1.The-Geographic-Concentration-of-Industry-Does-Na

AmericanEconomicAssociationTheGeographicConcentrationofIndustry:DoesNaturalAdvantageExplainAgglomeration?Author(s):GlennEllisonandEdwardL.GlaeserSource:TheAmericanEconomicReview,Vol.89,No.2,PapersandProceedingsoftheOneHundredEleventhAnnualMeetingoftheAmericanEconomicAssociation(May,1999),pp.311-316Publishedby:AmericanEconomicAssociationStableURL::29/01/201107:29YouruseoftheJSTORarchiveindicatesyouracceptanceofJSTOR'sTermsandConditionsofUse,availableat.://=aea..EachcopyofanypartofaJSTORtransmissionmustcontainthesamecopyrightnoticethatappearsonthescreenorprintedpageofsuchtransmission.JSTORisanot-for-profitservicethathelpsscholars,researchers,andstudentsdiscover,use,andbuilduponawiderangeofcontentinatrusteddigitalarchive.Weuseinformationtechnologyandtoolstoincreaseproductivityandfacilitatenewformsofscholarship.FormoreinformationaboutJSTOR,[e-mail]ioniscollaboratingwithJSTORtodigitize,preserveandextendaccesstoTheAmericanEconomicReview.*ScholarsinmanyfieldsofeconomicshavebecomeveryinterestedinSiliconValley-styleagglomerationsofindividualindustries(J.VernonHenderson,1988;MichaelE.Porter,1990;PaulKrugman,1991).Theseagglom-erationsarestrikingfeaturesoftheeconomiclandscapeandmayprovideinsightsintothenatureoftheincreasing-returnstechnologiesandspilloversthatarethoughtbymanytobebehindendogenousgrowthandbusinesscycles.Inourpreviouswork(EllisonandGlaeser,1997),wenotedthatagglomerationsmayariseintwoways.Inadditiontoexplanationsbasedonlocalizedindustry-specificspillovers,thereisasimpleralternative:anindustrywillbeagglomeratediffirmslocateinareasthathavenaturalcostadvantages.Forexample,thewineindustry(thesecondmostagglomeratedindustryinourstudy)issurelyaffectedbythesuitabilityofstates'climatesforgrowinggrapes.Iffirms'locationdecisionsarehighlysensitivetocostdifferences(asfoundbyDennisCarlton[1983],TimothyJ.Bartik[1985],andHenderson[1997],amongoth-ers),thennaturaladvantagesmayaccountforasubstantialportionofobservedgeographicconcentration.Inthispaperweusethetermnaturalad-vantagefairlybroadly.Somepossibleex-ampleswouldbethatnoneofthemorethan100,000shipbuildingworkersintheU.S.in1987workedinColorado,Montana,orNorthDakotaandthatthehighestconcentrationofaluminumproduction(whichuseselectricityintensively)isinWashington(whichhasthelowestelectricityprices).WewillalsospeakoftheconcentrationoftherubberandplasticfootwearindustryinNorthCarolina,Florida,andMaineanditsabsencefromAlaskaandMichiganaspossiblyreflectingnaturaladvan-tagesinthelabormarket.Theindustryisanintensiveuserofunskilledlaborandfacestre-mendouscompetitionfromimports;hencewewouldexpecttoseeitlocateinlow-wagestates.Thesimplestwaytofindeffectsofnaturaladvantagesonindustrylocationsistoregresseachindustry'sstate-levelemploymentonstates'resourceendowmentsasinSukkooKim(1999).Aproblemwiththisapproach,however,isthatonecaneasilythinkofmorepotentialadvantagesthantherearestatesintheUnitedStatesandfiteachindustry'semploy-mentdistributionperfectly.Weidentifyef-fectsofnaturaladvantageswithoutoverfittingbyimposingcross-industryrestrictionsre-quiringthesensitivityoflocationdecisionstothecostofaparticularinputtoberelatedtotheintensitywithwhichtheindustryusestheinput.Usingsuchanapproachtoestimatetheef-fectsofnaturaladvantageonthe1987loca-tionsoffour-digitmanufacturingindustries,wefindthatindustrylocationsarerelatedtoresourceandlabor-marketnaturaladvantages.Ourprimarygoalistoseehowmuchofthegeographicconcentrationofindustriesre-portedinEllisonandGlaeser(1997)canbeIDiscussants:Yannisloannides,TuftsUniversity;ThomasNechyba,StanfordUniversity;SukkooKim,WashingtonUniversity-St.Louis.*Ellison:DepartmentofEconomics,MassachusettsInstituteofTechnology,Cambridge,MA02139,andNBER;Glaeser:DepartmentofEconomics,HarvardUni-versity,CambridgeMA02138,andNBER.WethanktheNationalScienceFoundationforsupportthroughgrantsSBR-9515076andSES-9601764.Ellisonwasalsosup-portedbyaSloanResearchFellowship.WethankYannisloannidesforvaluablecommentsandDavidHwang,SteveJens,AlanSorensen,andMarcusStanleyforresearchassistance.311312AEAPAPERSANDPROCEEDINGSMAY1999attributedtonaturaladvantages.Inourpre-ferredspecification,weattributeaboutone-fifthoftheconcentrationtoobservablenaturaladvantages.Giventhatweareusingonlyasmallnumberofvariablesthatcaptureadvan-tagesveryimperfectly,wewouldguessthatatleasthalfoftheconcentrationreportedinEllisonandG