly.associatedwithaslightlylargerpublicschoolsectorŽorsmallerprivatesector..Hence,apparentlyemployeesofforeignplantsaredisproportionatelylikelytosendtheirchildrentoprivateschools.

Whatistheeffectofthereductionsinschoolspendingonmeasuredschoolservices?Toaddressthisissue,wecorrelatechangesinforeign-anddomestic-plantemploymentinacountywithchangesinmeasuredschoolservicesintheschooldistrictsinthatcountyusingprivate-accessdatafromtheSchoolsandStaffingSurveysadministeredbytheU.S.DepartmentofEducation.Whilenotapopulationsample,wehaveapanelofobserva-tionsfor52schooldistricts,morethanhalfofallschooldistrictsinthestateofSouthCarolina,forthe1990᎐91and1993᎐94academicyears,theclosestyearsthatwecouldgetto1990and1995.Despitetheloweraverageperpupilexpendituresassociatedwithforeignplants,thereisnopercepti-bleaveragereductioninrealteachersalariesŽfoundbyFigliow11xtobeassociatedwithhigherteacherqualitylevels.orteacher᎐studentratio.However,thereisconsiderableheterogeneityintherelationshipbetweenthesevariablesandforeigninvestment.Specifically,itturnsoutthatforbelow-median-incomeŽinthestate.schooldistricts,theestimatedeffectofforeign-plantemploymentonmeasuredschoolservicesissignificantlymorenegativethantheestimatedeffectofforeign-plantemploymentinabove-median-incomedistricts.Furthermore,itisonlythelower-incomeareasthatseedifferentialmovementtoprivateschoolsandreductionsinpublicschoolexpendituresassociatedwithforeigninvestment.Therefore,itappearsthatemployeesofforeignplantsinlower-incomeareastenddisproportionatelytoenrolltheirchildreninprivateschools,butthistendencyisnotobservedforhigher-incomeareas.Inthelower-incomeareas,increasedforeign-plantemploymentisstronglyassociatedwithlowerlevelsofschoolexpenditureandmeasuredservices.

Notonlydoesforeigninvestmentappeartochangespendingonschoolsinaffectedcommunities,butitalsoappearstochangeaffectedschools’prioritiesaswell.UsingdatafromtheSchoolsandStaffingSurveys,wefindthatschooldistrictswhosecountiesexperienceincreasesintheforeignshareofmanufacturingemploymentaresignificantlymorelikelytointro-ducepoliciesoffreeteacherretraininginmathematics,science,andforeignlanguagesoverthesameperiod.Therefore,wefindsuggestiveevidenceindicatingthatpublicschoolsinareaswithforeigninvestmentgrowthtendtoshifttheirfocustowardscience,mathematics,andforeignlanguageinstruction.

LOCALEFFECTSOFFOREIGNDIRECTINVESTMENT

359

4.3.EmpiricalAnalysisofTransportationandPublicSafetyEffects

Itappearsthatspendingonpubliceducationisnegativelyrelatedtoforeigninvestmentᎏatleastinlow-incomecommunities.Wenextexplorewhethertheseexpendituresareswitchedinparttootherbudgetcate-gories.Forexample,Coughlin,Terza,andArromodeew6xfindthatforeignplantsareattractedtostateswithmoreextensivetransportationinfrastruc-ture,whichsuggestslocalcommunitiesmaydirectmoremoniesintotransportationexpenditures.Toexaminethisissue,weestimatemodelssimilartotheonesdescribedabove,exceptthatnowthedependentvariablesarethefractionsoftotallocalexpendituresgoingtotransporta-tionortopublicsafety.19Here,wehaveobservationsonlyforthreeyearsᎏ1980,1990,and1995ᎏtakenfromtheCityandCountyDataBooksŽinthecaseof1980.ordataprovidedusbytheSouthCarolinaDepartmentofRevenuefortheotheryears.TheresultsoftheseregressionsarereportedinthelasttworowsofTable3.

Weobservethatwhilenewdomesticmanufacturingplantsdonotseemtoaffectthefractionoflocalexpendituresgoingtotransportationorpublicsafety,newforeignplantsapparentlysignificantlyincreasethefractionofexpendituresgoingtotransportation.Whiletherelationshipbetweenforeign-plantemploymentandpublicsafetyexpenditureisstatis-ticallyinsignificantatconventionallevels,thepointestimateonforeignplantsismuchlargerthanthatestimatedfordomestic-plantemployment.Hence,itappearsthatforeigninvestmentleadslocalgovernmentstoredistributefundsfromeducationspendingtospendingontransportationandpossiblypublicsafety.

5.DOESTHESIZEDISTRIBUTIONOFFOREIGN

PLANTSMATTER?

Ourprecedingevidencesuggeststhatforeigninvestmenthasasubstan-tiallydifferenteffectonwagesandbudgetsinlocalcommunitiesthandoesdomesticinvestment.Butthequestionremains:doesthesizedistributionofforeignplantsmatter,oristhesheerfractionofforeign-plantemploy-mentallthatmatters?Thatis,ifacountygets500newforeign-plantjobs,istheeffectonlocalwagesandbudgetsthesameifthe500newjobscomefromonemanufacturingconcern,asopposedto1050-employeemanufac-turingplants?Manyofthestudiesthathaveexamineddifferencesinforeignvsdomesticplants,includingGloberman,Ries,andVertinskyw16x

OurdependentvariabletakesadifferentformhereŽfractionoftotalspending,asopposedtoperpupilspending.thanitdoesregardingeducationduetodatalimitations.Welookatpublicsafetyandtransportationbecausethesearethetwobudgetlineitemsforwhichwehave3yearsofdatathatcorrespondtotheyearsforwhichwehaveinformationonforeigninvestment.

19

360

FIGLIOANDBLONIGEN

TABLE5

DifferentialEffectsofDomestic-andForeign-PlantManufacturingEmploymentonIndustry-Specific,County-SpecificRealWagesandRealCounty-LevelperCapitaBudgetsandBudgetItems,ControllingforConcentrationofForeignPlantsintheIndustryin

theCounty

Effectofanadditionalforeignmanufacturingjobwhenforeignconcentrationinmanufacturingisat25thpercentile0.752Žps0.136.y0.102Žps0.006.y0.100Žps0.028.y0.031Žps0.090.0.019Žps0.262.0.066Žps0.182.

Effectofanadditionalforeignmanufacturingjobwhenforeignconcentrationinmanufacturingisat75thpercentile1.510Žps0.000.y0.077Žps0.042.y0.094Žps0.047.y0.032Žps0.057.0.028Žps0.082.0.062Žps0.185.

DependentvariableRealannualwageŽ1984dollars.1RealpercapitarevenuesŽ1984dollars.2

RealpercapitaexpendituresŽ1984dollars.2

Realperpupilschool

expendituresŽ1984dollars.2FractionoflocalexpendituresgoingtotransportationŽ=100.2FractionoflocalexpendituresgoingtopublicsafetyŽ=100.2

1

Differencebetweencolumns0.758Žps0.002.0.025Žps0.001.0.006Žps0.348.y0.001Žps0.803.0.009Žps0.000.y0.004Žps0.449.

Modelcontrolsforcounty-specificfixedeffectsandindustry-year-specificfixedeffects.Standarderrorsareheteroskedasticity-robustandcorrectforwithin-county-timecorrelationintheerrors.pvaluesareinparentheses.2

Modelcontrolsforcounty-specificfixedeffectsandyeareffects.Standarderrorsareheteroskedasticity-robustŽthereisnowithin-county-timevariation..pvaluesareinparenthe-ses.

andDomsandJensenw8xfindthatcontrollingforsizecansignificantlyaffectestimateddifferences.Thus,ifforeignplantsaresystematicallylargerthantheaveragedomesticplant,ourresultsmaybeexplainingdifferencesinlargeandsmallplants,notforeignanddomesticones.Toexplorethesensitivityofourresults,weestimatemodelssimilartothosepresentedabove,exceptthistimeweallowthemarginaleffectsofforeign-plantemploymenttovarydependingonthemarketconcentrationofforeignplantsintheindustryŽorcounty..

ThefirstrowofTable5presentstheresultsofourestimationoftheequation:

wikts␣eiktq␤fiktq␪fikthikt␥ktq␦i,

LOCALEFFECTSOFFOREIGNDIRECTINVESTMENT

361

wherehrepresentsameasureofthemarketconcentrationofforeignplantsintheindustryinthecounty.Specifically,wecalculatehasthesumoverallforeignplantsintheindustryinthecountyoftheirsquaredmarketsharesŽasafractionoftotalemploymentintheindustry..Therefore,ahighervalueofhreflectsgreaterconcentrationoftheforeignplantsintheindustryinthecounty,andpresumablygreaterinfluenceofanygivenforeignplantinthecounty.

Weobservethattherelationshipbetweenforeign-plantemploymentandwagesisstronglyrelatedtotheconcentrationofforeignplantsintheindustry.Specifically,weestimatethatthemarginaleffectofanadditionalforeign-plantemployeeonwagesistwiceaslargeifthemarketconcentra-tionisatthe75thpercentileinthestate,relativetowhentheconcentra-tionisatthe25thpercentileinthestate.Therefore,itappearsthatwagesinthecountywillincreasemoreifonenewlargeforeignplantenters,asopposedtowhenanumberofsmallerforeignplantswiththesameaggregatelevelofnewemploymententertheindustry.

Wealsofindlimitedevidencesuggestingthatthemarketconcentrationofforeignplantsplaysaroleindetermininglocalbudgetsaswell.Specifi-cally,themarginaleffectofforeigninvestmentonpercapitarevenuesisthree-quartersashighwhenthemarketconcentrationisatthe75thpercentileinthestate,relativetowhentheconcentrationisatthe25thpercentileinthestate.However,thisdifferenceismuchsmallerŽandlesssignificant.inthecaseofpercapitaexpenditures.Withregardtobudgetcategories,theonlycaseinwhichconcentrationofforeignplantsseemstomatterinvolvesthefractionoflocalexpendituresgoingtotransportation.Inthatcase,themarginaleffectofforeigninvestmentonthetransporta-tionspendingshareis47%higherwhenthemarketconcentrationisatthe75thpercentileinthestate,relativetowhentheconcentrationisatthe25thpercentileinthestate.Inmanyways,onemightexpectthelastresultmorethananyoftheothers,astransportationspendingismorelikelytohavelocalizedeffectswithinacountythanwouldotherbudgetspendingcategories;hence,itisreasonabletoexpectthataplantwithrelativelyhighmarketpowershouldbemorelikelytoinfluencespendingontrans-portationthanwouldacollectionofplants,eachwithlowmarketpower.Giventhatthesizeconcentrationofforeignplantsappearstoaffecttheoutcomesdescribedinthepaper,onemightsuspectthattheresultspresentedinTable3arereallyjustlarge-planteffects,ratherthanforeign-planteffects.Inordertogaugethedegreetowhichthisisthecase,weestimateourmodelsreportedinTable3withanadditionalcontrolfortheaverageplantsizeŽmeasuredintermsofnumberofemployees.intheindustryinthecounty,regardlessofownership,calculatedusingdatafromCountyBusinessPatterns.Theresultsofthisexercisearehighlysimilarto

362

FIGLIOANDBLONIGEN

thosereportedinTable3;forinstance,thedifferencebetweentheeffectofforeignanddomesticmanufacturingjobsinthewageequationis1.497Žps0.000.,ratherthanthe1.502reportedinTable3.Otherdifferencesarecomparableaswell.Thissuggeststhatdifferentialplantsizeisunlikelytobethemajordeterminantofthedifferencebetweenourestimatedforeignanddomesticplanteffects.

6.CONCLUSION

Thispaperpresentsevidencethatforeigninvestmenthasconsiderablydifferenteffectsonlocalcommunitiesthandoesdomesticinvestment.UsingdetaileddataonforeignanddomesticinvestmentinSouthCarolina,acrossindustriesandcountiesandovertime,wefindthatforeignplantstendtosignificantlyincreasewagespaidtoworkersinanindustryinalocalcommunity,butalsoleadtosubstantiallylowerpercapitagovern-mentbudgets.Moreover,foreigninvestmentapparentlyinduceschangesinlocalgovernmentbudgetallocations;specifically,wefindevidencesuggest-ingthatcommunitiesexperiencingrelativeincreasesinforeigninvestmenttendtosubstitutefromeducationspendingtospendingontransportationandpublicsafety.

Weacknowledgethattherearelimitationstoouranalysisandresults.Forexample,whileourresultsshowthatthepresenceofforeignplantsisassociatedwithhigherwagesintheindustryinacommunity,ourresultscannotidentifywhetherthisisduetoforeignplantspayinghigherwagestoagivenworker,foreignplantsusinghigherskilledworkerswhocommandhigherpay,orsomealternativeexplanation.Likewise,whileforeignplantpresenceisassociatedwithlowerpercapitabudgets,wehavenotdirectlytestedwhetherofferedtaxincentivesarethesourceofthisresult.Never-theless,ourresultspointtosubstantialdifferencesinhowforeignmanu-facturingplantsaffectlocalcommunitiesvsdomesticones.

REFERENCES

1.B.Aitken,A.Harrison,andR.E.Lipsey,Wagesandforeignownership:Acomparative

studyofMexico,Venezuela,andtheUnitedStates,JournalofInternationalEco-nomics,40,345᎐371Ž1996..

2.T.J.Bartik,‘‘WhoBenefitsfromStateandLocalEconomicDevelopmentPolicies?’’W.

E.UpjohnInstituteforEmploymentResearch,Kalamazoo,MIŽ1991..3.M.Blomstrom,¨Hostcountrybenefitsofforeigninvestments,in‘‘ForeignInvestment,

TechnologyandEconomicGrowth’’ŽD.G.McFetridge,Ed..,UniversityofCalgaryPress,CalgaryŽ1991..

4.B.A.BlonigenandD.N.Figlio,Votingforprotection:Doesdirectforeigninvestment

influencelegislatorbehavior?,AmericanEconomicRe¨iew,88,1002᎐1014Ž1998..

LOCALEFFECTSOFFOREIGNDIRECTINVESTMENT

363

5.R.E.Caves,Multinationalfirms,competitionandproductivityinhost-countryindustries,

Economica,41,176᎐193Ž1979..

6.C.Coughlin,J.V.Terza,andV.Arromodee,Statecharacteristicsandthelocationof

foreigndirectinvestmentwithintheUnitedStates,Re¨iewofEconomicsandStatis-tics,73,675᎐683Ž1991..

7.J.B.Crihfield,Astructuralempiricalanalysisofmetropolitanlabordemand,Journalof

RegionalScience,29,347᎐371Ž1989..

8.M.E.DomsandJ.B.Jensen,Comparingwages,skills,andproductivitybetweendomestic

andforeignownedmanufacturingestablishmentsintheUnitedStates,in‘‘Geogra-phyandOwnershipasBasesforEconomicAccounting’’ŽR.E.Baldwin,R.E.Lipsey,andJ.D.Richardson,Eds..,UniversityofChicagoPress,ChicagoŽ1998..

9.R.C.FeenstraandG.H.Hanson,Foreigndirectinvestmentandrelativewages:Evidence

fromMexico’smaquiladoras,JournalofInternationalEconomics,42,371᎐393Ž1997..

10.Z.FelicianoandR.E.Lipsey,‘‘ForeignOwnershipandWagesintheUnitedStates,

1987᎐1992,’’NationalBureauofEconomicResearchWorkingpaperNo.6923Ž1999..

11.D.N.Figlio,Teachersalariesandteacherquality,EconomicsLetters,55,267᎐271Ž1997..12.P.S.FisherandA.H.Peters,Taxandspendingincentivesandenterprisezones,New

EnglandEconomicRe¨iew,109᎐130ŽMarchrApril1997..

13.E.L.Glaeser,H.D.Kallal,J.A.Scheinkman,andA.Shleifer,Growthincities,Journal

ofPoliticalEconomy,100,1126᎐1152Ž1992..

14.N.J.GlickmanandD.P.Woodward,Thelocationofforeigndirectinvestmentinthe

UnitedStates:Patternsanddeterminants,InternationalRegionalScienceRe¨iew,11,137᎐154Ž1988..

15.N.J.GlickmanandD.P.Woodward,‘‘TheNewCompetitors:HowForeignInvestorsAre

ChangingtheU.S.Economy,’’BasicBooks,NewYorkŽ1989..

16.S.Globerman,J.C.Ries,andI.Vertinsky,Theeconomicperformanceofforeign

affiliatesinCanada,CanadianJournalofEconomics,27,143᎐156Ž1994..

17.E.M.GrahamandP.R.Krugman,‘‘ForeignDirectInvestmentintheUnitedStates,’’

ThirdEdition,InstituteforInternationalEconomics,Washington,D.C.Ž1995..

18.K.Head,J.Ries,andD.Swenson,Agglomerationbenefitsandlocationchoice:Evidence

fromJapanesemanufacturinginvestmentsintheUnitedStates,JournalofInterna-tionalEconomics,38,223᎐247Ž1995..

19.N.G.HowensteinandW.J.Zeile,Characteristicsofforeign-ownedU.S.manufacturing

establishments,Sur¨eyofCurrentBusiness,74,34᎐59ŽJanuary1994..

20.E.Janeba,Taxcompetitionwhengovernmentslackcommitment:Excesscapacityasa

countervailingthreat,AmericanEconomicRe¨iew,forthcoming.

21.S.T.Marston,Twoviewsofthegeographicdistributionofunemployment,Quarterly

JournalofEconomics,50,57᎐79Ž1985..

22.B.Moulton,Anillustrationofapitfallinestimatingtheeffectsofaggregatevariableson

microunits,Re¨iewofEconomicsandStatistics,72,334᎐338ŽMay1990..

23.M.O’Connor,Cities,countiestacklingeconomicinfrastructureissues,IndustrialDe¨elop-mentandSiteSelectionHandbook,1028᎐1052ŽOctober1987..

24.D.V.Rainey,‘‘BusinessIncentives:ProjectedFiscalCosts,’’StromThurmondInstitute

WorkingPaperSeriesŽ1997..

25.D.G.TerklaandP.B.Doeringer,Explainingvariationsinemploymentgrowth:Struc-turalandcyclicalchangeamongstatesandlocalareas,JournalofUrbanEconomics,29,329᎐348Ž1991..

26.R.H.Topel,Locallabormarkets,JournalofPoliticalEconomy,94,S111᎐S143Ž1986..27.D.P.Woodward,LocationaldeterminantsofJapanesemanufacturingstart-upsinthe

UnitedStates,SouthernJournalofEconomics,58,690᎐708Ž1992..