our not-so-urban world

OurNot-So-UrbanWorldShlomoAngel,PatrickLamson-Hall,BibianaGuerra,YangLiu,NicolásGalarza,and

AlejandroM.Blei

WorkingPaperNo.42

TheMarronInstituteofUrbanManagement,NewYorkUniversity

August2018(Updated22August2018)

OurNot-So-UrbanWorld

2

Abstract

Arecentarticletitled“Everythingwe’veheardaboutglobalurbanizationturnsouttobewrong”(ThomsonReutersFoundation,10July2018),citingnewfiguresreleasedbytheEuropeanCommission, informedreaders that84%of theworld’spopulationnow lives inurban areas, not 55% as reported by the United Nations Population Division. In thisworkingpaperwepresentfourarguments,backedbyampleevidence,contendingthattheEuropeanCommission’snumberisimplausibleiftheword‘urban’istoretainanyfamiliarmeaningatall.First,theshareofthegloballaborforceemployedinagriculturewas27%in2015.Thisshare,coupledwithanestimated30%incrementtotheagriculturallaborforceinvillagenon-farmemploymentandanestimated15%largerhouseholdsizeinruralareas,,suggestthatnotmorethan60%oftheworld’spopulationlivedincitiesinthatyear.Second,regularitiesinthedistributionofthecitypopulationsizes,knownasZipf’sLaw,allowustoestimatethepopulationofcitiesof5,000peopleormorein2010as3.6billion.52%oftheworld’spopulationlivedinsuchcitiesandtownsthatyearandnot84%asclaimedbytheEuropeanCommission.Third, the low ‘urbandensity threshold’ adoptedby theEuropeanCommissionresultsintheinclusionofentirecroplandregionsasurban:InJava,Indonesia,forexample,96%ofthepopulationlivingoncroplandisclassifiedasurban.Andfourth,thelowurbandensitythresholdadoptedbytheEuropeanCommissionistoolowincomparisonwith observed population densities on the fringe of real-world cities. It implies that aresidential plot on the urban edge would have an area equivalent to 1.5 regulation-sizesoccer fields. The European Commission claims that in 2015 cities occupied 2.27millionkm2andcovered1.5%ofthelandmassofourplanet,morethan2.5timesourownback-of-theenvelopeestimate.IfoneintendstobelievetheEuropeanCommission’sestimates,thenone can safely conclude that cities should stop expanding right nowand that allmust bedonetocontainthem. If theworld isalready84%urbanthenonemayalsoconcludethattheurbanizationproject—therelentlessmigrationofpeoplefromvillagetocity,fromlivingclosertothelandtolivingclosertoeachother—isbasicallyover.We,onourpart,believethatit isbynomeansoverandthatwestillhaveawindowofopportunitytoprepareourcitiesforabsorbingmorethan2billionpeopleby2050,manystillresidinginruralareas.Inaddition to retrofitting existing cities—which, if we are to believe the EuropeanCommission, is all that is left for us to do—we believe that preparing cities for theirinevitable andmassive expansion anddensification in the decades to come is a very realchallengenowfacingusall.


3


Introduction

In2003,theUNSecretaryGeneral,relyingonextensiveUNHabitatanalysis,declaredthat“[a]lmost1billionpeople, or32per cent of theworld’surbanpopulation, live in slums…And if no serious action is taken, thenumberof slumdwellersworldwide is projected toriseoverthenext30yearstoabout2billion.“(TheChallengeofSlums,2003,page1)The1billion number was repeated and amplified uncritically by hundreds of media outlets aswell as by numerous academic publications and awidely circulated book titledPlanetofSlums. UNHabitat classifiedmore than one-third of EastAsia’s urbanpopulation as slumdwellers, for example, a highly disputable finding that no observant visitor to Chinese,KoreanorJapanesecitiescouldpossiblysupport.Since2003,UNHabitathasindeedreviseddownthat1billionnumber.Accordingtoits2016WorldCitiesReport,“theabsolutenumberstoodat881millionin2014,”i.e.atalowervaluethanthe1billionestimatedtohavelivedin slums in 2003. It is still too high, unfortunately, because it is census-based and notsettlement-based and the census metrics used—e.g. the share of the urban populationwithout improved water and sewer connections—do not correspond to our intuitiveunderstandingof‘slums’.

Amajormediaoutlet informed itsreadersrecentlyus that “Everythingwe’veheardaboutglobalurbanizationturnsouttobewrong”.1Thearticlebythatnamecitesnewestimates of the share of the world’s population living in urban areas released by theEuropean Commission’s Directorate General of Regional andUrban Policy. The EuropeanCommission disputes the 2018 UN Population Division numbers, according to which 4.2billionpeople,or55%oftheworldpopulation,livedinurbanareasin2018.ItsAtlasoftheHuman Planet-20162—using a new methodology and backed by satellite imagery and aglobalpopulationgrid—reports that asmanyas6.4billionpeople,or84%of theworld’spopulation, lived inurbanareas in2018. Yet again,we findourselveswith an alarminglylargenumberthatislikelytoberepeatedadnauseumbymediaoutletslookingforscoopsor by interested parties anxious to make use of this number to further their politicalagendas. Before leaping on the bandwagon,wemust ask ourselves ifwhat the European

1Scruggs,Gregory.ThomsonReutersFoundation.10July2018,onlineat:

http://www.thisisplace.org/i/?id=0150beca-e3f5-47e0-bc74-9ccc5ef1db8a 2 Pesaresi et al, eds., 2016. Atlas of the Human Planet—2016, Joint Research Centre, European

Commission,onlineat:https://ghsl.jrc.ec.europa.eu/documents/Atlas_2016.pdf?t=1476360675.


4

Commission considers urban can possibly correspond to our intuitive understanding ofurban,andifthisnumbercanpossiblybetrue.

TheproblemwiththeUNurbanpopulationestimates

The UN global urban population estimates, now released every four years by the UNPopulation Division, have come under criticism before, essentially on methodologicalgrounds.TheUN—unliketheWorldBankortheIMF—isboundbyitsmandatetoonlyusenumbers provided by its member states. And while the UN has made valiant efforts tostandardize census questionnaires and methods over the years, it has been unable tostandardize two key quantitative measures: First, what levels of spatial contiguity andpopulationdensityconstituteawell-defined,singular‘settlement’,asagainstacollectionofdisparate settlementsor, for thatmatter, individual farms; and second,what is theurbanpopulation threshold (UPT) that distinguishes an ‘urban’ from a ‘non-urban’ or ‘rural’settlement.ThusnoprofessionaldemographerseriouslyquestionstheUNestimatesofthetotalpopulationsofcountries,beyondquestioningthecapacityofnationalcensusbureausto correctly count their populations.Butmanydoquestion its published estimates of theshareofthepopulationlivinginurbanareas.

As each member state still employs its own classification of what constitutes asettlementaswellasitsownurbanpopulationthreshold,theshareofitspopulationthatitreportstotheUNasurbanisnotreallycomparabletowhatothermemberstatesreportasurban. TheUNPopulationDivision,which is chargedwith aggregatingnational estimatesinto global ones in its bi-annual World Urbanization Prospects, is thus faced with animpossible task: Adding populations that were defined in different ways into singularregionalandglobalestimatesthatarelabeled‘urban’,eventhoughthedefinitionof‘urban’varies from one state to another. That should not mean, however, that the UN urbanpopulationestimatesforworldregionsandfortheworldasawholeareoffbyasmuchas50 percent, as the European Commission claims. It may well be that the various urbanpopulation thresholds in individual country estimates cancel each other, and we havegatheredevidencethatindicatesthattheUNestimatesarenotthatfaroffthemark.

Thechallengeofdefining‘urban’settlements

From a theoretical perspective, any urban-rural distinction that includes a contiguity- ordensity-based definition of a settlement and an urban population threshold comes undersuspicion.ForBrennerandSchmidt,

it is clear that settlement-based understandings of the urban condition have nowbecomeobsolete.Theurbancannotbeplausiblyunderstoodasabounded,enclosedsiteofsocialrelationsthatistobecontrastedwithnon-urbanzonesorconditions.It


5

is time, therefore, to explode our inherited assumptions regarding themorphologies,territorializationsandsociospatialdynamicsoftheurbancondition…Today,urbanizationisaprocessthataffectsthewholeterritoryoftheworldandnotonlyisolatedpartsofit…Thereis,inshort,nolongeranyoutsidetotheurbanworld;thenon-urbanhasbeenlargelyinternalizedwithinanunevenyetplanetaryprocessofurbanization.3

Fromourownperspective,BrennerandSchmidtaregoingtoofar.Thereisstillvalueindefining settlementsaswell as cities in spatial terms.Anyonewhohasexaminedsatelliteimagerycanclearlyidentifyfreestandingsettlements,madeupofcontiguousbuilt-upareaswith urbanized open spaces in and around them. One can also clearly distinguish largesettlementswithcontiguousurban footprintsextendingover largeareas, fromout-of-the-way villages with a small number of structures in a sea of cultivated land. Moreover,examining large regions, one can clearly identify a hierarchy of settlements,with a smallnumberoflargecitiesspacedfurtherapart,alargernumberofmedium-sizedtownsspacedclosertogether,andalargenumberofsmallvillagesdistributedwidelythroughouttheland.Weseenoreasontoignoretheseclearlyobservablesettlementsandsettlementhierarchiessimplybecausetheworldhasbecomemoreurbanincharacter.

Thatsaid,wearefullyawarethatmanycitiesarelocatedwithinagriculturalzonesandthat, as they expand outwards, they incorporate villages and towns on their peripheries.Furthermore, freestanding villages and towns on the peripheries of cities becomeincreasinglyurbanastheircontacts,connections,andexchangeswiththecitybecomemoreintense,astheyproducemoregoodsandservicesforthecity,asmoreoftheirinhabitantswork in thecity, andasmoreurbanitesmake thesevillages theirhomesandcommute towork in thecity.Thephysicalboundarybetweenthe ‘urban’and the ‘rural’—whileusefulforpurposesofanalysis,planningandpolicymaking—canthusonlybeanapproximation,aforceddistinctionbetweenareasthataremoreurbanandareasthataremoreruralintheircharacterandfunction.

Fromourperspective,acontiguousbuilt-upareathatcontains100,000peopleormorecomprisesanurbansettlement.Startingwithourearlierworkin20054,wechose100,000peopleasoururbanpopulationthresholdnotbecauseweconsideredsmallersettlementsas

3Brenner,N.andSchmid,C.,2013.The‘UrbanAge’inQuestion,InternationalJournalofUrbanand

RegionalResearch,20-21,onlineat:http://www.urbantheorylab.net/site/assets/files/1014/brenner_and_schmid_ijurr.pdf

4 Angeletal,2005.TheDynamicsofGlobalUrbanExpansion,TransportandUrbanDevelopmentDepartment,WashingtonDC:TheWorldBank,onlineat:http://siteresources.worldbank.org/INTURBANDEVELOPMENT/Resources/dynamics_urban_expansion.pdf.


6

non-urbanbutbecausewewantedtoensureafocusoncities.Inalmostallcountries,exceptChina,asettlementof100,000peopleormoreisconsideredacity.In2010,forexample,weidentifiedauniverseof4,231citiesandmetropolitanareas thatwerecontiguousbuilt-upareas with populations of 100,000 people or more (see figure 1). We then focused ouranalysisonastratifiedsampleof200citiesfromthisuniverse(seefigure2).

Figure1:Theuniverseofallcitiesandmetropolitanareasthathas100,000peopleormorein2010.

Figure 2: The global sample of 200 cities from the universe of cities, stratified by 8 worldregions,4populationsizeranges,andthreecategoriesofthenumberofcitiesinthecountry.

WedefinedandmappedthephysicallimitsofthesecitiesbyclassifyingLandsatsatelliteimagery—imagerywith a30-meterpixel resolution—intobuilt-upandopen spacepixels,and then classifying the built-up pixels into urban andnon-urban ones.Wedrew a 1km2‘walkingdistancecircle’aroundeachbuilt-uppixel,andclassifiedallpixelsthathad25%or


7

moreoftheircirclebuilt-upas ‘urban’andallpixelsthathadlessthan25%oftheircirclebuilt-upas ‘non-urban’.Thatallowedus todrawadistinctboundaryaroundeverycity inthesampleattheedgeofitsurbanbuilt-uparea,definingthecitybyitsExtremaTectorum,as the ancient Romans did. Clearly, the same method can be used to map smallersettlements if a lower urban population threshold is preferred, as well as to map ruralsettlements, be they townsor villages, so as to distinguish them from freestanding farmsandhamlets.

Itisimportanttonoteherethatourdefinitionofasettlement,largeandsmall,doesnottakepopulationdensityintoaccount.Itonlyconsidersthecontiguityofbuilt-upareas.Wedefineurbanizationasthemovementofpeoplefrombeingclosertothelandtobeingclosertoeachother.Thecontiguityofbuilt-upareasisthusaproxyforthebasicfunctionofcities,thecreationofvaluebybringingpeopletogethertoworkwithoneanother, to learnfromoneanother,toexchangegoodswithoneanother,andtoserveoneanother.Moregenerally,people living invillagesand tilling the landstillneed tobe incloseproximity to the land.Peoplelivinginurbanareastypicallyengageinnon-farmoccupationsandnolongerneedtobe close to the land.Theymake a livingbybeing closer to otherpeople andby engagingwithotherpeople,largelyabandoningtheircontactwiththeland.Clearly,therefore,wecanobserve that in a hierarchy of settlements, the people living in smaller settlements areclosertothelandandworktheland,whilethepeoplelivinginthelargestsettlementsarefarfromthelandanddonotworktheland.Finally,thepropensityofpeopletoliveclosetoeach other in urban settlements does imply a certainminimum density of settlement: Aresidentialareaontheurbanedgewhereeachhouseoccupiesaone-hectareplot—roughlythesize1.5regulation-sizesoccerfieldsortwoNFL-approvedfootballfields5,forexample—shouldnotbeconsidered ‘urban’simplybecausepeopleinmosturbansettlementsdonotlivethatfarfromeachother.

Given these preliminary observations, we present four arguments that reject thepropositionthat84%oftheworld’spopulationnowlivesinurbanareas,asclaimedbytheEuropeanCommission:

1. Theshareofthelaborforceinagricultureandvillagenon-farmoccupations:27%ofthegloballaborforcewasemployedinagriculturein2015andanadditional30%ofvillageemploymentwas innon-farmoccupations.Aswewilldemonstrate,theseestimates—togetherwithdifferencesinhouseholdsize—suggestthatnomorethan60%oftheworld’spopulationlivedincitiesinthatyear;

2. Regularitiesinsettlementhierarchies:Regularitiesinthedistributionofthecity

5 TheminimumdimensionsofaFIFAapprovedsoccerfieldforinternationalmatchesare100-by-

64meters;theNFL-approveddimensionsofafootballfield(includingtheendzones)are360-by-160feetor109.7-by-48.8meters.


8

populationsizes,knownasZipf’sLaw,allowustoestimatethepopulationofcitiesof5,000peopleormorein2010as3.6billion.52%oftheworld’spopulationlivedinsuchcitiesthatyearandnot84%asclaimedbytheEuropeanCommission;

3. High density agricultural regions: We will show that the low urban densitythreshold—300personsperkm2—adoptedbytheEuropeanCommissionresultsintheinclusionofalmosttheentirepopulationofcroplandregionsas‘urban’;and

4. Typicaldensitieson theurban fringe:Wewillshowthatthelow‘urbandensitythreshold’ adopted by the European Commission is too low in comparison withobservedpopulationdensitiesonthefringeofreal-worldcities.

1.Theshareofthelaborforceinandvillagenon-farmoccupations

Ifoneacceptsthepremisethatthosewhoworkinagriculturearenoturbanitesandthatthelargemajorityofthosewhoareemployedinagriculturedonotliveinurbanareas,wegainthefirstsuspicionthattheEuropeanCommission’sestimatesmaynotberight.AccordingtotheInternationalLaborOrganization(ILO),27%ofallworkers intheworldat largewereemployedinagriculturein2015.

Inadditiontothoseemployedinagriculture,manyinhabitantsofruralvillagesengagedin non-farm occupations: Small-scalemanufacturing, construction, retail sales, education,publicduties,religiousdutiesandothermiscellaneousserviceoccupations.Studiesofnon-farmemployment in rural areas suggest that, as a rough estimate, non-farmemploymentnowaddsasmuchas30%tothegloballaborforceinruralareas.6Thatwouldsuggestthatthe shareof theglobal labor force inagricultureandvillagenon-farmoccupationswasoftheorderof35%in2015(0.27×1.3=0.35).

Wealsoknowthatruralhouseholdsare largerthanurbanones:In2015, forexample,rural households in 80 countries comprising 70% of the world’s population in that yearwere 15% larger, on average, than urban households.7Assuming that labor participationrateswere the same in rural and urban areas, thatwouldmean that 40% of theworld’s

6 Seeforexample,table1.2inHaggblade,S.,Hazelle,P.B.R.,andReardon,T.,2007.“Introduction”,

inHaggblade,S.,Hazelle,P.B.R.,andReardon,T.,eds.,TransformingtheRuralNon-FarmEconomy:OpportunitiesandThreatsintheDevelopingWorld, InternationalFoodPolicyResearchInstitute,Baltimore: Johns Hopkins University Press, 6; table A3 in Majid, N., 2015. “The GreatEmploymentTransformationinChina”,WorkingPaperNo.195,EmploymentPolicyDepartment,Geneva:InternationalLaborOrganization,43;andtable6inReddy,D.N.,Reddy,A.A.,Nagaraj,N.and Bantilan,M.C.S, 2014. “Emerging Trends in Rural Employment Structure and Rural LaborMarkets in India”,Working Paper Series No. 56, International Crop Research Institute for theSemi-AridTropics(ICRISAT).

7 Source:https://globaldatalab.org/areadata/2015/hhsize+regpopm/.


9

population lived in ruralareas in2015 (0.35×1.15=0.40).Therefore, atmost60%of theworld’spopulationlivedinurbanareasinthatyear,amuchsmallernumberthanthe84%postulatedby theEuropeanCommissionandmuchcloser to theUNPopulationDivision’s54%estimateoftheshareoftheworldpopulationthatwasurbanin2015.

2. Regularitiesinsettlementhierarchies

The European Commission has settled on 5,000 people as a common urban populationthreshold(UPT) forallcountriesandregions:Settlementsof5,000peopleormorewherecontiguous 1km2 grid cells have populations densities of at least 300 persons/km2 aredefinedasurbanwhile settlementswith fewer than5,000people aredefinedas rural, ornon-urban.Wehavenoquarrelwiththe5,000peoplethreshold—fewruralvillageswithanagriculture-basedeconomycontainmorethan5,000people—andwedoappreciatethatitovercomes the limitations in the UN Population Division’sWorld Urbanization Prospectsreports that require it to acceptmember states’ urban share estimates that use differentthresholds.Letusassume,intheabsenceofinformationtothecontrary,thatwecanindeedusethesameurbanpopulationthresholdforallcountries.Wecanaskourselvesthen:Howmanypeople in theworld live in settlementsof5,000peopleormore?Wedonothaveacomplete listofsuchsettlements,but forthepurposesof thisreview,wehavecompiledadataset that includes all cities that had 20,000 people ormore in 2010. The datasetwascompiled from www.citypopulation.de, a reliable and comprehensive website built andoperatedbyThomasBrinkhoffthatactsasaclearinghouseforcensusinformationoncities.Using the census populations and dates for individual cities, we interpolated orextrapolated city populations to 2010, obtaining a list of 19,289 cities that had 20,000peopleormorein2010,withatotalpopulationof3.01billionpeople,some84%oftheUNestimateoftheurbanpopulationintheworldinthatyear:3.59billion.

It iswell known thatwhen settlements are ranked by their population size, from thelargest to the smallest, their population can be readily estimated from their rank. Thedistributionofcitypopulationsizesfollowsapowerlaw,originallydiscoveredbyZipfandoftenreferredtoasZipf’sLaw.8Thisregularityimpliesthatifwedividecitiesintoranges—wherethelowerlimitofeveryrangeishalfthesizeof itsupperlimit, thenthenumberofcitiesinarangewillbetwicethenumberofcitiesinthenextrangewhiletheaveragecitypopulationinarangewillbehalf thatofthenextrange.Wedividedthe19,289citiesthathad20,000peopleormorein2010intotenranges:(1)20-40K;(2)40-80K;(3)80-160K;…(8) 5.12M; (9) 5.12-10.24M and (10) 10.24M+.We found that the power law holds—thenumber of cities in a given range is halved and the average population is doubled aswemovefromonerangetothenextoneup.Thisisillustratedinfigures3and4below.

8 SeeZipf,G.K,1949.Humanbehaviorandtheprincipleofleasteffort.Addison-Wesley.


10

Figure 3: The number of cities in each city population size range in 2010: Blue barscorrespondtoempiricalfindings;orangebarsareestimatedfromthetrendline.

Figure4:The average citypopulation in each citypopulation size range in2010:Bluebarscorrespondtoempiricalfindings;orangebarsareestimatedfromthetrendline.

Using the trend lines in each graph, we could estimate the number of cities and theaverage city population size in the two lower ranges—5-10K and 10-20K—with a highdegreeofaccuracy,asthetrendlinesbothhaveR2≥0.99.Multiplyingthenumberofcitiesineachrangebyitsaveragecitypopulationsize,weestimatethetotalpopulationineachofthesetworangestobe294Mand296Mrespectively.Thisgivesusanestimateofthetotalnumber of people that lived in urban settlements of 5,000 people ormore in 2010: 3.60billion,anumberthat iswithin0.2%oftheUNestimateof theglobalurbanpopulation inthat year (3.59 billion) andmuch smaller than global urban population estimated by theEuropeanCommission,whichin2010wasoftheorderof5.8billion.Usinganestablished

41,51920,992

8,1984,715

2,9821,845

847391

181

7739

14

y=89194e-0.701xR²=0.99358

10

100

1,000

10,000

100,000

5-10K

10-20K

20-40K

40-80K

80-160K

160-320K

320-640K

640K-1.28M

1.28-2.56M

2.56-5.12M

5.12-10.24M

10.24M+

Num

bero

fCitiesinRange

CityPopulationRanges

714

2856

112224

444877

1,7363,352

7,19514,165y=3.5473e0.6896x

R²=0.99995

1

10

100

1,000

10,000

100,000

5-10K

10-20K

20-40K

40-80K

80-160K

160-320K

320-640K

640K-1.28M

1.28-2.56M

2.56-5.12M

5.12-10.24M

10.24M+

AverageCityPop

ulationinRange('000)

CityPopulationRanges


11

regularityinthepopulationdistributionofsettlements,wefindtheEuropeanCommission’snumbertobewidelyoffthemark.

2. Highdensityagriculturalregions:

According to the European Commission’s Atlas of the Human Planet—2016, 9 ‘UrbanClusters’ are contiguous clusters of 1km2 cells, with each cell containing 300 or morepeople, and the total cluster containing at least 5,000 inhabitants. The threshold for anarea—say,a1km2ofland—tobeclassifiedasurbanisthusapopulationdensitythreshold:300personspersquarekilometer.Thisthresholdisappliedtoallcountriesandregionsintheworld regardless of the great variations in both urban and rural populationdensitiesamongthem.Thechoiceof this ‘urbandensity threshold’ isproblematicbecause, inmanyheavily populated agricultural regions, densities—typically more evenly distributed overthe land andwithout the peaks often observed in large cities—aremuchhigher than the300personsperkm2‘urbanthresholddensity’adoptedbytheEuropeanUnion.

To illustrate this, we take the island of Java in Indonesia as an example. Java is verydensely populated. The total population of the island in 2015was 144.9million. Its landareais138,332km2.Itsaveragepopulationdensitywas,therefore,1,048personsperkm2.Approximately 42% of the area of Java, 58,443km2is cultivated cropland.10Average farmsizeinCentralJavain2010was0.40hectares;inEastJavaitwas0.47hectaresinthatyear.LetusassumeanaveragefarmsizeinJavain2010wasoftheorderof0.5hectares.11Thisimplies that therewere some11.8million family farms in Java in 2010.With an averagehouseholdsizeof4.012andassumingthat20%ofthe laborforceisemployedinnon-farmoccupations, agricultural villages, hamlet, and individual farms housed some 56 millionpeople or 40% of the population of the island in that year. This estimate is reasonable:

9ExecutiveSummaryonlineat:

https://ghsl.jrc.ec.europa.eu/documents/Atlas_of_the_Human_Planet_2016_executivesummary.pdf?t=1476109824.

10 Fuglieestimatedittobe70,000km2in2001-2005.SeeFuglie,K.O.,“Chapter12:Indonesia:FromFoodSecuritytoMarket-LedAgriculturalGrowth”,table12.2,page347.Onlineat:https://www.card.iastate.edu/products/books/shifting_patterns/pdfs/chapter12.pdf.

11Yamauchi,F.,2014.“WageGrowth,Landholding,andMechanizationinAgriculture:EvidencefromIndonesia”,PolicyResearchWorkingPaper6789,WashingtonDC:TheWorldBank,February,table1,page17,onlineat:http://documents.worldbank.org/curated/en/584441468042884117/pdf/WPS6789.pdf.

12TheaveragehouseholdsizeinIndonesiain2012was4.0.UN,HouseholdSizeandCompositionAroundtheWorld2017,18http://www.un.org/en/development/desa/population/publications/pdf/ageing/household_size_and_composition_around_the_world_2017_data_booklet.pdf


12

According to the ILO, 39.1% of total employment in Indonesiawas in agriculture in thatyear.

According to U.S. Geological Survey data, Java had 58,443km2 of cropland in 2015.According to the European Commission’sAtlas of theHumanPlanet—2016 database, thetotalpopulationlivingonthiscroplandareaamountedto58.8million.Thisisnotfarfromourestimatebasedonfarmsize.Thispopulationinhabitedthecroplandareaoftheislandatanaveragedensityof1,006personsperkm2.Becauseofthishighdensityofsettlement,theEuropeanCommissionclassified56.3millionpeople—96%ofthetotalcroplandpopulationin Java—as ‘urban’. In terms of area, it classified half of the cropland area of Java,27,884km2,as ‘urban’.Thisisillustratedinfigure5below.Thefigureshowsfourdifferentzonesin2015:(1)The‘urban’areaofJavanotoncropland,atotalof23,352km2,whichishometo83.5millionpeople,showninblack;(2)the ‘urban’areaoncropland,27,884km2,whichishometo56.3millionpeople,showninred;(3)the‘rural’croplandarea,atotalof30,559km2,whichishometo2.5millionpeople,showninlightgreen;and(4)uncultivatedlandthatisnot‘urban’,atotalof56,637km2,shownindarkgreen.

Figure 5: Amap of the island of Java, Indonesia, derived from European Commission data,showing that in2015more thanone-thirdof itsarea (51,232km2)wasclassifiedas ‘urban’,halfofituncultivatedlandandhalfofitoncultivatedland.

Inotherwords,theEuropeanCommissionclassifiedhalfofthecroplandareaofJavaas‘urban’and96%ofthepopulationlivinginthisareaas‘urban’,aclassificationthatdoesnotcorrespond to any possible understanding of what constitutes an ‘urban’ settlement.Wehave not conducted similar analyses for dense agricultural regions in other countries—China,India,Bangladesh,Vietnam,SouthKorea,Egypt,theDemocraticRepublicofCongo,orEthiopia, to takea fewexamplesthatreadilycometomind—butwestronglysuspect thatthe results will be the same. The low urban density threshold adopted by the Europeancommissionleadstotheclassificationoflargecroplandregions,andthegreatbulkoftheir


13

populations,as‘urban’,resultinginanover-estimateoftheurbanpopulationsofcountries,regions,andtheworldatlarge.

4.Typicaldensitiesontheurbanfringe

The low urban density threshold—300 persons per km2—presents yet another difficultywhenappliedtodelimitingtheedgeoflargeurbanareas,saycitiesandmetropolitanareasthathad100,000peopleormore in2015.Weknowfromahostof studiesundertaken inmany cities since Colin Clark’s 1951 seminal article, “Urban Population Densities”,13thatdensitiesarehigherinthecentralcityanddeclineregularlyaswemoveawayfromthecitycenter to the inner suburbs and eventually to the outlying suburbs. According to theEuropeanCommission, thedensityattheurbanedgewouldbeas lowas300personsperkm2.Densitiesclosertothecitycenterwillbehigher.Areasadjacenttotheurbanedgebutfurtherawayfromthecitycenterwillhavedensitiesthatlowerthan300personsperkm2andwillthereforenotbeconsideredurban.Doesthisurbandensitythresholdcorrespondinanywaytowhatweknowaboutreal-worldcities?

As noted earlier, we defined and mapped the physical limits of cities by classifyingLandsat satellite imagery—imagerywith a 30-meter pixel resolution—into built-up14andopenpixels,andthenclassifiedthebuilt-uppixelsintourbanandnon-urbanones.Wedrewa1km2‘walkingdistancecircle’aroundeachbuilt-uppixel,andclassifiedallpixelsthathad25%ormoreoftheircirclebuilt-upas‘urban’andallpixelsthathadlessthan25%oftheircirclebuilt-upas‘non-urban’.Wethenaddeda100-meter-wideopenspacebufferalongthebuilt-up urban edge of cities to approximate the open space affected by the city. Thatprocedureallowedustodrawadistinctboundaryaroundeverycitywestudiedandtomapits ‘urban footprint’. Essentially, the urban edge of cities is thus determined by thecontiguityofbuilt-uppixels:Onlybuilt-uppixels thathavea lotofotherbuilt-uppixels intheir immediate vicinity (and that together accommodate 100,000 people or more) areconsidered‘urban’.

Incontrast,theEuropeanCommissiondelimitstheboundarybetweenurbanandnon-urbanareas using a density threshold instead of a contiguity threshold. We suspect that thisdensitythresholdistoolowinallbuttheplushest2-to-5-acre-lotsuburbsinNorthAmerica(0.8–2.0hectares).Thinkofa1km2residentialareaontheouterfringeofalargecity.Letusassumethat25%ofthisareaistakenupbystreets(agenerousassumptionconsideringthatwecalculatedtheaverageshareoflandinstreetsinthenewlybuilt-upareasofcities—

13 Clark, C., 1951. “Urban Population Densities”, Journal of the Royal Statistical Society Series A

(general),4,490-496.14Built-uppixelshavemorethanhalfoftheirareacoveredbyimpervioussurfaces,suchasroofsor

pavements.


14

areasdevelopedbetween2000and2014—tobe20.2±0.7%15).Thatwould leave0.75km2for residential plots. At the urban edge there will be 300 people inhabiting this area.Assumingthat they form4-personhouseholds, therewillbe75households inhabitingthearea.Thatwouldmeanthattheplotareaoccupiedbyeachhouseholdwouldbeonehectare(10,000m2 or 2.5 acres) of land, on average, roughly equivalent to 1.5 the area of a FIFAregulationsoccerfieldortwicetheareaofanNFL-approvedfootballfield.16

This ishighlyunrealistic in lightof the fact that themediannewresidential lotsize intheUnitedStates—lotstypicallybuiltattheouterfringesofcities—droppedfromin930m2in1992to800m2in2015.17ThisisanorderofmagnitudelowerthanthenewresidentiallotsizeontheurbanedgethatwouldbeimpliedbytheEuropeanCommission’surbandensitythreshold.Needlesstosay,averageplotsizesintheexpansionareasinourglobalsampleof200 cities—areas developed between2000 and2014—are smaller18: The average size ofplotsininformallandsubdivisionswasfoundtobe472±98m2,andtheaveragesizeofplotsinformallandsubdivisionswasfoundtobe537±72m2.Evenifweassumethathouseholdsontheurbanfringeoccupyplotsof1,000m2andthat,onaverage,thesehouseholdsare4-personhouseholds,asquarekilometeroflandontheurbanfringeofcitieswillhaveroomfor 3,000 persons—10 times the ‘urban density threshold’ adopted by the EuropeanCommission.Wemustconclude,therefore,thattheurbandensitythresholdadoptedbytheEuropean Commission is simply too low, and that this low threshold leads to countinghouseholdsthatarenoturban—eitherbecausetheymakealivinginagricultureorbecausethey live at non-urban densities; and that including these households results in massiveovercountingoftheurbanpopulationoftheworld.

Figure 7 below presents an example of such possible over-counting. It compares theurbanclusterswithinthestudyareaofDhaka,Bangladesh(yellowline)intheAtlasofUrbanExpansion—2016EditionwiththeurbanclustersintheEuropeanCommission’sAtlasoftheHumanPlanet—2016.Bothmapsarefortheyear2015.Theareaintheformer(darkred)is444km2 and the area in the latter (red+dark red) is 1,718 km2, i.e. four times as large.Within the entire image shown in figure 7, the area classified as urban in the latter,4,975km2,is11timestheareaclassifiedasurbanintheformer.

15 Angeletal,2016.AtlasofUrbanExpansion—2016Edition,Volume2:BlocksandRoads,table1-2,

onlineat:http://www.atlasofurbanexpansion.org/data.16 Seefootnote5.17 Siniavskaia,Natalia,2016.“Lotsin2015areSmallestonRecord”,EyesonHousing,11July,online

at:http://eyeonhousing.org/2016/07/lots-in-2015-are-smallest-on-record/18 Angeletal.,2016,op.cit.,table1-2.


15

Figure 7: A comparison of urban clusterswithin theDhaka, Bangladesh study area (yellowline) between the Atlas of Urban Expansion—2016 (dark red) and the Atlas of the HumanPlanet—2016(red+darkred)in2015.Theareaofthelatteris4timestheareaoftheformer.

Not surprisingly, the low ‘urban density threshold’ adopted by the EuropeanCommissionalsoresultsinverylargeestimatesofthetotalamountoflandinurbanuseintheworld at large. ItsAtlasof theHumanPlanet—2016 asserts that in 2015 urban areasoccupied 2.27million km2, up from1.86million km2 in 2000 and (according to our owninterpolation)2.12millionkm2in2010.Anurbanlandcoverof2.27millionkm2amountsto1.5%ofthelandmassofourplanet.19

TheEuropeanCommissionestimatesareseveraltimeslargerthanotherestimates.Themedianvalueoftheglobalurbanlandcoverin2000intheeightmapsexaminedbyPotereetal20,forexample,wasonly552,000km2.TheEuropeanCommission’sestimatefor2000is

19 The landmass of the planet is 149 million km2, and 2.27÷149=1.5%. The Atlas of the Human

Planet-2016calculated thispercentage incorrectlyas7.6%.Thiserrorhasbeenaddressedandwillbecorrectedintheforthcoming2018editionoftheAtlas.

20 See,forexample,Potereetal.,“Mappingurbanareasonaglobalscale:Whichoftheeightmapsnowavailableismoreaccurate?InternationalJournalofRemoteSensing30(24),6531-6558.


16

morethantriplethatnumber.Angeletal21estimatedgloballandcoverin2000tobeoftheorderof600,000km2andprojectedittoincreaseto780,000±90,000km2by2010.Ourmorerecentback-of-the-envelopeestimatefor2010isstillwithinthisrange,andourback-of-the-envelope estimate for 2015 is of the order of 840,000km2 (less than 0.6%of theworld’stotal land mass). Again, the European Commission’s estimates for these years are muchlarger,indeedmorethan2.5timeslargerthanourownestimates.Whilewearefullyawarethatmoreworkisneededtoprovideaccurateassessmentsoftheamountoflandinurbanuse,we strongly suspect that the low ‘urbandensity threshold’ adoptedby theEuropeancommission results in a gross overestimation of the amount of land in urban use, incountries,inregions,andintheworldatlarge.

Conclusion

WebegtodifferwiththeEuropeanCommission’sconclusionthat84%ofthepopulationofthe world now lives in urban areas. We do understand why the European Commissionquestions theUNurbanpopulationestimates.Weare fullyawareof the limitationsof theestimatesoftheshareofcountrypopulationslivinginurbanareasprovidedbytheUnitedNationsWorldUrbanizationProspects.TheselimitationsareinherentintheUN’smandate,restrictingittotheuseofnumbersprovidebymemberstates.Thatsaid,itneednotmeanthattheglobalestimatesprovidedbytheUNPopulationDivisionarewayoffthemark.Webelievethatitispossibletorecalculatethesesharesusingacommonyardstick—essentiallyacommon‘urbanpopulationthreshold’,coupledwithacleardefinitionofwhatconstitutesa‘settlement’—andweourselvesareplanningtoundertakethiseffortinduetime.Indeed,takingtheUNdataandmodelingittoestimatecountryurbanpopulationsusingacommonurbanpopulationthresholdpresentsasensibleyetnotinsurmountablechallengetourbandemographers. Itwill producedifferentnumbers and it is difficult to tell in advancehowdifferent theymaybe.What isrequirednowisclarity:Aclearexpositionofeachmemberstate’s method of calculating its urban population. Hopefully, researchers interested inmakinginternationalcomparisonsandarrivingatnewglobalestimatesoftheworld’surbanpopulationcouldthendevelopandtestamethodfortransformingthedisparatethresholdsusedbydifferentcountriesintoacommonurbanpopulationthresholdandthencalculatingnewestimatesbasedonthiscommonthreshold.

WehavenoquarrelwiththeEuropeanCommission’schoiceofapopulationof5,000asafirstattempttodefinean‘urbanpopulationthreshold’.Thisthresholdcanandshouldbe

21 Angel, S., Parent, J., Civco,D. L., Blei, A. andPotere,D., 2011. “The dimensions of global urban

expansion: Estimates and projections for all countries, 2000–2050”, Progress in Planning, 53–107,table6.2,98.


17

tested alongwith others, before deciding on themost sensible one. Are there villages orsmalltownswithmorethan5,000personswherethemajorityofworkersareemployedinagriculture?Wehavenodatathatcanshedlightonthisquestion.Doesthepopulationsizeof settlements where the majority of people are employed in non-farm jobs varysystematically fromone country to another?Wehave no data that can shed light on thisquestioneither.Unfortunately,theEuropeanCommissiondoesnottelluswhyitchosethisthreshold.Webelievethatitneedstobegroundedinasolidtheoreticalfoundationofwhatmakesasettlementurban,especiallyinlightofrecentclaimsthatallsettlementslargeandsmallarenowurban.

WebelievethattheEuropeanCommission’sestimateoftheshareofthepopulationthatresidesinurbansettlementsisinaccurate,andthatthesourceoftheerrorisnotthe5,000persons thresholdbut thedensity threshold that isused to identify a settlementof5,000persons,namelytheurbanthresholddensityof300personsperkm2.Thisdensitythresholdistoolow,resultingin identifyingmanyareasthatareclearlynoturbanasurbanwithoutgiving a solid justification as to why such areas are urban. In other words, agriculturalregions with small plots tended by farmers living in villages and hamlets are lumpedtogetherwithlargemetropolitanareasandlabeledurban.Asaresult,afartoolargeshareof theworld’spopulationanda far too largeshareof theglobal landmass is classifiedasurban.

Thismatters.Why? Because if one is led to believe that in 2015 cities occupied 2.27millionkm2andalreadycovered1.5%ofthelandmassofourplanetthenonemayconcludethat this number is excessively large and that cities should be prevented, by all meanspossible, from ‘sprawling’ any further. If that is true,we shouldnowrestrict ourselves toaccommodatingtheshiftingurbanpopulationsthroughinfillanddensification,butcertainlynot through urban expansion, expansion thatwill further increase the already exorbitanturban land cover on our planet. This is the position adopted by many planners,environmentalists, and city officials that are appalled by urban ‘sprawl’ and have sought,unsuccessfully, to contain it for the past thirty years. Where containment has beensuccessful—say, in Seoul or San Francisco—it has usually resulted in critical land supplybottlenecksandskyrocketing landandhouseprices.Morecommonly,wherecontainmenthas failed, it has resulted in unplanned and disorderly urban expansion that has beenimpossibletoretrofitwithtrunkinfrastructurelinesorwithpublicopenspacesonceithastaken place. It is for this reason that we must insist that the European Commission’sestimatesgrosslyexaggeratetheamountoflandinurbanuseandaresimplywrong.

Furthermore,ifoneisledtobelievethattheworldisalready84%urbanthenonemayconclude that theurbanizationproject—the relentlessmigration of people fromvillage tocity, from living closer to the land to living closer to each other—is basically over. Thisproject, which started in earnest at the end of the 18th century, is now expected tograduallycometoaclosebytheendofthe21stcentury.IftheEuropeanCommissionisto


18

bebelieved,thereisverylittleleftforustodobuttoinfillandretrofitexistingsettlements.We,onourpart,believethatwestillhaveawindowofopportunitytoprepareourcitiesforabsorbingmore than2billionpeople by2050, the greatmajority of them in the cities ofless-developedcountries.Inadditiontoretrofittingexistingcities,wecanpreparecitiesfortheirinevitableexpansionbeforetheyarebuilt.Wecanprojecttheamountoflandthatwillbeneededtoaccommodatethenewurbandwellersand identifyappropriateareaswheregrowth can take place; we can plan arterial infrastructure grids in expansion areas andsecuretherights-of-wayforthesegridstoday,beforesettlementtakesplace;wecanprotectsensitive environmental assets and secure adequate lands for a hierarchy of public openspaces in advance of urban development; and we can engage landowners on the urbanfringeinimprovingtheirlandsubdivisionpractices.

Butforthistohappen,weneedtounderstandthatourworldisnotyetasurbanastheEuropeanCommissionwouldhaveusbelieve.Weneedtounderstandthattheworldisstillurbanizing, and that most future urbanization will take place in the less developedcountries,especiallyinAsiaandAfrica,countriesthatareleastpreparedforit.Ifwedeludeourselvestobelievethattheurbanizationprojectisoverthenweareunlikelytoengageinrealisticandpragmaticpreparationsofourcitiesandmetropolitanareasfortheirinevitableexpansion—handinhandwiththeirretrofittinganddensification,ofcourse—intheyearstocome.

Likeourcolleagues in theEuropeanCommission,weare frustratedby the inabilityofUnitedNationsmember states to adopt a commonurbanpopulation threshold.Weagreethatweseriouslyneedanddonotyethavearigorousandreliablemethod forestimatinghowurbanourcountries,ourregions,andourworldare.TheEuropeanCommissionistobeapplaudedfortakingthefirstattemptatamethodthat,incontrastwiththeUNPopulationDivision, uses a common urban population threshold for all countries. Unfortunately, itcoupled this threshold with an unrealistic urban density threshold’ and producedimplausibleestimatesinlightoftheargumentsabove.Thisshouldbeaninvitationtootherinterested parties to do better and to do better soon. For the time being, however, werecommendthat the84%numberproposedby theEuropeanCommissionnotbebandiedaboutbutbelaidtorest,andweurgeitscompetentresearcherstogobacktothedrawingboardandstartafresh.Theworktheyareengaged in isbothvaluableand important,andtheyhaveboththehumanandthefinancialresourcestogetitright.

***

our not-so-urban world

Documents