AlexanderC.LembckeOECD/GOV,RegionalDevelopmentPolicyDivision

DivergentCitiesConferenceCambridge,16July2015

Productive citiesEvidence from the OECD

Motivation

2

• Acountry’sproductivityis,inlargepart,determinedbytheproductivityofitscities;understandinghowtoincreasetheproductivityofthesecitiesisthereforeanurgentpolicyquestion.

• Largeurbanagglomerationsaccountforover50%oftotalGDPwhiletakinguplessthan5%oftotalsurfacearea.

• Itiswellknownthat,inmanycountries,economicproductivityincreaseswithcitysize.Thisisinpartaresultofsorting,asbettereducatedindividualshaveatendencytoliveandworkinlargercities.However,productivityincreasesarepresentevenwhencontrollingforsorting‐ aswedoforthisproject.

• Usingcitydefinitionsbasedonfunctionsratherthanadministrativeboundaries(FUA),thepapercontraststhecomplexityofadministrativeboundariesbyFUAandhenceexaminestheroleofurbangovernanceintheeconomicproductivityofacity.

Evidencefor5+2OECDcountries:• Germany,Mexico,Spain,UKandUS• Netherlands andJapaninprogress

Coherentandcomparativemethodology• Usewagemicrodata &OECDMetropolitanAreasDatabase:Citydefinitionbasedoneconomiclinks(commutingflows),ratherthanadministrativeboundaries

• Two‐stepeconometricframework accountforindividualcharacteristicsthatdetermineproductivity

• Estimateagglomerationbenefits,examinetheroleofurbangovernanceandothercharacteristicsforcityproductivityand

Background for the presentation

3

• Sizematters:Estimatedelasticityforagglomerationbenefitsaround0.02‐0.06(roughly:2‐6%higherproductivityforadoublinginpopulation)

• Productivityishigherincitieswithmoreskilledworkers; withlarger employmentsharesinhigh‐techmanufacturing,financeandbusinessservices; largerneighbouringcitiesandlowerlevelsofgovernmentalfragmentation.

• Citieswithfragmentedgovernancestructuresexhibitlowerlevelsofproductivity:Doublingthenumberofmunicipalitieswithinametropolitanareaisassociatedwith6%lowerproductivity

• Thepresenceofagovernancebodyhalvesthispenalty

City productivity: Results from 5 (+1) OECD countries

4

• DefinitionofFunctionalUrbanAreasbasedonpopulation densityin1km2 cellsthatarematchedtomunicipalboundariesandconnectedviacommutingpatterns.

• Urbancentresareidentifiedbyaggregatingdenselypopulated1km2 cells.Urbancentreswithatleast50,000inhabitantsarekept.

• Theyarematchedwiththeboundariesofthelowestadministrativelevelforwhichstatisticaldataistypicallyavailable(NUTS5/LAU2)

• UrbancentresandthelessdenselypopulatedmunicipalitiesinthecommutingzonearecombinedintoFunctionalUrbanAreasbasedoncommutingflows(>15%).

• Moreinfo:OECD(2012)RedefiningUrban

A functional definition for cities

5

OECD metropolitan areashttp://measuringurban.oecd.org

6

Larger Metropolitan Areas are more productive (higher GDP per worker)

7

Source:OECDMetropolitanExplorer.LabourProductivityismeasuredasGDP(MillionsofUS$constantPPP,constantprices,referenceyear 2005)dividedbythetotalnumberofemployeesinaFunctionalUrbanArea.Datareferto2010ortheclosestavailableyear.

• ResultsfromAhrend,Farchy,Kaplanis andLembcke (2014).EstimationfollowsCombes,Duranton andGobillon(JEconGeo,2011)

• Usemicrodataandfollowa2‐stepapproach:i.applyindividualwageregressionsinordertoestimatethedifferentialproductivitylevelsofcities,controllingforsorting(onobservables)ii.explainthedifferentialcityproductivitylevelsfoundinthefirststepbyregressingthemonanumberofcityexplanatoryvariables

• Accountsforbiasarisingfromnon‐randomsortingofmoreproductive(e.g.skilled)individualsacrosscities

Empirical strategy: Two step approach

8

• Firststep:

yiat:wage(ln)Xiat:gender,education,experience,occupation,part‐timingdiat:city‐yearFE

• Secondstep:

• Allowsforflexiblecountrytimetrendsdct• Samplelimitsampletoyearsavailableforallcountries(2005‐2007)

Empirical strategy: Two step estimation

9

• ExplainfirststageestimateswithasetofrelevantFUAlevelcontrols( ; )– agglomeration:population;densityandarea(logged)– humancapital:shareofuniversitygraduates– industrialstructure:specialisation,compositionandtechnologicalintensity

– urbangovernance:administrativefragmentation(lnnumberofmunicipalities;governancebody)

– location“fundamentals”:port;capitalcity;popincitieswithin300kmradius

Empirical strategy: Explaining productivity differentials

10

• Microdata withinfoonwagesandworkplacelocation• UK:ASHEfirststage(1%sample);city‐yearcontrolsfromQLFS;2003‐2010

• Spain:MCVL;administrativedata(4%sample);2005‐2011• Germany:EmploymentpaneloftheGermanFederalEmploymentAgency(2%sample);German weakly anonymous BA-Employment Panel (version 1998-2007)

• US:IPUMSdata;Census1990,2000andAmericanCommunitySurvey2005‐2007

• Mexico:ENE2000‐2004;ENOE2005‐2010

Data

11

City productivity increases with city size- even after controlling for sorting

12

Pooled regression results, 2005-2007

13

(I) (II) (III) (IV)ln(population) 0.038***

(0.005)ln(density) 0.038*** 0.045*** 0.045***

(0.006) (0.006) (0.006)ln(area) 0.038*** 0.070*** 0.070***

(0.006) (0.009) (0.009)ln(municipalit.) -0.032*** -0.035***

(0.006) (0.006)Log (popul. in 0.017**catchment area) (0.008)R-Squared 0.76 0.76 0.78 0.78Observations 1,290 1,290 1,290 1,290FUAs 430 430 430 430

‐ Effectisdisaggregatedintoapuresizeeffectmeasuredby(log)areacoveredbyFUAandagglomerationbenefitscapturedby(log)populationdensity

‐ Analysisonthepooleddataestimateselasticityofaround0.04

• SourcesofagglomerationfromMarshall(1890);reviewsbyRosenthalandStrange(2004),Puga (2010)

• Thickerlabourmarkets:labourmarketpooling;bettermatching• gainfromreducedlabouracquisitionandtrainingcostsinthicklocal

labourmarketswithabundantspecialisedlabourforce

• Sharing facilities,inputs,gainsfromspecialisation• firms may face lower costs for specialised non-traded inputs that are

shared locally in a geographical cluster.

• Knowledgespillovers• face-to-face contact can enable tacit knowledge spillovers through

increases in the intensity of the interactions with other firms or individuals

Sources of agglomeration benefits

14

Heterogeneity: bigger is better

15

Spain

UnitedStates

Heterogeneity: borders matter(ed)

16

Germany

Mexico

Heterogeneity: distance matters

17

Netherlands

UnitedKingdom

Heterogeneity: distance matters

18

Netherlands

UnitedKingdom

excludingFUAsthatborderametropolitanarea(lightblue)

Distance matters

19

ProductivitydifferentialsanddistancetoLondon

AnnualpercapitaGDPgrowthrates(1995‐2010)anddrivingtimetotheclosestmetroareaof2millionormoreinhabitants

Spillovers of large cities: Economic growth in EU regions

20

• Inmanycountries,thegovernancestructureofurbanareasisinadequate,asadministrativebordersdonotcorrespondtotheeconomicarea.

• Thisdiscrepancymightleadtoconflictsofinterestamongthevariousagents nottherightscaleforpolicy(lackofpoliciesorduplication)(CheshireandGordon,1996;CheshireandMagrini,2009)

• Someareasaffectedbycoordinationproblems:– transport;stymietransportinfrastructureinvestmentscongestion

– landuse;inadequatelanduseplanning urbansprawl– easeofdoingbusiness;additionalbureaucracyunderinvestment

The role of governance

21

Berlin

Madrid

Administrative fragmentation in two OECD metropolitan areas

22

City productivity decreases with fragmentation

23

All FUAs Metro areas Metro areas Metro areas

ln(density) 0.048*** 0.064*** 0.065*** 0.047***

(0.006) (0.012) (0.012) (0.012)

ln(area) 0.064*** 0.082*** 0.085*** 0.087***

(0.008) (0.012) (0.012) (0.013)

ln(municipalit.) -0.032*** -0.032*** -0.057*** -0.066***(0.006) (0.010) (0.016) (0.017)

ln(municipalit.) 0.031** 0.036**x govern. body (0.014) (0.015)

governance body -0.079** -0.092**

(0.034) (0.038)

Add.controls no no no yesR-Squared 0.779 0.847 0.855 0.880Observations 1,290 420 420 420FUAs 430 140 140 140

Fragmentation and governance bodies

24

• Theempiricalanalysisconfirmsthattheproductivityofresidentsincreaseswithcitysize.Butitalsoshowsheterogeneityacrosscountries

• Humancapital,high‐techandknowledgeintensiveservices orientedcitiesmaketheirresidentsmoreproductive

• Smallercitiescan“borrow”agglomerationbenefits.• Importantly,theanalysisshowsthatadministrativefragmentationreducesproductivity,apenaltythatisalleviated(butnotnegated)bygovernancebodies.

Concluding remarks

25

Thepresentationdrawsfrom:Ahrend,Farchy,Kaplanis andLembcke (2014)“WhatMakesCitiesMoreProductive?Agglomerationeconomiesandtheroleofurbangovernance:Evidencefrom5OECDCountries”Ahrend andSchumann(2014)“DoesregionaleconomicGgowthdependonproximitytourbancentres?”OECD(2015)TheMetropolitanCentury:UnderstandingUrbanisationanditsConsequencesOECD(2015)GoverningtheCityOECD(forthcoming)MetropolitanReview:Rotterdam‐TheHagueOECD(2012)RedefiningUrban:anewwaytomeasuremetropolitanareas

Thank you

26

Variables USA MEX DEU GBR ESPpopulation 2.47 0.88 0.47 0.46 0.39

(3.115) (2.211) (.615) (1.169) (.810)

area 11.2 4.3 1.2 0.7 1.6(12.907) (6.989) (1.136) (.878) (2.205)

pop.density 0.35 0.41 0.51 0.74 0.63(.379) (.542) (.513) (.518) (.913)

% university 0.28 0.16 0.14 0.19 0.12graduates (.056) (.051) (.041) (.07) (.054)

Capital 0.01 0.01 0.01 0.01 0.01(.12) (.115) (.096) (.1) (.115)

Port 0.48 0.19 0.17 0.16 0.32(.503) (.392) (.381) (.367) (.468)

Herfindahl 0.06 0.06 0.05 0.05 0.07index (.006) (.018) (.013) (.009) (.025)

# local governments 5.0 5.4 44.3 4.3 31.8per 1,000 inh. (3.8) (10.1) (50.4) (7.0) (52.2)

% of pop.in 0.67 0.81 0.52 0.72 0.72largest municip. (.233) (.232) (.183) (.221) (.165)

Governance 0.87 0.81 0.83 0.29 0.13body (.341) (.402) (.381) (.469) (.354)

Summary statistics:Regression controls

27

UK Productivity differentials and latitude

28