Agglomeration economies and productivity differenc

AgglomerationeconomiesandproductivitydifferencesinUScitiesDavidL.Rigby*andJu¨rgenEssletzbichler**AbstractPlant-leveldatafromtheLongitudinalResearchDatabaseoftheUSBureauoftheCensusareemployedtoestimatetheimpactofagglomerationeconomiesonindustryproductivityacrossUSmetropolitanareas.Thisanalysisseekstoremedythreeshortcomingsofpreviousempiricalstudiesofagglomerationeconomies:relianceonaggregatespatialorsectoraldata;lackofattentiontospatialdependenceindata;andrepresentationofagglomerationeconomieswithvagueproxiessuchascity-size.Weshowhowanumberofestablishment-,industry-,andcity-specificfactorsinfluencelaborproductivityacrossUScities,andwepayparticularattentiontoseparatingtheinfluenceofdifferentkindsofagglomerationeconomiesonfirmefficiency.HerewefollowMarshall’sPrinciplesofEconomicsinexaminingthespatialconcentrationofinput^outputlinkages,thecharacteroflocallaborpoolsandembodiedtechnologicalspillovers.Keywords:agglomerationeconomies,laborproductivity,UScitiesJELclassifications:C21,L6,O30,R12Datesubmitted:15November2001Dateaccepted:8May20021.IntroductionInthispaperweexaminelaborproductivitydifferentialsacrossmetropolitanareasintheUS.Wefindsignificantgeographicalvariationsinproductivitylevelswithinindividualmanufacturingsectorsatdifferentlevelsofaggregation.Usingplant-leveldatafromtheLongitudinalResearchDatabaseoftheUSBureauoftheCensus,weshowhowproductivitylevelsareinfluencedbyavarietyofestablishment,firmandcity-specificfactors.Muchofourattentionisdevotedtoexaminingtheinfluenceofdifferenttypesofagglomerationeconomies,followingMarshall(1920).Inparticular,weshowhowthespatialconcentrationofinput–outputlinkages,theoccupationaldistributionofacity’slaborpoolandembodiedtechnologicalspilloversaffectmetropolitanproductivitylevelsindifferentmanufacturingindustries.ConsiderableresearchefforthasbeendevotedtoexplainingregionalvariationsinproductivityacrosstheUS(Sveikauskus,1975;Moomaw1981a,b;Williams,1985;FogartyandGarofalo,1988;BeesonandHusted,1989;andseethereviewsbyMoomaw,1983a;andGerking,1994).Forthemostpart,however,thisresearchisplaguedbyanumberofseriousdeficiencies.First,itistypicallyconductedatrelativelyhighlevelsofindustrialandgeographicalaggregation.Thisisunderstandablegiventhe&OxfordUniversityPress2002JournalofEconomicGeography2(2002)pp.407–432*DepartmentofGeography,UniversityofCaliforniaLosAngeles,LosAngeles,CA90095–1524,[e-mail]**DepartmentofGeography,UniversityofSouthampton,Southampton,SO171BJ,[e-mail]ta,thoughresearchersinthisareahavebeenslowtotakeadvantageofnewerdatasources.Sveikauskus(1975)calledattentiontotheneedformoredisaggregateindustrialanalysismanyyearsago,andeconomicgeographershavefrequentlyarguedforafocusonfunctionaleconomicregionsratherthanconvenientpoliticaldivisionsasobservationalunits(Isard,1956;Richardson,1978).Second,statisticalanalysisofgeographicaldifferencesinproductivitylevelshasignoredthequestionofspatialdependenceindataandthuspastresultsmustbetreatedwithsomecaution.Third,whiletheexistenceofagglomerationeconomiesiscentraltothestudiesabove,itisalwayspoorlymeasured:typicallyproxiedusingmetropolitanpopulationortheproportionofastate’spopulationresidinginthelargesturbanareas.Intheanalysisbelowweremedytheseshortcomings.AudretschandFeldman(1996,p.630)writerecently,‘Noone,todate,hasexaminedtheunderlyingpropensityforindustrialactivitytoclusterspatially’.Thisassertionismorethanalittleoutrageousgiventhelonghistoryofanalysisineconomicsand,moreparticularly,ingeographythatfocusespreciselyonthisquestion.Perhapsthemostwell-knownandoriginalstatementsaboutthespatialclusteringofindustryandthebenefitsofagglomerationareprovidedbyMarshall(1920).FivechaptersofhisPrinciplesofEconomicsfocusonindustrialorganizationandthelocalizationofindustryfiguresprominentlyinhisdiscussion.‘Whenanindustryhasthuschosenalocalityforitself,itislikelytostaytherelong:sogreataretheadvantageswhichpeoplefollowingthesameskilledtradegetfromnearneighbourhoodtooneanother.Themysteriesoftradebecomenomysteries;butareasitwereintheair,andchildrenlearnmanyofthemunconsciously’(1920,p.271).Marshall’sevocativediscussionof‘industrialatmosphere’drawsonthreespecificbenefitsthatsustainthespatialconcentrationofparticularbranchesofindustry.Hewritesofthegrowthofsubsidiarytrades,ofbuyerandsuppliernetworks,andthedevelopmentofhighlyspecializedmachinerywithinaneighborhood;oflocalmarketsforspecial(labor)skills;andofthediffusionoftechnicalknowledgeortheexchangeofinformation(seealsoMarshall,1885,p.53).Weber(1929)andHoover(1948)werelatertodiscusssomeofthesesameissuesinamorerestrictivelocation-costformat(Richardson,1978).Inrecentyearsthestudyofagglomerationeconomieshasbeenreinvigorated(seethereviewsofMalmberg,1996;SweeneyandFeser,1998;Hanson,2001).Partly,thisisaconsequenceofempiricalworkthatreaffirmedtheasymmetriesofeconomicgrowthfollowingthedemiseoftheFordistregimeofaccumulation(PioreandSabel,1984;Scott,1988,1993).Attemptstodeterminetheformofapost-Fordistfutureconcentratedonthelocationalandorganizationaldynamicsofnewindu