Structural Change and Productivity Growth in Cities Ron Martin*, Peter Sunley**, Ben Gardiner*, Emil Evenhuis* and Peter Tyler*** *Department of Geography, University of Cambridge, UK **School of Geography and Environment, University of Southampton, UK *** Department of Land Economy, University of Cambridge, UK Research Paper 2 December 2016 Structural Transformation, Adaptability and City Economic Evolutions An ESRC Urban Transformations Project (ES/N006135/1)
41
Embed
Structural Change and Productivity Growth in Cities · characterised in terms of the quantitative relations between input and output, between the output, employment and productivity
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Structural Transformation, Adaptability and City Economic Evolutions An ESRC Urban Transformations Project (ES/N006135/1)
1
Abstract
Itisnowwidelyacknowledgedthatstructuralchangeisintegraltotheprocessofeconomicgrowthandtheevolutionofcapitalistdevelopment.Giventheveritableexplosionofinterestinrecentyearsincitiesas‘engines’ofwealthcreation,trade,innovation and creativity, the issue of structural change would seem highlyrelevant to understanding the evolving economic performance of cities,particularly given the ongoing debates over structural specialisation versusdiversity. Thispaperexaminesthedifferingproductivitygrowthpathsofsome85Britishcitiessincethebeginningofthe1970s,andexploreshowfarthesepathsreflectdifferencesacrosscitiesinthepaceandnatureofstructuralchange.Wefirst find evidence that while productivity tended to converge across citiesbetween1971-1991, thereafterconvergenceceasedandwasreplacedbyweakdivergence. The paper then analyses the extent and nature of structuraltransformation in the various cities, using particular measures applied to 82sectors of activity, between 1971 and 2014.We find evidence of considerablestructural convergence across cities and a general tendency for the degree ofspecialisationtofall.Thisthenleadstoadecompositionanalysiswhichidentifiesthe relative contribution of within-sector and between-sector (sectoral re-orientation and relocation) effects to city productivity growth. The analysisrevealsthatwithin-sectorproductivitydevelopmentsoutweighstructuralchangeinaccountingfordifferencesinproductivitygrowthacrossmostBritishcities.Assuch, the paper raises questions over the importance often assigned tospecialisation as amotor of city growth, and points to the role of city-specificfactorsthatinfluencethegrowthperformanceacrossmostofacity’ssectors.
Keywords:CitiesStructuralchangeProductivitygrowth
JELClassificationsR10O47R11
2
Productivity isn't everything, but in the long run it is almosteverything.(PaulKrugman,1994,p.11)Productivity is the challenge of our time… The gap in labourproductivitybetween theUK’s two largestcityeconomies,LondonandManchester,islargerthaninanyotherG7countryandmorethandoublethatinbothGermanyandJapan.Adynamiceconomyneedsthrivingcities.(HMTreasury,2015,pp.3,68.)
1.IntroductionItisnowwidelyacknowledgedthatstructuralchangeisintegraltotheprocessofeconomicgrowthandtheevolutionofcapitalistdevelopment:asRoncolatoandKucera (2014,p.399)put it, “sustainableeconomicgrowth requires structuraltransformation”.Akeycharacteristicofstructuraltransformationisthechangingsectoral composition of employment and output over time, a stylised fact forwhichthereiscopiousevidence(Kuznets,1957,1971;Pasinetti,1993;FreemanandLouca,2001;CornwallandCornwall,1994;Metcalfe,FosterandRamlogan,2006; Kruger, 2008). Development economists of the structuralist persuasionhave long been concerned with theorising major long-run shifts ortransformationsineconomicstructure,particularlytheprogressiveshiftfromlowproductivity activities to high productivity ones, and the long-run progressionfromprimaryproductiontomanufacturingtotertiaryactivity(see,forexample,Ocampoetal,2009).Partofthisconcern,infactthefocusofearlydebate,focusedon the relative role of sectoral specialisation versus sectorally ‘balanced’developmentas thebasis forachievingproductivityadvance. Incontrast,withrespecttoadvancedeconomies, traditionalgrowththeoryalwayshaddifficultyincorporating structural change. Technical advance, a key driver of economicgrowth,wasintroducedintothattheoryinsuchawayastomakeitcompatiblewiththebasicconceptionofaneconomyinwhichtherewasnostructuralchange,characterised in terms of the quantitative relations between input and output,between theoutput,employmentandproductivityofdifferent sectors,and thepattern of relative prices and incomes. In order to be consistent with thisconception,technicalchangehadtobeassumedastakingplaceatauniformratein all sectors, and demand had to be made to grow at a uniform rate for allproducts,bothassumptionsthatwereclearlyunrealistic.It took the path-breakingwork of authors such as Kaldor (1966, 1967, 1968),Kuznets(1973)andPasinetti(1981,1993)tomovestructuralchangetocentrestageingrowththeory.ThusaccordingtoKuznets
rapid changes in production structure are inevitable – given thedifferentialimpactoftechnologicalinnovationsontheseveralproductionsectors, the differing income elasticity of domestic demand for variousconsumergoods,andthechangingcomparativeadvantageinforeigntrade
3
(1973,p.250).Likewise,inPasinetti’sscheme,structuralchangeisconceivedasamulti-sectoraleconomy evolving through time under the influence of technical progress andchangesinfinaldemandconsumption.Technicalchangeoccursunevenlyamongsectors,sothattherateofchangeofproductivitydiffersfromsectortosector(andby implication from region to region).1 Correspondingly, demand changes atdifferentratesamongdifferentproducts.Moreover,technicalchangemaytaketheform of the introduction of new products, and hence the emergence of newactivities and new sectors. In short, structural dynamics are inherent to thegrowth process. In a later contribution, Baumol et al (1989) discuss theconsiderablediversityofproductivitydevelopmentsthatcanbeobservedacrossindustriesandsectors,andemphasisenotonlythefactthatstructuralchangeisan ongoing long-run phenomenon, but also that productivity growth isparticularly relevant in the long run. In Kaldor’s seminalworks on economicgrowththeory,manufacturingwasassignedparticularimportanceasthedriverof economic growth: in his view the more rapid the expansion of themanufacturingsector relative to the restof theeconomy, the faster the rateofgrowthofproductivitynotjustinmanufacturingitself,butintheeconomyasawhole, primarily because of dynamic returns to scale and spillovers to othersectors(Kaldor,opcit;seealsoThirlwall,1983).Yet,asKruger(2008)pointsout,despitethekeyrelevanceofstructuralchangeforgrowththeory,businesscycletheory and labourmarket theory, as well as for economic policy, the topic ofstructural change and its potential relevance for productivity growth, arefrequentlyneglectedtopicsineconomicresearch.On the face of it, these same topics should also be of central relevance forunderstandingtheevolvingeconomicperformanceofcities.Giventheveritableexplosionofinterestinrecentyearsincitiesas‘engines’ofwealthcreation,trade,innovationandcreativity(Jacobs,1984;Glaeser,2011;Moretti,2013),theissueofstructuralchangewouldseemhighlyrelevanttounderstandingcitygrowthanddevelopment. Within the overlapping subfields of urban economics, regionalscienceandthe ‘neweconomicgeography’,avastbodyof literaturenowexiststhatexplores the importanceof theexternaleconomiesand increasingreturnseffectsthatarisefromtheconcentrationofeconomicactivity(firms,workersandconsumers)incities,andthepositiveimpactofthatconcentrationonproductivity,innovationandwages.Thereisalsoagrowingliteratureonhowcityeconomiesevolveovertimeandontheircomparativegrowthpaths(seeforexample,Glaeser,2005;MarkusenandShrock,2006;Moretti,2012;Poweretal,2010;Hobor,2013;Dijkstraetal,2013;Michaelsetal,2013;Cowell,2014,Storperetal,2015;Martin 1 Interestingly,inexplainingthestimulusforhisnewtheory,Pasinettiattributeditinpartto“theextremelyunevendevelopment–fromsectortosector,fromregiontoregion-oftheenvironmentinwhichIlived(post-warEurope)atthetimeIbeganmytrainingineconomics”(Pasinetti,1981;p.xi).
4
etal,2016).Oneofthemanyfactstohaveemergedfromthisburgeoningbodyofworkisthatcitiesappeartodifferintheirratesofgrowthofoutput,employmentandproductivity. Inmanyinstancesthegrowthpathsofcitieshavebeenquitedivergent.IntheUnitedStatesforexample,Storperetal(2015)findthatsincethebeginningofthe1970sthetwolargecitiesofSanFranciscoandLosAngeleshavefolloweddistinctlydifferentgrowthpaths,withtheformerpullingprogressivelyahead of the latter in terms of per capita income. The authors attribute thisdivergenceinlargeparttodifferencesinspecialisationbetweenthetwocities.Asimilar pattern of divergence in per capita incomes and employment in facttypifiesnumerousUScities(Moretti,2013;Martin2016).IntheUK,Martinetal,(2016)findasimilarpicture:citygrowthpathsthereovertheperiod1971-2014havebeenhighlydivergent,withmostcitiesinthesouthofthecountrygrowingfasterintermsofbothemploymentandoutputthanthenationalaverageandmostnortherncitiesgrowingslower,withtheresultthatthetwogroupshavepulledprogressively apart over the past four to five decades. Other studies of citiesacrossEuropeandmorewidelyacrosstheOECDnationshavealsofoundmarkeddifferencesingrowthrates(Dijkstraetal,2013).Thereareofcoursenumerouspossiblefactorsthatmightaccountforsuchunevengrowthacrosscities,evenwithinthesamenationalmacro-economy. Citiesaresubject to national-level fiscal, monetary, legal, regulatory and politicalarrangements, all of which influence city growth rates. And most cities areindirectly,ifnotdirectlyaffectedbyglobalforces,eventsandrelationships.Butthespecificsofeachcityalsomatter,andservenotonlytoshapetheparticularimpactofnationalandglobalfactorsandprocessesfromcitytocity,butalsoactas indigenous sources of growth and development in their own right. Severaldeterminants of city prosperity and productivity have been suggested in theliterature, including city and size and density, both being assumed to conferpositive agglomeration economies; human capital, especially skilled, highlyeducated,andcreativeworkers;goodconnectivity,bothinternallyandexternallywith other cities; and supportive and strategic institutions and governancearrangements,tonamebutsome.Inaddition,andthefocusinthispaper,isthequestion of how and to what extent the structural configuration of a city’seconomy(itsensembleofactivities,industriesandfirms),andmoreparticularlyhow changes in that structure over time, contribute to the evolution of cityproductivitytrajectories.Morespecifically,usinganewanddetaileddatasetthepaperinvestigateshowaggregateproductivityhaschangedandevolvedacrosssome85Britishcitiesovertheperiod1971-2014,andhowfarthesepatternsreflectdifferencesacrosscitiesintheimpactofstructuralchange.Thepasthalfcenturyhaswitnessedhistoricshifts and changes in the British economy, most notably intensedeindustrialisation and the rapid growth of services, especially knowledge-intensive business services.How these shifts haveworkedout acrossBritain’s
5
citiesisnotonlyofinterestinitsownright,butmayhaveadirectbearingoncityproductivitygrowthpaths.ThenatureofthosepathshasaparticularsalienceintheBritishcontext,giventhatthelacklustreproductivityperformanceofnationaleconomy has itself been a recurring issue of concern. Since 2010, the UKGovernmenthasrepeatedlyvoiceditsdesireto‘spatiallyrebalance’theeconomy,havingacknowledgedthatnationaleconomicgrowthhasbecometoodependentonanarrowrangeofactivities,especiallyfinance,concentratedinonepartofthecountry, namely London and the surrounding South East region. Promotinggrowthandproductivityoutsidetheseareas,andespeciallyincitiesinthenorthofthecountry,hasbecomeakeytopicofnationalpolicydiscourse.Anewmodelof limited fiscal and political devolution has been become part of the policyresponse(Sandford,2016).Yetmorerecently,theGovernmenthasdeclareditsintention to develop an industrial strategy aimed at rebalancing the economysectorally and spatially,with a view to raising the productivity of the nationaleconomy as a whole (HM Treasury, 2015; other refs). Further, the UK’sreferendumvotetoleavetheEuropeanUnion–so-called‘Brexit’-makestheneedtoimprovetheproductivityofthenation’scitiesandregionsallthemoreurgent,giventhattheyarecouldwellfacetariffsontheirexportstoEuropeandwillalsoneedtolooktootheroverseasmarketstoexporttheirgoodsandservices.
Webeginbyfirstreviewingthe literatureonstructuralchangeandhowthis relates to the ongoing debate within economic geography and urbaneconomicsontherelativemeritsofstructuralspecialisationversusdiversityasthebasisforcitysuccess.WethenexaminetheproductivityevolutionsofBritishcitiesoverthepastfortyyearsorso.Thisisfollowedbyananalysisofwhathasbeenhappeningtotheeconomicstructuresofthecitiesinquestion.Akeyissuewithinthiscontextiswhethercityeconomicstructureshavebecomemoreorlessspecialised,andwhethertheyhaveconvergedordiverged.Thepaperthenmovesontoassessthecontributionofstructuralchangetotheevolutionofproductivityacrossthestudycities.2.StructuralChangeandProductivityThereisnogeneralorwidelyacceptedeconomictheoryofstructuralchangeandits effects onproductivity, but instead a variety of theoretical approaches (seeKruger,2008,foranexcellentsurvey).Nevertheless,twotypesofargumenttendtorecur.Inone,itisthetime-evolvingincomeelasticitiesofdemandthatshapethepaceanddirectionofstructuralchange:overtime,thedemandforparticularproductsorservicesapproachessaturation,andthis,togetherwithrisingincomeeffects,stimulatestheemergenceofnewproductsandnewsectorsofactivity.Thegrowthindemandexperiencedbythesesectorsallowsincreasingreturnseffectswhich in their turnpromotehigherproductivity.Other thingsbeingequal, thisincreaseinproductivitywilltendtoreducerelativeprices,whichthenstimulates
6
yetfurtherdemandforthenewproducts,sothattheseactivitieswillaccountforlargersharesofemploymentandoutput(thisisessentiallytheargumentbehindthe Kaldorian model of export-led, increasing returns-based, regionaldevelopment). In the other type of argument, it is the differential rates oftechnological advance that drive structural change: sectorswith a high rate oftechnological progress and labour productivity gain in importance over time,whereassectorswithalowrateoftechnologicalprogressandlabourproductivitylose ground and therefore suffer in terms of employment and value added.However,inpracticetheremaybeimportantfeedbackeffectsbetweenchangesinstructureandproductivity.Increasesinproductivityleadtoincreasesinincome,andthecompositionofthelatterwillinfluencethepatternofdemand,andhencestructuralchange.Therelationshipbetweenstructuralchangeandproductivitygrowthisthuslikelytoinvolvecirculardependencies.
Atthesametime,technologicaladvancesassociatedwiththeemergenceof new industries and sectors can spill over to existing activities, raising theirproductivity,thoughthiscanresultinreducedlabourinputs.Indeed,thetechnical,organisationalandproductupgradingwithinexisting(‘old’)sectorsmayleadtomajorproductivityadvancesintheseactivities,thoughpossiblyattheexpenseofemployment.Yetfurther,newsectorsofactivitymayexperiencesustainedgrowthindemand,andhenceinoutput,buthavelimitedscopeforproductivitygrowth,andmayevenhavelowerthanaverageproductivity.Inthiscase,theoutputandemployment shares of these sectors may rise, but the consequence for theaggregateeconomymayactuallybeaslowingofproductivitygrowth.
Therehasinfactbeenconsiderabledebateovertheapparentslowdowninproductivitygrowth inmanyadvancedeconomiesover thepast threedecades.Someattributethisapparentdeclinetomeasurementproblems,tothefactthattechnological advances and shifts simply do not show up in conventionalmeasures of (both labour and total factor) productivity (the so-called ‘SolowProductivity Paradox’ – see Triplett, 1999; Crafts, 2002). Others dispute thisargument, and argue that the slowdown is real (Cowen 2016; Gordon, 2016;Syverson,2016).AccordingtoGordon(2016)andCowen(2016),forexample,innovationhasstalled,andtechnologicalprogressnolongerproducethegainsinGDPthatitoncedid.AsimilarviewisespousedbyCowen(2016),thathigh-techdevelopments have not saved advanced economies from a slowdown inproductivity.Yetanotherexplanationpointstothefallinbusinessdynamismoverthepasttwotothreedecades(EuropeanCentralBank,2016),asreflectedinnewfirm formation rates: new firms are assumed to embody more advancedtechnologyandtobemoreproductivethanoldexistingfirms.WhilemuchofthedebatehascentredontheUnitesStateseconomy,similarconcernanddiscussionhave also surrounded the productivity performances of other OECD countries(OECD,2015),includingtheUK.
provokeddivergentviews,withsomeobserversarguingthatmanyservices(suchasretail,hospitality,personalservices,andevensomeprofessionalandbusinessservices)have limitedpotential forhighproductivitygrowth,andmayevenbe‘stagnant’asfarasproductivityisconcerned(Baumol,1967;Baumoletal,1985;Williamson,1991;Kim,2006).Otherauthors,however,pointtothefactthatmanyservicesfunctionasintermediaryinputstothemanufacturingsector,andmaynotonlyhelptoraisetheproductivityofthelatter,butthemselvesmayhaveasmuchscope for increasing their own productivity (Oulton, 2001). The trend formanufacturing firms to outsource certain routine service activities that werepreviouslycarriedout ‘inhouse’,whileatthesametimeoftendevelopingtheirowncustomer-orientatedserviceactivities(fromfinancetoafter-care),maywellalsohaveimpactedonthemeasurementandallocationofproductivityadvancebetween‘manufacturing’and‘services’incomplexandnotunambiguousways.
What is clear is that the relationship between structural change andproductivitygrowthisnotasimplenorself-evidentone:productivitygrowthcanbeofa‘within-sector’kind,arisingfromtechnologicalandrelatedimprovementsamongstfirmsinthatsector;and‘between-sector’,orstructural-changeinduced,arisingfromtheshiftandreallocationofcapitalandlabourfromlowproductivitygrowthsectorsintohigherproductivitygrowthones(seeTable1). Therehavebeenvariousrecentattemptstoassesstherelativecontributionofsuch‘within’and‘between’effectstoaggregateproductivitygrowth.Someofthesehaveusedsuch a ‘within’ and ‘between’ sectoral decomposition of a particular nation’sproductivity growth, andothers to examinedifferences inproductivity growthbetweencountries(Fagerberg,2000;Pieper,2000;Peneder,2003;Kruger,2006;Ocampoetal,2009;TimmeranddeVries,2009;andKuceraandRoncolato,2012).Althoughtheresultsvaryacrosstimeperiods,datafrequency,whetherstructureis measured by employment shares or output shares, the choice of labourproductivity or total factor productivity, and the particular decompositiontechnique used, the balance of the findings is that the ‘within-sector’ effectdominatesthe‘between-sector’effect(ie.theeffectduetostructuralchange),withthestudybyMcMillanandRodrikalsofindingthisforsome,thoughnotall,ofthecountriesintheirstudy.Haltiwanger(2000)hasarguedthatstructuralchangeismuchmore intensewithin industries (that is among the firmswithindifferentsub-branchesofanindustry)thatbetweenindustries,evenatdetailedlevelsofindustrialdisaggregation.Yet thevarious studies thathaveuseda ‘within’ and‘between’firmdecompositiontoinvestigateproductivitygrowthofagivenbroadsector(Bailyetal,1992,2001;Fosteretal,1998;BartlesmanandDoms,2000;Disneyatel,2003,CantnerandKruger,2006)alsotendtofindthe‘within-firm’effectisgreaterthanthe‘between-firm’effect.Asfarasweareaware,thissortofdecomposition analysis has not been conducted to examine the productivitygrowthpathofcities,whichisouraiminwhatfollows.First,however,usingournovel data set, we explore how British cities have actually fared in terms ofproductivitygrowth.
3.ProductivityGrowthPathsofBritishCitiesAsmentionedabove,theissueofproductivitygrowthintheBritisheconomyhaslongbeenarecurringtopicofacademicandpoliticaldebate.Figure1showshowlabourproductivitygrowthintheUKeconomyhasbeenhighlyvariablefromyearto year since 1950. However, certain features are evident. First, the rate ofproductivitygrowthactuallytrendedupwardsfrom1950to1973,butsincethenthe trend has been on a downward path. In fact,while theUK annual rate ofgrowthover1971-1991averaged2.13percent,over1991-2014theaveragefellto1.46percent.Howfarthe‘turningpoint’intheunderlyingtrendofproductivitygrowtharound1973reflectedtheendofthepost-warexpansionofemploymentin manufacturing, or the impact of the recession of the early-1970s (itselfassociatedwith the firstof theOPECoilpricehikes),are interestingquestions.While all of the major sectors of the economy have experienced productivitygrowth,itisevidentthatsincethebeginningofthe1980s,productivityadvancein manufacturing or production (manufacturing plus construction) has beenfaster than inmarket (private)services,especially if financeand insuranceareexcludedfromthelatter(Figure2).Shiftsintheeconomicstructureofcitiesawayfrommanufacturingtowardsservicesmightbeexpected,therefore,tohavehadpotentiallysignificantimplicationsforcityaggregateproductivity.Figure1:AnnualGrowthRateofLabourProductivityintheUK,1950-2015
9
Source:OfficeforNationalStatistics
With regard to British cities, an immediate major issue is the lack of anycomprehensive and consistent official economic time series data. As part of alargerprogrammeofresearchoncityeconomicevolutions,wehaveconstructednewdataseriesonemployment,grossvalueadded(GVA),andlabourproductivity(GVAperemployedworker)forsome85citiesfor82sectorsofactivity,yearlyfrom1971to2014(andon249sectors from1991to2014). Wehavedefinedthese cities spatially on the basis of the 2011 travel-to-work area (TTWA)boundariesasdelineatedbytheUKOfficeforNationalStatistics.Therearesome228TTWAscoveringthewholeoftheUK.ThecriteriafordefiningTTWAsisthatgenerally75percentofanarea’sresidentworkforceworkintheareaandatleast75 percent of those who work in the area also live there. For TTWAs with aworking population of 25,000 or more, self-containment rates as low as 66.7percentareaccepted.OnlythoseareasbasedonanidentifiablecitywithatotalTTWApopulationofat least200,000wereselectedasourfinalsetof85cities.Together these cities, in 2014, accounted for some 83 percent of Britishemployment and85percent of British output (gross value added).2They thusmakeupthebulkofthenationaleconomy.
2 Because of lack of basic data we were unable to include any cities in Northern Ireland.Correspondingly,weuseGreatBritainratherthantheUKasthebaseforsuchcalculations.
!4#
!3#
!2#
!1#
0#
1#
2#
3#
4#
5#
6#
1950#
1953#
1956#
1959#
1962#
1965#
1968#
1971#
1974#
1977#
1980#
1983#
1986#
1989#
1992#
1995#
1998#
2001#
2004#
2007#
2010#
2013#
Annu
al#Growth#in#Labou
r#Produ
c>vity#(G
VA#
per#E
mployed
#Person),#percent#
Trend#Line#is#Best#Fit#Poynomial#(Second#Order)#
10
Figure 2: Labour Productivity in Selected Major Sectors of the BritishEconomy (Gross Value Added per Employedworker, 2005 Prices), 1971-2009
Thevariationinlabourproductivitylevelsacrossthe85citiesin1971andin2014isshowninFigure3.Thecitieshavebeengroupedinto ‘northern’andsouthern’setsaccordingtotheregionoftheirlocation,usingtheconventionalwayof dividing theUK into these twobroad geographical areas.Also shown is thenational average (Great Britain) productivity level for the two years. What isstrikingisthatallbarthreenortherncities(Crew,ChesterandTelford)areinthebottomleft-handquadrantoftheFigure,havingproductivitylevelslessthanthenationalaveragebothatthebeginningoftheperiodandattheend.However,atthe same time, the correlation between productivity levels in 1971 and 2014,thoughsignificant(R=0.686)isnotparticularlyhigh, indicatingcertainshiftsinrelativepositionovertheperiod;inotherwords,productivitygrowthratesacrosscitieshavenotbeenproportionate.
Inthiscontext,an interestingfeatureemerges intherelationshipacrosscitiesbetweentheirinitialproductivitylevelsandtheirsubsequentproductivitygrowthwhenthewholestudyperiodisdividedintothetwosub-periods(Figure4). In the first sub-period, 1971-1991, which corresponds to when nationalproductivity advance was higher, the relationship is negative (R=-0.581). Anegative relationship indicates that cities which had initially low labourproductivitylevelstendedsubsequentlytoexperiencefasterproductivitygrowth,thatisto‘catchup’withcitiesthatinitiallyhadhigherproductivitylevels.Thisisin linewith thepredictionsofstandardneoclassicalgrowththeory.But for thesecondsub-period,1991-2014,therelationshipisnolongernegativebutweaklypositive(R=0.109).ThechangingpatternevidentinFigure4indicatesthatwhileBritish cities converged in terms of productivity in the 1970 and 1980s, thistendencydisappearedover the1990sand2000s. Thedetailed timepathsofproductivityforsomeselectedcitiesillustratethischangeddynamic(seeFigure5,inwhichtheseriesareindexedatGreatBritain=100).
Forexample,Oxford,Reading,MedwayandStevenage,initiallyhighproductivitycities, all lost ground against the national average until around 1991, whileSunderland, Doncaster and Merthyr Tydfil, initially low productivity cities, all
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
10000 15000 20000 25000 30000 35000
AverageAn
nualGrowthofLab
our
Prod
uctiv
ity(perce
nt)
LabourProductivity,1971
Period1971-1991
Oxford
Reading
MedwayLeamington
HighWycombe
MilfordKeynes
London
Sunderland
BlythMansfield
NorthernCities
York
MerthyrTydfil
SouthernCities
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
20000 25000 30000 35000 40000
AverageAn
nualGrowthofLab
our
Prod
uctiv
ity,1991-2014(Percent)
LabourProductivity,1991
Period1991-2014NorthernCitiesSouthernCities
LondonReading
TunbridgeWells
Oxford
GuilfordStevenage
Preston YorkSwansea
Blackpool
Exeter
Cardiff
13
gainedground.Thegeneralpatternofmoderateconvergence isclearlyevident(aside from London). From 1991 onwards, however, divergence is apparent.London has continued to pull ahead throughout the period, while Liverpool’sproductivityhasbarelykeptupwiththatnationally.Further,andsignificantly,ifwegroupthe85citiesintothoseinthe‘south’ofBritain,andthoseinthe‘north’,thereisclearevidenceofa‘switch’inrelativeproductivitygrowthbetweenthesetwogroupsbetween1971-1991and1991-2014,withnortherncitiesoutpacingsoutherncitiesinthefirstsub-period,butthelatterout-performingtheformerinthemorerecentsub-period(Table2).
Thekeyquestionof interesthere iswhether,how farand inwhatways, thesechangingpatternsofproductivityacrosstheBritishcitieshavebeeninfluencedstructuraleconomicchangesincethebeginningofthe1970s.Toexplorethisissue,wefirstexaminehowtheeconomicstructuresofthecitiesthemselveshavebeenevolvingoverthisperiod.4.CityEconomicStructureandStructuralChangeThere are numerous indicators which can be used to capture local economicstructureandwithwhichtocomparedifferentareasatanyparticularpointintime.EarlydiscussionsofarangeofalternativemeasurescanbefoundinIsardetal(1960)andBahletal(1971).Morerecently,Palan(2010)reviewednofewerthanninedifferentmeasures. Someof thesewere ‘absolute’measures, inwhich theactualsectoraldistributionofalocalarea’semploymentoroutputiscomparedtoa defined ‘base-line’ metric, usually that of a perfectly equal distribution ofsectoralshares;whileotherswere‘relative’measures,wherethecomparisoniswithsomechosen‘norm’or‘reference’economy,forexamplethenationalsectoralstructure.Palanassessedthevariousmeasuresfortheextent
Source:Authors’owndatatowhichtheysatisfycertaincriteria(orwhatshecalledaxioms)3andthenusedthemtoexamineandcomparetheeconomicstructuresofEuropeancountries.Shefoundthatthedifferentmeasuresdonotyieldconsistentresults,andthatnooneapproach could be said to be superior: to some extent the choice ofmeasuresdependsonthespecificpurposeinmind.
One of themost commonly usedmeasures of city or regional economicstructureisthecoefficientofrelativespecialization(seeIsard,1960;ThirlwallandDixon,1975).ThishasdeployedbyKrugmanonanumberofoccasionstoexaminecity and regional specialization (Krugman, 1991, 1993), and for that reason isoftencalledthe‘KrugmanIndex’.Ittakestheform
3 Palan lists four such axioms: anonymity, or invariance to ordering of sectors; progressivetransfers, or the principle of rank-preserving equalisation; having defined upper and lowerbounds;anddecomposabilityintowithinandbetweencomponents(suchassectorsorfirms).
of sectors involved in the analysis. The sum is over the absolute differencesbetween the industry shares of a given city’s employment (or output) and thecorresponding industryshares in thereferenceeconomy.Asdefined, the indextakesthevalueofzerowhenacity(orregion)hasexactlythesamestructureasthereferenceeconomy(sinceeachabsolutesectoralsharedifferencein(1)woulditselfbezero).Theindexincreasesthemorethatthecity’s(region’s)economicstructurediffersfromthereferenceeconomy.Ifthecity(region)sharednosectorincommonwiththereferenceeconomy(whichmightbeanothercityorregion),then the maximum value the index would be 2, since each absolute sectoraldifference in (1)wouldbe counted in full.This isnot the case,however, if thenationaleconomyistakenasthereferencenorm,sincebydefinitionthenationaleconomymustshareatleastonesectorincommonwithatleastoneofitscities(regions).Inthiscasetheupperboundoftheindexis2[(N-1)]/N.Alternatively,the ‘national reference economy’ could be calculated individually for each city(region)astheweightedsumofallothercities(regions)minusthecity(region)inquestion.Inthiscaseitispossiblethatthecityand‘adjusted’nationalreferenceeconomydonotshareallofthesamesectors,sothattheupperlimitof2wouldagainbeappropriate.
AccordingtoKrugman,theindexisa“roughwayofquantifyingdifferencesinstructures,andhenceregionalspecialization”(1993,p.250).Strictlyspeaking,however,ittellsusmoreaboutstructuraldissimilaritybetweenregions,orcities,thanaboutregionalorcityspecializationperse,sinceeveniftheindexforacityis close to zero, suggesting little difference from the reference economy, thereferenceeconomyitselfcouldbenarrowlyspecialisedinparticularsectors,sointhiscaseboththecityandthenationwouldbeequallyandsimilarlyspecialized.
Thusanadditionalmeasureisrequiredinordertocapturewhetheracityis absolutely specialisedordiversified economically. Theobvious approach tomeasuring the degree of diversity of a city’s economic structure is to compareactual sectoral (employment or output) shares against an equi-proportiondistributionofshares,thatisastateofcompletediversityorbalancedstructure.Thereareseveralpossiblemeasuresthatcanbeusedforthispurpose(see, forexample,Isardetal,1960;GibbsandPostan,1975;Kruger,2006;Palan,2010).TheseincludeShannon’sEntropyIndex(forexample,AigingerandDavies,2004;AigingerandPfaffermayr,2004),theIndexofInequalityinProductionStructure(see Cuadrado-Roura et al., 1999; Haaland et al., 1999; Landesmann, 2000;Percocoetal.,2005), theTheil Index(BrulhartandTraeger,2005;EzcurraandPascual,2007),and theHirschman-Herfindahl Index(forexample,Sapir,1996;Davis, 1998; Storper et al, 2002; Aiginger and Pfaffermayr, 2004; Beine andCoulombe, 2007). As Palan (2010) shows, although a simple measure, theHirschman-Herfindahlsatisfiesmostofthekeycriteriarequiredforausefuland
16
meaningfulindexofdiversity.The Hirchman-Herfindahl index is defined as the sum of the squared sectoralshares, 𝐻𝐻𝐼$% = 𝑠)$%2(
)*+ (2)
where,asinEquation(1),theshares𝑠)$%areexpressedasproportionsofunity.The index ranges from aminimum of 1/N, when all sectoral shares are equal(maximumdiversity)toanupperboundof1,inwhichcaseacitywouldbemono-specialised,that isallof itsactivityis in justoneindustry.Becausethesectoralsharesaresquared,theindexgivesmoreweighttolargesectors.Forthisreason,thesquarerootoftheindexissometimesused(forexample,ChisholmandOeppen,1973).
BoththeCRS(Krugman)IndexandtheHirschman-HerfindahlIndexcanbeusedforidentifyingandtrackingstructuralchangeinindividualcitiesandregionsby comparing values of the indices at different points in time.4 If structure ischanging,thefirstplacetolookisatthepatternsoftheemploymentandoutputsharesandthechangestheyevinceovertime.InthecaseoftheCRS,byusingthenationaleconomyas thereferenceeconomy, the indexcan illuminatewhether,howfarandhowfast,cityeconomicstructuresareconverging(decliningvaluesoftheindex),ordiverging(increasingvaluesoftheindex).NotethattheCRScanbealsousedtochartthechangingeconomicstructureofacityrelativetoitsown‘starting’structure,atsayt=0,bysettingthereference‘norm’𝑠)$%∗ in(1)to𝑠)$3.Inthisinstance,structuralchangewouldbeindicatedbyrisingvaluesoftheindexovertime,asthecityincreasinglydivergedfromitsoriginalmixofsectors.WithrespecttotheHHI, ifthereisproportionalgrowthacrosssectors,andhencenostructural change, the indexwould remain constant over time (Metcalfe et al,2006).Changesintheindexthusindicatestructuralchange:successivevaluesthatmovedtowards1/Novertimewould indicate increasingequality(diversity) ineconomic structure, whereas a trend towards 1 would indicate increasingspecialisation. The rate of change of theHHI is proportional to the covariancebetween employment (or output) shares and employment (or output) growthrates:
That pronounced ongoing structural changes have transformed the UKeconomyoverthepastfiftyyearsorsoisclearlyevidentfromFigures6and7.Inbroadterms,themostprofoundchangehasbeentheshiftfromaneconomybasedon production industries (manufacturing, construction and utilities) to onedominated by privatemarket services. The decline in production employmentfromitspeakofjustover11million(or41percentoftotaljobs)in1966to5.5million (19 percent) in 2010 represents one of the most rapid rates ofdeindustrialisation inthewesternworld(Ref).Likewise,having increasedoverthetwodecadesaftertheSecondWar,theshareofproductionindustriesintotaloutputsteadilyincreasedtoreachapeakof38percentin1969,andthereafterprogressively declined, falling to 22 percent by 2009. At the same timeemploymentinprivatemarketservicesincreasedfrom8.8million(34percent)in1969 to 14.4 million (50 percent) in 2009. If we add in local and centralgovernment,theserviceeconomyincreaseditsshareoftotalemploymentfrom53percent in1969to80percent in2009,and itsshareof totalnationalGrossValueAddedfrom38percentto55percentoverthesameperiod. Themacro-structureofthenationaleconomyontheeveofthefinancialcrisisin2007lookedverydifferentindeedfromthatin1970.
Howhave thesestructuralshiftsandchangesworkedoutacrossBritishcities? The calculated CRS (Krugman) indices of structural specialisation byemploymentandbyoutputacross82sectorsforthe85citiesfor1971,1991and2014 are given in Tables 3 and 4 respectively. For each city, the referenceeconomyinEquation(1)wasdefinedasGreatBritainminusthecityinquestion,soastoavoiddoublecounting(whichwouldnotbeinsignificantinthecaseofthelargest cities such as London, Birmingham, Manchester, Sheffield, Liverpool,Glasgow and Edinburgh). Several key features stand out. First, in 1971, citiesdifferedmarkedlyinthedegreeofrelativestructuralspecialisation,whether interms of employment or output. Second, in both instances, the large cities(regional capitals) and Londonwere less specialised thanmost other, smallercities.Third,inthecaseofemploymentstructure,allbutonecity(Slough)haveexperiencedadeclineinrelativespecialisationsince1971.Thetrendsinoutputstructures are broadly similar, although some thirteen cities experienced an
ThecorrespondingHHindicesfor1971,1991and2014aregiveninTables5and6,withthedetailedtemporaltrendsforselectedcitiesshowninFigures10and11. Theseshowseveral interesting features. Ingeneral, cities tend tobemorespecializedintermsofoutputstructuresthaninemploymentstructures;thiswasespecially the case in the 1970s and 1980s. With respect to employmentstructures,morethanhalfofthecitiesexperiencedadeclineinspecializationovertheperiod1971-2014.Thosecitiesthatweremorespecializedinitiallyunderwentthelargestdeclines,whilesomeofthecitiesleastspecialisedinitiallyexperiencedan increase inemploymentspecialization.TheHHstructural indices foroutputshares,however, showamoredefinitepattern,withmost citiesbecoming lessspecializedoverthefourdecades.AsinthecaseoftheKrugmanindices,itwouldappearthatthedeclineinspecialisationwasmostevidentinthe1971-1991subperiod,andthatstructuralchangesincethenhasbeenslower.Table 5: Hirschman-Herfindahl Employment Specialisation Indices forBritishCities,(82sectors),1971,1991and2014(Citiesrankedindescendingorderofspecialisationfor1971________________________________________________________________________________________________
Figure 10: Sectoral Diversification in Selected Cities: The Hirschman-HerfindahlIndexofEmploymentShares,1971-2014.(AsmallervalueofHHindicatesgreatersectoraldiversityorbalance)
Figure11:SectoralDiversificationinSelectedCities:TheHHindexofOutputShares, 1971-2014 (A smaller value of HH indicates greater sectoraldiversityorbalance)
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.11971
1973
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
2005
2007
2009
2011
2013
HHIn
dex
Oxford
Sunderland
Huddersfield
Stoke-on-trent
Halifax
Dudley
Trowbridge
Bradford
Middlesbrough
Reading
0"
0.02"
0.04"
0.06"
0.08"
0.1"
0.12"
0.14"
0.16"
1971"
1973"
1975"
1977"
1979"
1981"
1983"
1985"
1987"
1989"
1991"
1993"
1995"
1997"
1999"
2001"
2003"
2005"
2007"
2009"
2011"
2013"
HH"Inde
x"
Oxford"
Trowbridge"
Plymouth"
Gloucester"
Portsmouth"
Basingstoke"
Reading"
Exeter"
Swansea"
Medway"
25
5. Structural Change and Productivity Growth Across BritishCitiesToprovidebackgroundtothecityanalysis,Figure12showsthenational–levelrelationshipbetweenproductivitygrowthandthechangeinemploymentshareforour82sectorsoverthewholeperiod1971-2014.Severalfeaturesareevident.Almostallof themanufacturingsectorshaveexperienceda fall inemploymentshare,consistentwithdeindustrialisationtrenddepictedinFigure6.theinmostcases of between 2-4 percent per annum. But all except ‘coke and refinedpetroleum products’ also experienced productivity growth, in most cases ofbetween2-4percentperannum.Somekeysectors,however,achievedratesmuchhigher than this, of between 4 and 6.5 percent per annum: pharmaceuticals,transportequipment,chemicals,computerequipment,andvehicles.Inthecaseofprivatemarketservices,thesefallintotwomaingroups:oneinwhichannualproductivity growth has averaged around 2.5 percent, but employment shareshavechangedonlymarginally(eg.FinancialServices,Insurance,),andanotherinwhichemploymentshareshavegrownbuttheannualrateofproductivitygrowthhas averaged only around 1.0 percent per annum (eg. Food Services, BuildingServices, Head Offices, Management Consultancies). Information services, airtransport,financeandcomputerprogrammingserviceshaverecordedhighratesofproductivitygrowth,thoughonlythelastofthesesawanysignificantincreasein employment share. What is particularly striking is that, overall, there is anegativerelationshipbetweenproductivitygrowthandemploymentshareacrosssectors(R=-0.201):theemploymentstructureofthenationaleconomyhasbeenawayfromthosesectors–manufacturing,constructionandutilities-whichhaveshownthehighestratesofproductivitygrowthintothosesectors–privateandpublicservices–mostofwhichhaveexperiencedtheslowestratesofproductivitygrowth.HowfarthishelpstoaccountforthedecliningtrendinnationalaggregateproductivitygrowthshowninFigure1isthusaninterestingquestioninitsownright.Ourinteresthere,however,isintherelationshipbetweenstructuralchangeandproductivitygrowthacrossBritain’scities.Asmentionedabove,variousauthorshavesoughttoidentifytherelativecontributionof‘within-sector’and‘between-sector’ (structural change) contributions to country-level and industry-levelproductivity growth. For example, in their study of differences in productivityperformance among developing countries, McMillan and Rodrik (2011) use asimpledecompositionoftheabsolutechangeinproductivity,
)*+ 𝑦)$% (4)where𝑌$%and𝑦)$%refer,inourcase,tototalandsector-specificlabourproductivitylevels(realGVAperemployedworker)incityjattimet.The∆operatordenotesthechange, inproductivityandinemploymentshares,betweent-kandt. Thefirsttermin(4),whatMcMillanandRodrikcallthe‘withinsector’contribution,istheweighted sum of productivity changewithin individual sectors, arising forexample from improvements in production technique or the other factorsidentifiedinTable1,wheretheweights𝑠)$%arethesharesofcityemploymentinthose sectors at the beginning of the time period concerned. The second termcapturestheproductivityeffectofemploymentreallocationsacrossorbetweendifferentsectors.Itisessentiallytheinnerproductofsectoralproductivitylevels(attheendofthetimeperiod)withthechangeinemploymentsharesacrossthosesectors(sincethebeginningoftheperiod).ItisthiscomponentthatMcMillanandRodrikrefertoas ‘structuralchange’.Whenchangesinemploymentsharesarepositively correlated with productivity levels, this term will be positive, andstructuralchangewillhaveoperatedtoincreasethecity-widerateofproductivity
AvariantofthisdecompositionapproachisusedbyKruger(2006),inhisstudyofstudyofproductivitygrowthinUSmanufacturing(seealsoDisney,etal,2003;Foster,etal,1998;KuceraandRoncaolato,2012;RoncolatoandKucera,2014). In this version, a city’s aggregateproductivity growth (rather than theabsolutechangeinproductivitystudiedbyMcMillanandRodrik)isdecomposed,usingKruger’snotation,intothreecomponents:∆GHIJKGHI
= LMHI∆NMHIJKOMPQ
GHI+ ∆LMHI(NMHI:GHI)
OMPQ
GHI+ ∆LMHIJK∆NMHIJK
OMPQ
GHI
(5)
where, as in (4), 𝑌)$% and 𝑦)$% refer to total and sector-specific labourproductivity levels in city j, and the∆operator denotes the change inproductivityoremploymentsharesbetweentandt+k.Thefirsttermontheright-hand sideof (5) is interpretedas the ‘within-sector’ effect,which is the share-weightedaverageproductivitygrowthof the individual industries incity j.Thesecondtermrepresentsthe‘between-sector’effect.Itispositiveifsectorsinitiallywith above average productivity levels experience increasing shares betweenperiodtandt+1onaverage,andindustrieswithbelow-averageproductivitylevelsexperiencefallingsharesoftotalcityemployment,onaverage.Thethirdtermisaninteractionor‘covariance-effect’whichispositiveifindustrieswithhighratesofproductivitygrowthtendtogainintermsoftheirshares(ormoregenerally,ifsharechangeandproductivitygrowthtendtohavethesamesign).The‘betweeneffect’ and ‘covariance effect’ together reflect the role of structural change inaggregatecityproductivitygrowth.5
HereweuseKruger’s formula (Equation (5)). The results are shown inFigure 12 and Table 7. Figure 12 plots the three right-hand components ofpercentage productivity change in Equation (5) against the total percentageproductivitychange,forthewholeperiod1971-2017foreachofthe85cities.
5 Itisfurtherworthnotinginpassingthatintheliterature,severalmodificationsandextensionsofthetwodecompositionmeasuresinEquations(4)and(5)havebeenproposed(Bailyetal,1992;GrilichesandRegev,1995;OlleyandPakes,1995;Fosteretal,1998;FagerbergandPeneder,2000;Disneyetal,2003).Forexample,Bailyetal(1992)andFosteretal(1998)deriveversionswithadditional terms that represent the contributions of entering and exiting establishments toaggregateproductivitygrowth.Theseeffectscannotbeinvestigatedhereforthetimeperiodthatisofinterest.
28
As is clear, the overwhelming contribution to total productivity change acrossalmost all of the cities was from ‘within-sector’ improvements. The ‘between-sector’contributionwhilstgenerallypositive,ismuchsmallerinvirtuallyeverycase.Thereareonlyfivecitiesinwhichthebetween-sectorcontributionisgreaterthan the within-sector contribution: Birkenhead, Chester, Crawley, HighWycombeandHull.Thecovariancecomponent,bycontrast,isnegativeinmostcities, and is larger (in absolute terms) than the between-sector or structuralcontributioninmorethanhalfofthecities.Thisisconsistentwithadeclineovertimeintheemploymentsharesofthosesectorswithhigherproductivitygrowth–thatis,manufacturingandproduction.Giventhatthisnegativecovarianceeffectgenerally exceeds the small positive between-sector component, the overall‘structuraleffect’onproductivityacrossmostcitieswouldappeartohavebeennegative. Table 7,which shows the detailed decomposition for the top 10 andbottom10 cities in termsof total productivity growth for the two sub-periods1971-1991and1991-2017,confirmsthesefindings.Figure12.DecompositionofCityProductivityGrowth,1971-2014,into
These findings are somewhat surprising, in that they suggest that structuralchangehasplayedaminor,andinmostcasesanegativeroleinaccountingforthevariationsinproductivitygrowthacrossBritishcities.However, inthisrespect,theydo tend tobe consistentwith the resultsobtained fromstudies thathavesoughttoassessthesignificanceofdifferentialstructuralchangeinaccountingforthepatternsofproductivitygrowthacrosscountries,asreferredtoinSection2above. And given that our analysis in Section 4 suggests that city economic
30
structureshaveconverged,theimplicationisthatcitydifferencesinproductivitygrowth havemainly reflected differential productivity performancewithin thesame sectors: put anotherway, itwould appear that the same sector performsdifferentlyacrossdifferentcities.Amuch finer levelof industrialdisaggregationmight change this conclusion, and give more (possibly positive) weight tobetween-sectorandstructuraleffects.On theotherhand, the finer thegrainofsectoraldisaggregation,thesmallerarethecorrespondingsectoralemploymentweights are likely to be, which may not therefore do much to change theimportanceofwithin-sectorandstructuralchangeeffects.6.ConclusionsThispaperhasidentifiedan‘urbanproductivitypuzzle’toaddtothenumerousotherpuzzlesandproblems thatbeset researchonproductivity.Wehaveseenthat while sectoral structures in British cities have converged since the early1970s,afterconvergingover19971-1991,cityproductivitydifferenceshavesincediverged.Wehavefoundagrowingimbalanceintheproductivityperformanceofcitieswhichundoubtedlyhasserious implications for the long-termseverityofspatialinequalitiesinBritain.Itmightbeexpectedthattheconvergenceofurbaneconomies on service sectors would have led to a convergence in theirproductivity,butthishasnotbeenthecase.Howthencanwestarttoexplainthisconundrum? For a start, productivity variations within service industries aredifficulttomeasurebutappeartobelarge(BailyandSolow,2001).Moreover,theresultsofourdecompositionanalysisofcityproductivitygrowthshowshowsthatgrowth has been dominated by within-industry productivity developments,rather than by the transfer of employment and output across industries, orstructural change. Underlying this are the substantial variations in bothproductivitylevelsandgrowththatexistacrossplantsandindustrieswithinthesameindustryorsector.Inthissense,ourresultsconfirmthefindingsofearlierstudies that emphasise that in mature industrialised economies there arepersistent and large productivity differentialswithin individual industries andsectorswhichtendtodominateproductivitygrowth(BartelsmanandDoms,2000;Haltiwanger,2000;Krüger,2006).Thechallengethisposes ishowcanwebestexplainthesourcesoftheseintra-industryspatialvariationsinproductivity?Thereareseveralpossiblesourcesofthisurbangeographyofproductivitygrowth.The first, of course, is that agglomeration economies and localised increasingreturnsmaybenefit firmstoagreaterdegree insomecitiesthanothers.WhiletherearenosimplerelationshipsbetweencitysizeandproductivitygrowthintheBritish case (see Martin et al, 2014), the possible effects of agglomerationeconomiesandespecially thoseacross related industries,need tobeexamined
31
morecarefully(forexample,seeEssletzbichlerandRigby,2002),aswellastheimportanceoftraveltimestoLondon(Webberetal,2009).However,giventhattheweight of economic research shows that agglomeration only has at best amodest or marginal effect on productivity – the bulk of the evidence on thissuggeststhatadoublingofcitysizeseemstobeassociatedwithanincreasesinproductivityofonlybetweenabout4-8percent(seealso,RosenthalandStrange,2003)-wesuspectthatagglomerationeconomiesareatbestasmallcontributoryfactorratherthanamajordriver.6ReturningtoTable1,itisevidentthattherearetwokeycausesofwithin-sectorproductivitychange.Thefirstisa‘recompositionorreallocationeffect’andinvolvestheentryandexitoffirmsandthere-allocationofmarketsharesbetweenincumbentfirms.Ingeneralahigherrateoffirmandplant entry leads to faster productivity growth as new entrants tend to havehigherproductivitythanthosethatexitorareclosed.Iflargeandwell-organizedfirmsandplantsgainmarketsharethiswillalsoofcoursepushupproductivitygrowth.Thusvariedentrepreneurialdynamicsandlargefirminvestmentsinnewplants across cities will strongly shape their productivity growth. The secondmajorsetofprocessescentresontechnologicalandorganizationalchangeamongsurviving firms which includes both the adoption of innovations as well asmanagement,organizationalpracticesandformats.Typicallytheseareshapedbythe intensity of competition faced by firms, and by their regulatory andinstitutionalcontext,andintheUKtheyareoftenproxiedbytheamountofcapitalemployedperworkerandlinkedtoforeignownershipofthefirm(SeeWebberetal,2009).Existingindustryresearchimpliesthatbothofthesetwoprocessesarelikelyresponsiblefortheintra-industryurbanvariationsinproductivitythatwehave found (Disneyetal,2003),although the relative importanceof these twoprocessesmaychangeindifferentperiods(see,forexample,RileyandBondibene,2016).HarrisandMoffat(2015)arguethatfirmentryandclosurehavebeenthemost importantcauseofchange to total factorproductivitydifferentialsacrossLocal Enterprise Partnership areas in the UK, but struggle to link this to theeconomic characteristics of these areas. We clearly need more research todisentangletherelativeimportanceandcausesofthesetwosetsofprocessesindifferentcities.Itishighlylikelythatthetwoarecombinedincitiesinreinforcingwayssothatthosecitieswiththemostcompetitiveandglobalisedmarketsandhighestratesofplantcreationandforeigndirectinvestmentwillseehigherratesoffirmchurnandentry,aswellashigherratesofinnovationandorganisationalimprovement. The strength of these processes in some cities may haveimplicationsforunderstandingourfindingsoncityproductivitygrowth.
6There is also the intriguing issue of whether the effects of agglomeration may increase up to a certain size of city but thereafter either increase at a decreasing rate or even begin to decline. Forexample, PotterandWatts(2011)findthatagglomerationeconomiesmayfirstriseandthenfallovertime.
32
Oneofthekeylimitationsofourdecompositionanalysis,ofcourse,isitsrelianceonahistorically-derivedclassificationsoffirmsintodistinctindustries.Whilewehaveusedthemostdetailedsectoralbreakdownpossibleforthelongtimespanstudiedhere, thesecategoriesandrepresentationsnevertheless continue tobebasedontheassumptionthatindustrialcategoriesaremeaningful,andthatfirmswithinindustriessharesomekeycharacteristicsandfeatures.Thereisagrowingrecognition,however,thattheseclassificationsarenotcapturingformsofactivitychange and restructuring that are widening differences within particularindustries.Manyindustriesnowincludefirmsthatarequitedifferentintermsoftheoccupationstheyinvolve,themarketstheyreach,andthetasksandfunctionsthat theyperform(seeBaldwin,2016).Partly,ofcourse, this isduetothenewdivisions of labour emerging from supply chain re-organisation and thespecialisationsofareasandcitiesinspecifictasks,stagesandoccupationsratherthan in complete supply chains. In several ways the uneven diffusion ofglobalisationhaswideneddifferenceswithinindustries.ItalsomaybeduetothewaysinwhichICTanddigitisationareboundupwithfirmentryandexit,andarechangingfirmactivitiesandleadingtotheemergenceofdigitalactivitiesthatareblurring industry boundaries (including, in some instances, between whatconstitutes ‘manufacturing’ and ‘services’), and pulling apart the productivityperformancesof firms. In this context, revisions to industryclassificationsarelaggingwellbehindthegrowthofnewactivities.Whatthismayimplyisthatthegeography of ‘structural change’ is no longer well measured by changes inindustrialclassesandcategoriesbutneedstobeanalysedinamorefinegrainedwaywithin particular industries, for example in terms of occupational or task‘bundles’.Atpresenthowever,theseconclusionsarespeculative,andourfutureresearch will endeavour to examine in more depth some of these processesunderlyingthepuzzleofdivergenceinproductivityacrosscontemporaryurbanBritain.