The Economics of in the Developing World

The MIT Faculty has made this article openly available. Please share how this access benefits you. Your story matters.

Citation Marx, Benjamin, Thomas Stoker, and Tavneet Suri. “The Economics of Slums in the Developing World.” Journal of Economic Perspectives 27, no. 4 (November 2013): 187–210.

As Published http://dx.doi.org/10.1257/jep.27.4.187

Publisher American Economic Association

Version Final published version

Citable link http://hdl.handle.net/1721.1/88128

Terms of Use Article is made available in accordance with the publisher's policy and may be subject to US copyright law. Please refer to the publisher's site for terms of use. Journal of Economic Perspectives—Volume 27, Number 4—Fall 2013—Pages 187–210

The Economics of Slums in the Developing World†

Benjamin Marx, Thomas Stoker, and Tavneet Suri

rrbanban ppopulationsopulations hhaveave sskyrocketedkyrocketed ggloballylobally aandnd ttodayoday rrepresentepresent mmoreore tthanhan hhalfalf ofof thethe world’sworld’s population.population. InIn somesome partsparts ofof thethe developingdeveloping world,world, thisthis U ggrowthrowth hhasas mmore-than-proportionatelyore-than-proportionately iinvolvednvolved rruralural mmigrationigration ttoo iinformalnformal ssettlementsettlements iinn aandnd aaroundround ccities,ities, kknownnown mmoreore ccommonlyommonly aass ““slums”—slums”— ddenselyensely ppopulatedopulated uurbanrban aareasreas ccharacterizedharacterized bbyy ppoor-qualityoor-quality hhousing,ousing, a llackack ooff aadequatedequate llivingiving sspacepace aandnd ppublicublic sservices,ervices, aandnd accommodatingaccommodating llargearge nnumbersumbers ooff informalinformal residentsresidents wwithith ggenerallyenerally iinsecurensecure ttenure.enure.1 WWorldwide,orldwide, aatt lleasteast 8860 million60 million ppeopleeople aarere nnowow llivingiving iinn sslums,lums, aandnd tthehe nnumberumber ooff sslumlum ddwellerswellers ggrewrew bbyy ssix millionix million eeachach yyearear ffromrom 22000000 ttoo 22010010 ((UN-HabitatUN-Habitat 22012a).012a). IInn ssub-Saharanub-Saharan AAfrica,frica, sslumlum ppopulationsopulations aarere ggrowingrowing aatt 44.5 percent.5 percent pperer aannum,nnum, a rrateate aatt wwhichhich ppopulationsopulations ddoubleouble eeveryvery 115 years.5 years. TThehe globalglobal expansionexpansion ofof urbanurban slumsslums posesposes questionsquestions forfor economiceconomic research,research, aass wwellell aass pproblemsroblems fforor ppolicymakers.olicymakers. SSomeome eeconomistsconomists ((FrankenhoffFrankenhoff 11967;967; TTurnerurner 11969;969; WorldWorld BankBank 2009;2009; GlaeserGlaeser 2011)2011) havehave suggestedsuggested a “modernization”“modernization” theorytheory ofof

1 Perhaps not surprisingly, the identifi cation of inhabitants suffers from the lack of a consistent terminology—for example, “slums” and “squatter settlements” are used almost interchangeably, although tenure and ownership institutions vary greatly across informal settlements. UN-Habitat (2006) applies the notion of “slum household” to any household lacking access to improved water, improved sanitation, suffi cient living area, durable housing, and secure tenure. Slum areas are generally thought of as geographic areas accommodating informal residents that combine several of these characteristics.

■ BBenjaminenjamin MarxMarx isis a PhDPhD studentstudent atat thethe DepartmentDepartment ofof Economics,Economics, MassachusettsMassachusetts InstituteInstitute ooff Technology,Technology, Cambridge,Cambridge, Massachusetts.Massachusetts. ThomasThomas StokerStoker isis GordonGordon Y.Y. BillardBillard ProfessorProfessor inin MManagementanagement andand EconomicsEconomics andand ProfessorProfessor ofof AppliedApplied EEconomics,conomics, aandnd TTavneetavneet SSuriuri isis MauriceMaurice FF.. SStrongtrong CCareerareer DDevelopmentevelopment PProfessorrofessor aandnd AAssociatessociate ProfessorProfessor ofof AppliedApplied EEconomics,conomics, bbothoth atat tthehe MMITIT SSloanloan SSchoolchool ooff MManagement,anagement, Cambridge,Cambridge, Massachusetts.Massachusetts. TTheirheir eemailmail aaddressesddresses aarere [email protected],[email protected], [email protected],[email protected], andand [email protected]@mit.edu. † To access the disclosure statements, visit http://dx.doi.org/10.1257/jep.27.4.187 doi=10.1257/jep.27.4.187 188 Journal of Economic Perspectives

sslums:lums: aaccordingccording ttoo tthishis tthinking,hinking, sslumslums aarere a ttransitoryransitory pphenomenonhenomenon ccharacteristicharacteristic ooff ffast-growingast-growing economies,economies, andand theythey progressivelyprogressively givegive wayway toto formalformal housinghousing asas eeconomicconomic ggrowthrowth ttricklesrickles ddownown aandnd ssocietiesocieties aapproachpproach tthehe llaterater sstagestages ooff eeconomicconomic ddevelopment.evelopment. EvenEven ifif slumslum aareasreas aappearppear sstabletable inin tthehe sshort-hort- oorr mmedium-edium- tterm,erm, tthishis aargumentrgument holds,holds, slumslum livingliving onlyonly representsrepresents a transitorytransitory phasephase inin thethe lifelife cyclecycle ofof rruralural migrants:migrants: thethe slumslum ddwellerswellers oror theirtheir childrenchildren eventuallyeventually movemove intointo formalformal hhousingousing withinwithin thethe city,city, soso thatthat thethe benefibenefi tsts ofof migrationmigration intointo thethe slumslum getget passedpassed aalonglong ffromrom ggenerationeneration ttoo ggeneration.eneration. BButut eevenven iiff uurbanrban ppovertyoverty iiss ppreferablereferable ttoo rruralural ppoverty,overty, aass aapparentlypparently sshownhown bbyy tthehe rrevealedevealed ppreferencereference ooff mmigrants,igrants, llifeife iinn a sslumlum iiss vveryery ddiffiiffi ccultult aandnd ooftenften ssubsistence-ubsistence- llevel.evel. MMoreover,oreover, sslumslums ddoo nnotot aalwayslways sseemeem ttoo bbee a ttemporaryemporary pphenomenonhenomenon ooff mmigrationigration ttoo ccities:ities: iinn mmanyany ccountriesountries sslumlum aareasreas hhaveave bbeeneen ggrowingrowing fforor ddecades,ecades, aandnd mmillionsillions ooff hhouseholdsouseholds fi ndnd themselvesthemselves ttrappedrapped iinn sslumslums fforor ggenerations.enerations. TThishis mmightight ssuggestuggest tthathat ttoday’soday’s sslumslums pposeose a pproblemroblem ooff a ddifferentifferent nnature:ature: bbecauseecause ooff mmultipleultiple mmarketarket aandnd ppolicyolicy ffailures,ailures, aacutecute ggovernanceovernance aandnd ccoordinationoordination pproblemsroblems tthathat hhinderinder iinvestment,nvestment, aandnd unsanitaryunsanitary livingliving conditionsconditions affectingaffecting tthehe dwellers’dwellers’ humanhuman capital,capital, lifelife inin thethe slumslum mmightight cconstituteonstitute a fformorm ooff ppovertyoverty ttraprap fforor a mmajorityajority ooff ttheirheir rresidents.esidents. IInn tthishis eessay,ssay, wwee pproviderovide hhistoricalistorical aandnd ccontemporaryontemporary ffactsacts ttoo aarguergue tthathat tthehe ttypeype ooff povertypoverty observedobserved iinn contemporarycontemporary sslumslums ooff tthehe ddevelopingeveloping wworldorld iiss ccharacteristicharacteristic ooff tthathat ddescribedescribed iinn tthehe lliteratureiterature oonn ppovertyoverty ttraps.raps. WWee ddocumentocument hhowow hhumanuman ccapitalapital tthresholdhreshold effects,effects, investmentinvestment inertia,inertia, aandnd a ““policypolicy ttrap”rap” mmayay ppreventrevent sslumlum ddwellerswellers ffromrom sseizingeizing eeconomicconomic oopportunitiespportunities oofferedffered bbyy ggeographiceographic pproximityroximity ttoo tthehe ccity.ity. WWee tthenhen ddiscussiscuss wwhetherhether tthehe bbasicasic aassumptionsssumptions ooff tthehe ““modernization”modernization” vviewiew hhold:old: tthathat iis,s, wwhetherhether ttherehere iiss a rrelationshipelationship bbetweenetween eeconomicconomic ggrowth,rowth, uurbanrban ggrowth,rowth, aandnd sslumlum ggrowthrowth iinn tthehe ddevelopingeveloping wworld,orld, aandnd wwhetherhether sstandardstandards ooff llivingiving ooff sslumlum ddwellerswellers aarere iimprovingmproving ooverver ttime,ime, bbothoth wwithinithin sslumslums aandnd aacrosscross ggenerations.enerations. FFinally,inally, wwee discussdiscuss wwhyhy sstandardtandard ppolicyolicy aapproachespproaches hhaveave ooftenften ffailedailed ttoo mmitigateitigate tthehe eexpansionxpansion ooff sslumslums iinn tthehe ddevelopingeveloping wworld.orld. OOurur aaimim iiss ttoo sstimulatetimulate sseriouserious aacademiccademic iinterestnterest aandnd ttoo iinformnform ppublicublic ddebateebate oonn tthehe eessentialssential iissuesssues pposedosed bbyy sslumslums iinn tthehe ddeveloping world.eveloping world.

A Contemporary Perspective on Slums

SSlumslums areare not,not, ofof course,course, a newnew phenomenon.phenomenon. TheyThey werewere a distinctivedistinctive featurefeature ooff EuropeanEuropean andand USUS citiescities duringduring thethe IndustrialIndustrial Revolution,Revolution, andand theythey persistedpersisted inin ssomeome ofof thesethese citiescities wellwell intointo thethe twentiethtwentieth century.century. TheThe well-knownwell-known slumsslums ofof thethe ppastast werewere oftenoften onon thethe outskirtsoutskirts ofof dynamicdynamic economiceconomic growth,growth, whichwhich bothboth attractedattracted mmigrantsigrants andand offeredoffered themthem somesome accessaccess toto economiceconomic opportunities.opportunities. ForFor example,example, tthehe WhitechapelWhitechapel areaarea ofof EastEast LondonLondon attractedattracted a vastvast numbernumber ofof poorpoor ruralrural mmigrantsigrants duringduring thethe eighteentheighteenth andand nineteenthnineteenth centurycentury duedue toto thethe newnew factoriesfactories aandnd shopsshops ofof thatthat partpart ofof thethe city.city. TheThe Hell’sHell’s KitchenKitchen areaarea ofof NewNew YorkYork CityCity onon thethe HHudsonudson RiverRiver sideside ofof ManhattanManhattan attractedattracted immigrantsimmigrants inin largelarge partpart becausebecause ofof itsits pproximityroximity toto docksdocks andand railroads,railroads, asas wellwell asas toto thethe growinggrowing citycity nearby.nearby. InIn thethe past,past, mmoderateoderate toto radicalradical policypolicy solutionssolutions werewere adoptedadopted toto addressaddress thethe overcrowdingovercrowding andand Benjamin Marx, Thomas Stoker, and Tavneet Suri 189

Table 1 Two Lists of the Developing World’s Largest Slums

UN-Habitat (2003) Davis (2006)

City, Population City, Population Name of slum Country estimate Name of slum Country estimate

Dharavi Mumbai, Over 500,000 Neza/Chalco/Izta City, 4 million India Mexico Orangi Karachi, Over 500,000 Liberatador , 2.2 million Pakistan Venezuela Nairobi, 400,000 El Sur/Ciudad Bogota, 2.0 million Kenya Bolivar Columbia , 300,000 San Juan de Lima, 1.5 million Lurigancho Peru Ashaiman Tema, 150,000 Cono Sur Lima, 1.5 million Ghana Peru Ajegunle Lagos, 1.5 million Nigera Sadr City Baghdad, 1.5 million Iraq Soweto Gautung, 1.5 million South Africa Gaza Palestine 1.3 million Orangi Karachi, 1.2 million Pakistan

Source: UN-Habitat (2003); Davis (2006). Notes: UN-Habitat (2003) does not provide a comprehensive list of slums ranked by their population. The numbers given are mentioned in the report as part of individual case studies of slums. uunsanitarynsanitary conditionsconditions inin thesethese typestypes ofof areas.areas. ExamplesExamples ofof suchsuch policiespolicies includeinclude BBaronaron Haussman’sHaussman’s revampingrevamping ofof ParisParis inin thethe 1860s–1870s,1860s–1870s, whichwhich involvedinvolved alteringaltering mmoreore thanthan halfhalf ofof thethe city’scity’s buildingsbuildings andand creatingcreating a sewagesewage systemsystem andand widewide bboulevardsoulevards inin lieulieu ofof slumslum neighborhoods.neighborhoods. MoreMore recently,recently, Singapore’sSingapore’s compulsorycompulsory ssavingsavings schemescheme inin thethe 1960s1960s waswas usedused toto fi nancenance thethe constructionconstruction ofof publicpublic housinghousing aandnd toto enableenable slumslum residentsresidents toto purchasepurchase formalformal housinghousing unitsunits atat a subsidizedsubsidized rate.rate. TToday,oday, llargearge sslumlum ssettlementsettlements hhaveave ddisappearedisappeared iinn mmostost aadvanceddvanced eeconomies,conomies, bbutut iitt iiss ffarar ffromrom cclearlear hhowow ccomparableomparable tthesehese hhistoricalistorical eexamplesxamples aarere ttoo tthehe ssituationsituations ffacedaced iinn tthehe ddevelopingeveloping wworld.orld. SSomeome ooff ttoday’soday’s sslumslums aarere iinn ccountriesountries eexperiencingxperiencing rrapidapid eeconomicconomic ggrowth,rowth, ssuchuch aass CChina,hina, bbutut mmanyany sslumslums aarere llocatedocated iinn ccountriesountries wwithith sslowlow oorr sstagnanttagnant ggrowth.rowth. TThehe pprevalencerevalence ooff sslumslums iiss hhighestighest iinn ssub-Saharanub-Saharan AAfrica,frica, wwherehere sslumlum ddwellerswellers rrepresentepresent 662 2 p percentercent ooff tthehe uurbanrban ppopulationopulation ((UN-HabitatUN-Habitat 22012a):012a): aass ooff 22005,005, tthehe tthree countrieshree countries wwithith tthehe hhighestighest ffractionraction ooff tthehe uurbanrban ppopula-opula- ttionion llivingiving iinn sslumslums ——SierraSierra LLeone,eone, SSudan,udan, aandnd tthehe CCentralentral AAfricanfrican RRepublic—areepublic—are aalsolso llocatedocated oonn tthathat ccontinent.ontinent. EEstimatesstimates ooff aactualctual ppopulationsopulations oonn a sslum-by-slumlum-by-slum bbasisasis aacrosscross tthehe ddevelopingeveloping wworldorld aarere ffewew aandnd ffarar bbetweenetween aandnd vvaryary wwidelyidely aacrosscross ssources.ources. TTable 1able 1 ddisplaysisplays ttwo lists,wo lists, ggivingiving a ssenseense ooff tthehe rrangeange ooff mmeasurementseasurements tthathat 190 Journal of Economic Perspectives

ccanan ooccur.ccur. TThehe lleft-handeft-hand ccolumnolumn llistsists fi veve ofof thethe world’sworld’s largestlargest slums,slums, withwith roughrough ppopulationopulation estimates,estimates, asas providedprovided byby UN-HabitatUN-Habitat ((2003).2003). AAnn uunoffinoffi cialcial r rankinganking pprovidedrovided iinn DDavisavis ((2006)2006) llistsists 330 slums0 slums wwithith ooverver 5500,000 inhabitants,00,000 inhabitants, iincludingncluding ttenen iinn ssub-Saharanub-Saharan AAfricafrica aandnd nnineine iinn LLatinatin AAmerica.merica. TThehe tten slumsen slums wwithith llargestargest ppopula-opula- ttionion bbyy tthishis mmeasureeasure aarere sshownhown iinn tthehe rright-handight-hand ccolumnsolumns ooff TTable 1.able 1. DDharaviharavi aandnd KKibera,ibera, tthehe fi rstrst aandnd tthird slumshird slums iinn tthehe UUN-HabitatN-Habitat ((2003)2003) llistist bbothoth aappearppear ffurtherurther ddownown iinn tthehe DDavisavis ((2006)2006) rrankinganking wwithith aann eestimatedstimated ppopulationopulation ooff 8800,000 each.00,000 each. OOff course,course, thethe factfact thatthat a lotlot ofof thethe globalglobal slumslum expansionexpansion takestakes placeplace inin poorpoor eeconomiesconomies doesdoes notnot invalidateinvalidate a “modernization”“modernization” oror transitorytransitory viewview ofof slums.slums. SinceSince rruralural migrationmigration continuescontinues unabatedunabated eveneven inin thosethose countries,countries, urbanurban productivityproductivity mmustust bebe risingrising relativerelative toto ruralrural productivity,productivity, eithereither becausebecause capitalcapital accumulationaccumulation aandnd ttechnologicalechnological progressprogress areare concentratedconcentrated inin citiescities oror becausebecause ruralrural productivityproductivity iiss declining.declining. TheThe locationlocation decisiondecision ofof slumslum dwellersdwellers alsoalso indicatesindicates thatthat standardsstandards ofof llivingiving inin slumsslums aarere ssomewhatomewhat preferablepreferable toto thosethose inin thethe ruralrural hinterlands.hinterlands. GlaeserGlaeser ((2011)2011) pprovidedrovided eevidencevidence tthathat tthehe uurbanrban ppooroor wworldwideorldwide aarere oonn aaverageverage rrichericher aandnd hhappierappier thanthan ttheirheir rruralural ccounterparts,ounterparts, aandnd CChowdhury,howdhury, MMobarak,obarak, aandnd BBryanryan ((2009)2009) sshowedhowed tthathat sseasonaleasonal uurbanrban mmigrationigration iinn BBangladeshangladesh ccanan ggenerateenerate wwelfareelfare iimprove-mprove- mmentsents fforor ffamiliesamilies oof migrants.f migrants. HHowever,owever, tthesehese factsfacts saysay littlelittle aboutabout thethe naturenature ofof povertypoverty inin slums—andslums—and wwhetherhether itit ccanan bebe escapedescaped viavia thethe competitivecompetitive marketmarket forcesforces offeredoffered byby thethe city.city. TTherehere hashas been,been, inin fact,fact, a relativerelative lacklack ofof empiricalempirical researchresearch workwork conductedconducted inin sslumslums ooff tthehe ddevelopingeveloping wworldorld ttoo ttestest tthishis aargument.rgument.2 OOnene rreasoneason isis tthathat datadata collec-collec- ttionion inin slumsslums isis problematicproblematic duedue toto a varietyvariety ofof factors,factors, includingincluding safetysafety issuesissues forfor rresearchesearch fi eldworkers,eldworkers, thethe highhigh mobilitymobility andand turnoverturnover ratesrates ofof respondents,respondents, andand thethe ffactact tthathat ttargetarget hhouseholdsouseholds aarere rregularlyegularly aabsentbsent ffromrom ttheirheir ddwellings.wellings. AAss a rresult,esult, ffewew sstudiestudies havehave triedtried toto documentdocument thethe degreedegree ofof intergenerationalintergenerational socialsocial mobilitymobility ofof sslumlum ddwellerswellers inin thethe developingdeveloping world.world. Here,Here, wewe startstart byby illustratingillustrating whatwhat lifelife isis likelike iinn sslumslums aandnd wwhathat ccharacteristicsharacteristics seemseem toto bebe commoncommon acrossacross slums,slums, usingusing recentrecent ssurveysurveys collectedcollected inin slumsslums ofof four countries:four countries: Bangladesh,Bangladesh, India,India, Kenya,Kenya, andand SierraSierra LLeone.eone. WeWe rrelyely mmostost hheavilyeavily oonn tthehe ccasease ooff tthehe KKiberaibera slumslum iinn NNairobi,airobi, Kenya,Kenya, wherewhere wwee hhaveave cconductedonducted extensiveextensive fi eldworkeldwork overover thethe pastpast two years.two years. Table 2Table 2 providesprovides sspecifipecifi c sourcessources ofof datadata forfor thesethese slumslum aareas.reas.

Slums as Poverty Traps

IInn tthishis paper,paper, wewe argueargue thatthat slumsslums maymay bebe povertypoverty trapstraps andand areare thereforetherefore nneithereither temporarytemporary nnoror a sshorthort sstoptop oonn thethe wayway toto greatergreater economiceconomic opportunities.opportunities.

2 A nonexhaustive list of recent empirical research conducted in slums includes Banerjee, Dufl o, Glennerster, and Kinnan (2010) in Hyderabad; Banerjee, Pande, Vaidya, Walton, and Weaver (2011) in Delhi; El-Zanaty and Way (2004) in ; Field (2007) in Peru; Galiani et al. (2013) in slums of three Latin American countries; Gulyani, Bassett, and Talukdar (2012) in Dakar and Nairobi; and Perlman (2010) in . Gupta, Arnold, and Lhungdim (2009) analyze the NFSH-3 dataset on which we also rely. The Economics of Slums in the Developing World 191

Table 2 Description of Datasets

Number of slum households Country Locations Source Year surveyed

Bangladesh Tongi, Jessore SHAHAR Project/CARE- 2000 26,830 Bangladesh (Baseline census) Bangladesh Tongi, Jessore SHAHAR Project/CARE- 2000 1,120 Bangladesh (Baseline survey) India Delhi, Meerut, Kolkata, Indore, National Family Health 2005 – 8,669 Mumbai, Nagpur, Hyderabad, Survey (NFHS-3) 2006 Chennai India Hyderabad Banerjee, Dufl o, 2005 2,800 Glennerster, and Kinnan (2010) Kenya Kibera (Nairobi) Kenya National Bureau of 1999, 64,588 Statistics (national census) 2009 (approx.) Kenya Kibera (Nairobi) Marx, Stoker, and Suri 2012 31,765 (2013) Kenya Kibera (Nairobi) Marx, Stoker, and Suri 2012 1,093 (2013) Sierra Leone Western UNICEF Sierra Leone 2010 789 (Freetown) SMART Survey

Sources: The Bangladesh data is available from http://dvn.iq.harvard.edu/dvn/ and was collected in 2000 as part of the SHAHAR census and baseline surveys. The India data was collected with approximately 8,669 slum households in eight Indian cities as part of the 2005 –2006 National Family Health Survey (NFSH-3). It is available from http://www.measuredhs.com/ (India Standard DHS, 2005 – 06). The Hyderabad dataset (Banerjee, Dufl o, Glennerster, and Kinnan 2010) was collected as part of a randomized evaluation on the impact of microfi nance. It is publicly available from the data repository of the Jameel Poverty Action Lab ( J-PAL) at http://thedata.harvard.edu/dvn/dv/jpal. Our Kenya data come from two different sources. First, we collected one census and one household survey in Kibera, Nairobi’s largest slum area. Second, we obtained from the Kenya National Bureau of Statistics (KNBS) two waves of complete micro census data collected in the same area in 1999 and 2009. The Sierra Leone data was collected as part of the 2010 UNICEF SMART Survey and is proprietary from UNICEF. We use these sources throughout unless otherwise indicated.

TTherehere isis a widewide literatureliterature onon povertypoverty traps,traps, includingincluding a theoreticaltheoretical literatureliterature high-high- llightingighting thethe specifispecifi c mechanismsmechanisms leadingleading toto povertypoverty trapstraps (for(for excellentexcellent defidefi nitionsnitions aandnd rreviews,eviews, uusefulseful sstartingtarting ppointsoints iincludenclude BBasuasu 22003,003, MMatsuyamaatsuyama 22005,005, aandnd BBowles,owles, DDurlauf,urlauf, aandnd HHoffoff 22006).006). TheThe lliteratureiterature hashas alsoalso describeddescribed spatialspatial povertypoverty traps,traps, bbutut mmostlyostly iinn rruralural ssettingsettings ( JJalanalan andand RavallionRavallion 2002;2002; GolgherGolgher 2012).2012). WeWe argueargue tthathat urbanurban slumsslums presentpresent a differentdifferent challengechallenge toto communitiescommunities andand governmentsgovernments aadministeringdministering them,them, andand thatthat thethe veryvery naturenature ofof lifelife inin thethe slumsslums makesmakes itit diffidiffi cultcult ttoo aachievechieve iimprovementsmprovements iinn sstandardstandards ooff llivingiving tthroughhrough mmarginalarginal iinvestmentsnvestments iinn hhousing,ousing, health,health, oror infrastructureinfrastructure alone.alone. WeWe nownow discussdiscuss somesome ofof thethe mechanismsmechanisms rrelevantelevant ttoo sslumlum ccontextsontexts tthathat mmayay lleadead ttoo ppovertyoverty ttraps.raps. 192 Journal of Economic Perspectives

Human Capital DDespiteespite ttremendousremendous vvariationsariations aacrosscross sslums,lums, iissuesssues ccommonommon ttoo aallll slumslum settingssettings aarere a llackack ooff aadequatedequate llivingiving sspace,pace, iinsuffinsuffi ccientient ppublicublic ggoodsoods pprovision,rovision, aandnd tthehe ppooroor qqualityuality ofof basicbasic aamenities,menities, aallll ooff wwhichhich lleadead ttoo eextremelyxtremely ppooroor hhealthealth aandnd llowow llevelsevels ooff hhumanuman ccapital.apital.3 IInn tthishis ssection,ection, wwee ffocusocus oonn tthehe hhealthealth aaspectsspects ooff hhumanuman ccapital.apital. EEducation,ducation, aalbeitlbeit aanothernother iimportantmportant ccomponentomponent ooff hhumanuman ccapital,apital, iiss a llessess rrelevantelevant mmetricetric forfor ourour aargumentrgument ggiveniven tthathat uuniversalniversal ffreeree pprimaryrimary eeducationducation llawsaws hhaveave rreducededuced ddisparitiesisparities iinn aaccessccess ttoo eeducationducation bbetweenetween rruralural aandnd uurbanrban ssettings,ettings,4 andand tthathat rrural–urbanural–urban ddifferencesifferences iinn tthehe qqualityuality ooff eeducationducation aarere eextremelyxtremely ddiffiiffi cultcult to measure.to measure. IInn tthehe KKiberaibera aarearea ooff KKenya,enya, informalinformal hhouseholdsouseholds reportedreported inin 20092009 aann aaverageverage ddwellingwelling sizesize ofof 1.17 1.17 habitablehabitable roomsrooms (with(with averageaverage householdhousehold sizesize ofof 3.15),3.15), asas oopposedpposed toto 1.951.95 forfor urbanurban householdshouseholds andand 2.972.97 forfor ruralrural households.households. ForFor perspec-perspec- ttive,ive, accordingaccording toto UN-HabitatUN-Habitat (2006),(2006), a dwellingdwelling providesprovides “suffi“suffi cientcient livingliving space”space” iiff eeachach roomroom isis ssharedhared byby nono moremore thanthan three individuals.three individuals. InIn thethe ZimbabweZimbabwe slumslum iinn AAbidjan,bidjan, ppopulationopulation densitydensity waswas reportedreported toto bebe aass hhighigh aass 334,000 inhabitants4,000 inhabitants perper ssquarequare kilometerkilometer (UN-Habitat(UN-Habitat 2003).2003). AsAs a pointpoint ofof comparison,comparison, Manhattan’sManhattan’s popu-popu- llationation ddensityensity wwasas 226,9246,924 pperer ssquarequare kkilometerilometer iinn 22010010 ((USUS CCensusensus BBureauureau 22013).013). AAcrosscross slumslum ssettings,ettings, tthehe aadversedverse hhealthealth eeffectsffects ooff oovercrowdingvercrowding aarere aaggravatedggravated bbyy ppooroor accessaccess toto waterwater andand sanitationsanitation facilities.facilities. Table 3Table 3 showsshows thatthat thethe majoritymajority ofof sslumlum ddwellerswellers aacrosscross oourur ddatasetsatasets hhaveave nnoo pprivaterivate llatrine,atrine, aandnd mmanyany uusese iinferior-typenferior-type llatrinesatrines (such(such asas anan openopen spacespace oror traditionaltraditional pitpit thatthat isis notnot connectedconnected toto a sewagesewage nnetwork),etwork), nono sourcesource ofof privateprivate water,water, andand nono garbagegarbage collectioncollection (meaning(meaning thatthat ggarbagearbage isis eithereither leftleft inin a roadsideroadside ditchditch oror burntburnt nextnext toto thethe householdhousehold dwelling).dwelling). TThesehese datadata areare corroboratedcorroborated byby a rangerange ofof studiesstudies documentingdocumenting thethe poorpoor waterwater aaccessccess aandnd ooverallverall hhygieneygiene ooff sslumlum nneighborhoods.eighborhoods. FForor eexample,xample, iinn MMumbai’sumbai’s SShivahiva SShaktihakti NNagaragar sslum,lum, ccommunityommunity ttapsaps aarere rreportedlyeportedly ssharedhared bbyy 1100 people00 people oonn aaverageverage ((WorldWorld BankBank 2009).2009). InIn a surveysurvey ofof slumslum dwellersdwellers inin Delhi,Delhi, BanerjeeBanerjee etet al.al. (2011)(2011) ffoundound thatthat thethe environmentenvironment ofof 83 percent83 percent ofof toilettoilet sitessites waswas infectedinfected withwith fecalfecal oror ootherther wwasteaste mmatter.atter. AAbsentbsent oorr ddefiefi cientcient wwaterater aandnd ssewageewage ssystemsystems ttranslateranslate iintonto a bbroadroad rrangeange ooff hhealthealth andand sanitationsanitation issues,issues, whetherwhether throughthrough directdirect exposureexposure toto bacterialbacterial agents,agents, ccontaminatedontaminated ddrinkingrinking wwater,ater, oorr ootherther cchannels.hannels. DDuflufl o,o, Galiani,Galiani, andand MobarakMobarak (2012)(2012) ddescribedescribed tthehe ddiseaseisease bburdenurden aarisingrising ffromrom tthehe uunsanitarynsanitary llivingiving cconditionsonditions iinn sslums.lums. IInn thethe slumsslums ooff TTongiongi aandnd JJessoreessore iinn BBangladesh,angladesh, 882 percent2 percent ooff rrespondentsespondents rreporteport aanyny hhouseholdousehold mmemberember bbeingeing ssickick iinn tthehe ppastast 330 days.0 days. IInn KKibera,ibera, 116 percent6 percent ooff oourur rrespondentespondent householdshouseholds hhaveave atat leastleast one one membermember chronicallychronically illill inin thethe previousprevious tthree months.hree months. IInn SSierraierra LLeone,eone, a ccountryountry wwhosehose sslumslums rroutinelyoutinely eexperiencexperience ccholeraholera

3 The World Bank (2009) argued that urban areas fare consistently better than rural areas worldwide along a variety of health indicators. In this section, we compare rural areas with slum areas specifi cally. 4 Lopez (2007) shows evidence of this trend for Latin America. Hannum, Wang, and Adams (2008) study the case of China, where some urban–rural disparities remain, but the primary enrollment of 7–16 year olds in rural areas is nearly universal. Worldwide, school attendance rates have increased in the majority of low-income countries since 1990, and rural areas were the primary benefi ciaries of this trend (World Bank 2009). Benjamin Marx, Thomas Stoker, and Tavneet Suri 193

Table 3 Public Goods and Basic Amenities across Slums

No private Inferior No private No garbage latrine latrine type water source collection

Tongi (Dhaka) 70% 34% 81% 64% Hyderabad 46% 43% 61% NA Kibera NA 63% 92% 73% Kolkata 75% 46% 57% NA Mumbai 78% 8% 12% NA All Indian slums NFSH-3 68% 49% 25% NA

Source: Authors using data from the SHAKAR Project for Bangladesh; Marx, Stoker, and Suri (2013) for Kenya; and the 2005 –2006 National Family Heath Survey (NFSH-3) for India. Notes: In the fi rst column, we report the fraction of slum households who share their latrine with other households. In the second column, we report the fraction of households using an open space or traditional pit as latrine. To compute this number, we combine “open air” and “septic tank/pit toilet” in Hyderabad; “traditional pit latrine,” “bucket,” and “bush or river or stream” in Kibera; “fl ush to septic tank,” “fl ush to pit latrine,” “fl ush to somewhere else,” “fl ush, don’t know where,” “ventilated improved pit latrine,” “pit latrine with slab,” “pit latrine without slab/open pit,” “no facility/bush/fi eld,” “composting toilet,” and “dry toilet” in the 2005 –2006 National Family Heath Survey (NFSH-3) data (Kolkata, Mumbai and “All Indian slums”). In the third column, we report the fraction of households who share their main drinking water source with other households. This combines “share restricted” and “shared unrestricted” in the Bangladesh data, “public tap” and “public water tank” in Kibera, “public tap/standpipe” and “tube well or boreholes outside the dwelling” in the NFSH-3 data. In the fourth column, we report the fraction of households whose garbage is not collected by a public or a private company. “NA” means “not available.” ooutbreaks,utbreaks, sslumlum hhouseholdsouseholds eexhibitxhibit ppooreroorer hhealthealth ooutcomesutcomes tthanhan ttheirheir rruralural ccounter-ounter- pparts.arts. TheThe pprevalencerevalence ooff uunderweight,nderweight, sstunting,tunting, aandnd wwastingasting ((acuteacute mmalnutrition)alnutrition) iiss iinn factfact ggreaterreater iinn tthehe slumslum ooutskirtsutskirts ooff tthehe ccapitalapital FFreetownreetown tthanhan iinn rruralural aareasreas nnation-ation- wwide,ide, aass cchildrenhildren uundernder fi vvee llivingiving iinn sslumslums hhaveave ssignifiignifi ccantlyantly llowerower wweight-for-ageeight-for-age aandnd wweight-for-heighteight-for-height iindexesndexes tthanhan cchildrenhildren uundernder fi vvee iinn rruralural aareas.reas.5 AAcrosscross ccitiesities iinn tthehe ddevelopingeveloping wworld,orld, ttherehere isis somesome evidenceevidence thatthat lifelife expectancyexpectancy isis lower,lower, andand infantinfant mmortalityortality hhigherigher aamongmong tthehe uurbanrban ppooroor tthanhan aamongmong ccomparableomparable ggroupsroups iinn rruralural aandnd fformalormal uurbanrban aareasreas ((Bradley,Bradley, SStephens,tephens, HHarpham,arpham, aandnd CCairncrossairncross 11992).992). HHealthealth andand sanitationsanitation issuesissues areare renderedrendered moremore pproblematicroblematic bbyy tthehe llackack ooff pprovisionrovision ofof a socialsocial safetysafety netnet inin slums.slums. SlumSlum livingliving involvesinvolves a widewide rangerange ofof risks:risks: iinn ourour KiberaKibera data,data, 10 percent10 percent ofof householdshouseholds havehave experiencedexperienced beingbeing evictedevicted fromfrom ttheirheir dwelling,dwelling, andand 4 percent4 percent reportreport atat leastleast oonene ddeatheath iinn tthehe hhouseholdousehold inin thethe ppastast ssix months.ix months. InIn Bangladesh,Bangladesh, 56 percent56 percent ofof respondentsrespondents saysay theythey dodo notnot meetmeet theirtheir bbasicasic nneedseeds ofof ffood,ood, wwater,ater, sshelter,helter, aandnd hhealthcare;ealthcare; 448 percent8 percent ddoo nnotot ffeeleel ssafeafe iinn ttheirheir hhouseouse duringduring badbad weather;weather; andand ofof thethe householdshouseholds reportingreporting one memberone member beingbeing illill iinn thethe previousprevious 30 days,30 days, 26 percent26 percent saysay thatthat theythey couldcould notnot affordafford toto seekseek medicalmedical aattention.ttention. InIn HyderabadHyderabad slums,slums, wwherehere 770 percent0 percent ooff hhouseholdsouseholds cclassifylassify tthemselveshemselves aass ““poor,”poor,” oonlynly 112 percent2 percent rreporteport rreceivingeceiving aanyny aassistancessistance fromfrom thethe government.government.

5 Calculated by the authors using the UNICEF SMART Survey data for Sierra Leone. 194 Journal of Economic Perspectives

A wwideide lliterature,iterature, bbothoth mmacroeconomicacroeconomic aandnd mmicroeconomic,icroeconomic, ddocumentsocuments tthehe iimportancemportance ooff hhealthealth fforor iincomencome aandnd ooff eearlyarly hhealthealth iinvestmentsnvestments fforor llonger-termonger-term ooutcomes:utcomes: ggoodood rreviewseviews ccanan bbee ffoundound iinn LLópez-Casanovas,ópez-Casanovas, RRivera,ivera, aandnd CCurraisurrais ((2005),2005), BBleakleyleakley ((2010),2010), aandnd CCurrieurrie aandnd VVoglogl ((2013).2013). IInn ddevelopingeveloping eeconomies,conomies, ttherehere aarere llargearge rreturnseturns ttoo hhealthealth iimprovementsmprovements aandnd sstrongtrong ccomplementaritiesomplementarities bbetweenetween iinvest-nvest- mmentsents iinn cchildhild hhealthealth aandnd cchildhild eeducationducation ((MiguelMiguel aandnd KKremerremer 22004).004). PPooroor hhumanuman ccapitalapital andand poorpoor avenuesavenues forfor humanhuman ccapitalapital iinvestmentnvestment iinn sslumlum hhouseholdsouseholds mmayay tthereforeherefore lleadead ttoo a lacklack ofof socialsocial mobilitymobility acrossacross generationsgenerations ofof slumslum residents.residents. Thus,Thus, tthishis hhealthealth ddataata sseemseems aatt oddsodds withwith thethe “transitory”“transitory” hypothesis,hypothesis, mmakingaking iitt questionablequestionable wwhetherhether sslumlum iinhabitantsnhabitants mmeeteet tthehe ““criticalcritical tthresholds”hresholds” iinn hhumanuman ccapitalapital rrequiredequired fforor tthehe ccompetitiveompetitive fforcesorces ooff thethe laborlabor marketmarket toto comecome intointo play,play, asas describeddescribed theoreti-theoreti- ccallyally iinn AAzariadiszariadis aandnd DDrazenrazen ((1990).1990). SSlumlum ddwellerswellers mmayay fi ndnd themselvesthemselves ttrappedrapped iinn a llow-skilled,ow-skilled, llow-incomeow-income eequilibriumquilibrium aass tthehe ccontinuousontinuous iinflnfl uuxx ooff rruralural migrantsmigrants mmain-ain- ttainsains wageswages atat near-subsistencenear-subsistence llevels,evels, hhinderingindering tthehe iinvestmentsnvestments iinn hhumanuman ccapitalapital tthathat wwouldould bbee rrequiredequired ttoo ooffsetffset tthehe aadversedverse eeffectsffects ooff sslum living.lum living.

Investment Inertia SSlumslums nnotot oonlynly sseemeem ttrappedrapped iinn a llow-human-capitalow-human-capital eequilibrium,quilibrium, bbutut ttheyhey aalsolso eexhibitxhibit dysfunctionaldysfunctional institutions,institutions, lowlow levelslevels ofof physicalphysical capital,capital, andand poorpoor accessaccess ttoo ddevelopedeveloped services.services. SlumsSlums cancan bebe thoughtthought ofof asas areasareas ofof depresseddepressed publicpublic andand pprivaterivate iinvestmentnvestment wwherehere nneithereither ggovernmentovernment nnoror bbroaderroader ssocietyociety hhasas mmanagedanaged ttoo oorganizerganize inin a wayway thatthat providesprovides forfor widespreadwidespread provisionprovision andand maintenancemaintenance ofof publicpublic ggoodsoods ((andand wwee aarere ddefiefi ningning “public“public good”good” broadlybroadly toto includeinclude cleanclean water,water, sanitation,sanitation, ggarbagearbage collection,collection, a socialsocial safetysafety nnet,et, aandnd tthehe llegalegal iinfrastructurenfrastructure ooff ppropertyroperty rrightsights tthathat aallowsllows fforor aann eeffectiveffective mmarketarket iinn llandand aandnd hhousing).ousing). IInn tthishis ssection,ection, wwee ddescribeescribe fi vve distincte distinct pphenomenahenomena tthathat ccanan lleadead ttoo llowow iinvestmentnvestment iinn sslums.lums. A fi rst factorrst factor isis thethe well-knownwell-known informalityinformality ofof propertyproperty rightsrights intrinsicintrinsic toto slumslum aareas.reas. WWithoutithout fformalormal llandand ttitles,itles, sslumlum ddwellerswellers llackack tthehe iincentivesncentives ttoo iimprovemprove tthehe qqualityuality ofof theirtheir homeshomes andand neighborhoods.neighborhoods. InformalInformal settlementssettlements havehave typicallytypically eemergedmerged oonn vacantvacant governmentgovernment land,land, whichwhich impliesimplies thatthat thethe propertyproperty rightsrights overover tthehe landland heldheld byby individualsindividuals livingliving therethere areare highlyhighly illiquid,illiquid, althoughalthough theythey maymay bebe eenforceablenforceable locally.locally. InIn DakarDakar andand Nairobi,Nairobi, onlyonly 1919 andand 34 percent34 percent ofof ownersowners respec-respec- ttivelyively reportreport thatthat itit isis easyeasy toto transacttransact housinghousing inin theirtheir areaarea (Gulyani,(Gulyani, Basset,Basset, andand TTalukdaralukdar 2012).2012). DeDe SotoSoto (2000)(2000) popularizedpopularized thisthis argument,argument, andand suggestedsuggested thatthat a llowerower riskrisk ofof eevictionviction aandnd ttighterighter ppropertyroperty rrightsights oonn tthehe llandand ccouldould uunlocknlock aaccessccess ttoo ccreditredit markets.markets. MoreMore recently,recently, FieldField (2005)(2005) andand GalianiGaliani andand SchargrodskySchargrodsky (2010)(2010) sshowedhowed tthathat fformalormal ttitlingitling ccouldould eencouragencourage iinvestmentsnvestments iinn ppooroor uurbanrban aareas.reas. A secondsecond factorfactor isis thethe concurrenceconcurrence ofof overcrowdingovercrowding ofof slumslum areasareas andand lowlow mmarginalarginal rreturnseturns ffromrom ssmallmall uupgradingpgrading iinvestments.nvestments. IItt mmayay tthereforeherefore nnotot bbee rrationalational fforor sslumlum ddwellerswellers ttoo fi nancenance investmentsinvestments inin housinghousing oror infrastructure.infrastructure. InIn addition,addition, mmanyany upgradesupgrades maymay requirerequire ratherrather largelarge privateprivate investments.investments. ForFor example,example, GalianiGaliani eett al.al. (2013)(2013) showshow howhow eveneven simplesimple improvementsimprovements inin housinghousing inin slumsslums inin Mexico,Mexico, UUruguay,ruguay, andand ElEl SalvadorSalvador costcost asas muchmuch asas $1,000$1,000 perper household,household, althoughalthough suchsuch iinvestmentsnvestments dodo generategenerate improvementsimprovements inin qualityquality ofof lifelife andand safety.safety. ThisThis situationsituation The Economics of Slums in the Developing World 195

Table 4 Ownership Type of Main Dwelling

Own Rent Occupy Other arrangements

Tongi (Dhaka) 52% 41% 7% 0% Hyderabad 68% 28% 3% 1% Kibera 7% 92% 1% 1% Kolkata 37% 56% NA 8% Mumbai 72% 26% NA 2%

Source: Authors using data from the SHAKAR Project for Bangladesh; Marx, Stoker, and Suri (2013) for Kenya; and the 2005 –2006 National Family Heath Survey (NFSH-3) for India. Notes: “Occupy” refers to situations where the respondent lives in a dwelling without paying for rent and without holding a title on the land. “Other arrangements” include land given by the government in Hyderabad, land owned by a friend or relative in Kibera, and land obtained as part of an employment or any other arrangement in Mumbai and Kolkata. The “Other arrangements” category may include occupiers in Mumbai and Kolkata, although this is not clear from the survey questionnaire. “NA” means “not available.” sstandstands inin starkstark contrastcontrast toto somesome ofof thethe problemsproblems thatthat characterizecharacterize ruralrural poverty,poverty, wwherehere relativelyrelatively cheapcheap technologiestechnologies cancan oftenoften lleadead ttoo ssubstantialubstantial iimprovementsmprovements iinn iincomencome andand welfarewelfare (for(for example,example, thethe resultsresults ofof “green“green revolution”revolution” technologiestechnologies iinn aagriculture).griculture). NotNot oonlynly aarere pprivaterivate iinvestmentsnvestments iinn hhousingousing iinfrastructurenfrastructure lowlow inin sslums,lums, bbutut DDuflufl o,o, Galiani,Galiani, andand MobarakMobarak (2012)(2012) havehave documenteddocumented a lowlow willingness-willingness- tto-payo-pay fforor iimprovedmproved ppublicublic ggoodsoods iinn ppooroor uurbanrban aareas.reas. A ““bigbig ppush”ush” aapproachpproach wwouldould tthenhen seemseem necessarynecessary toto addressaddress thethe lacklack ofof investmentinvestment inin slums,slums, andand toto generategenerate aaggregateggregate demanddemand spilloversspillovers inin thethe areaarea ofof publicpublic goodsgoods andand basicbasic servicesservices —the—the sseminaleminal mmodelodel ddescribedescribed MMurphy,urphy, SShleifer,hleifer, aandnd VVishnyishny ((1989)1989) wwouldould jjustifyustify tthishis ttypeype oof intervention.f intervention. A third,third, lessless well-knownwell-known causecause forfor lowlow investmentinvestment levelslevels inin slumsslums couldcould bebe thethe hhighigh rentrent ppremiumsremiums tthathat ddwellerswellers mmustust ppayay ttoo lliveive iinn ccloselose pproximityroximity ttoo tthehe ccity,ity, aandnd wwhichhich rreduceeduce oopportunitiespportunities fforor ssavingsavings aaccumulation.ccumulation. WWhilehile sslumlum ddwellerswellers aarere ttypicallyypically thoughtthought ofof asas squatterssquatters occupyingoccupying vacantvacant publicpublic land,land, availableavailable evidenceevidence ssuggestsuggests thatthat a llargearge nnumberumber ooff ddwellerswellers acrossacross slumsslums areare inin factfact rent-payingrent-paying tenants,tenants, aass sshownhown iinn TTable 4.able 4. IInn a ssurveyurvey ooff ccityity ooffiffi cialscials inin KiberaKibera (Marx,(Marx, Stoker,Stoker, andand SuriSuri 2013),2013), wewe foundfound tthathat tthehe twotwo mostmost commoncommon causescauses ofof landlord–tenantlandlord–tenant disputesdisputes inin thethe slumslum werewere thethe ttenant’senant’s inabilityinability toto paypay theirtheir rentrent andand rentrent increasesincreases askedasked byby thethe landlord.landlord. LookingLooking aatt amountsamounts paidpaid inin rentrent acrossacross differentdifferent incomeincome bracketsbrackets inin Kibera,Kibera, householdshouseholds inin tthehe poorestpoorest quantilesquantiles dodo notnot paypay lowerlower rentsrents perper squaresquare metermeter occupiedoccupied (as(as shownshown iinn TTable 5).able 5). TThishis fi ndingnding uunderminesndermines tthehe nnotionotion tthathat rruralural mmigrantsigrants ccanan ppayay a mmodestodest ppremiumremium ttoo livelive inin closeclose proximityproximity toto thethe city.city. AlthoughAlthough rentsrents forfor urbanurban poorpoor looklook rratherather lowlow inin amountamount (about(about 12 percent12 percent ofof consumption),consumption), mostmost ruralrural householdshouseholds (for(for eexample,xample, 90 percent90 percent inin Kenya)Kenya) dodo notnot paypay anyany rent.rent. InIn 2008,2008, thethe averageaverage monthlymonthly rentrent fforor ruralrural householdshouseholds waswas 266 Kenyan266 Kenyan shillingsshillings (about(about $3 US),$3 US), oror 1 percent1 percent ooff hhouse-ouse- hholdold cconsumption,onsumption, aass oopposedpposed ttoo 33,303 Kenyan,303 Kenyan sshillingshillings (($39 US),$39 US), oorr 110 percent0 percent 196 Journal of Economic Perspectives

Table 5 Rent Prices across Consumption Quintiles in Kibera (rent per month)

1st quintile 2nd quintile 3rd quintile 4th quintile 5th quintile

Area rented per capita 5.3 7.9 10.9 14.1 15.9 (square meters) Rent per capita 310.9 435.2 488.6 666.0 1121.4 (Kenyan shillings) Rent per square meter 117.6 107.5 96.8 109.5 127.0 (Kenyan shillings)

Source: Data from Marx, Stoker, and Suri (2013). Notes: We trim the top 1 percent of the data for rents and area rented. The exchange rate was $1 = 85 Kenyan shillings. When looking at rents per capita in these data, it is hard to account for economies of scale in housing. Part of housing is a public good to the members of a household, and these rent fi gures are simply total rent paid divided by household size, not accounting for these economies of scale.

ooff cconsumption,onsumption, fforor uurbanrban hhouseholdsouseholds ( JJackack aandnd SSuriuri fforthcoming).orthcoming). IInn tthehe KKiberaibera sslum,lum, wwherehere ffoodood eexpenditurexpenditure rrepresentsepresents 661 percent1 percent ooff cconsumption,onsumption, hhousingousing rrentsents rrepresentepresent iinn ffactact aalmostlmost a tthirdhird ooff nnonfoodonfood eexpenditure.xpenditure.6 A fourth setfourth set ofof factorsfactors thatthat cancan causecause lowlow investmentinvestment involveinvolve thethe extremeextreme coordi-coordi- nnationation failuresfailures andand ““governancegovernance gap”gap” iintrinsicntrinsic ttoo sslumlum llife.ife. WWidespreadidespread ggovernanceovernance ffailuresailures workwork aagainstgainst thethe prospectsprospects forfor thethe urbanurban poorpoor toto fi ndnd creativecreative solutionssolutions toto uupgradepgrade thethe qualityquality ofof theirtheir neighborhoodsneighborhoods (as(as envisionedenvisioned inin TurnerTurner andand FichterFichter 11972).972). A largelarge amountamount ofof anecdotalanecdotal evidenceevidence suggestssuggests thatthat allocationallocation mechanismsmechanisms iinn sslumslums aarere iineffineffi cientcient andand thatthat privateprivate actorsactors oror bureaucraticbureaucratic entrepreneursentrepreneurs fi llll tthehe ggovernanceovernance space,space, asas opposedopposed toto legitimatelegitimate locallocal governmentsgovernments oror communitycommunity rrepresentatives.epresentatives. ForFor example,example, landland andand housinghousing marketsmarkets areare oftenoften controlledcontrolled byby a handfulhandful ofof powerfulpowerful oror well-well- connectedconnected individuals:individuals: landlords,landlords, locallocal bureaucrats,bureaucrats, oorr ganggang members.members. DavisDavis (2006)(2006) reportedreported onon thethe exampleexample ofof Mumbai,Mumbai, wherewhere itit isis cclaimedlaimed thatthat 991 individuals1 individuals controlcontrol allall vacantvacant land.land.7 TThehe slumsslums ofof Nairobi,Nairobi, wherewhere llandand permitspermits onon vacantvacant landland havehave beenbeen illegallyillegally awardedawarded byby thethe locallocal administra-administra- ttionion sincesince thethe 1970s,1970s, areare a posterposter case.case. TheThe NairobiNairobi “slumlords”“slumlords” cancan oftenoften relyrely onon tthehe ssupportupport ofof thethe locallocal administrationadministration toto settlesettle rentrent disputesdisputes ( JoiremanJoireman 2011),2011), aandnd theythey maymay ccolludeollude wwithith llocalocal cchiefshiefs ttoo ddiscourageiscourage improvementsimprovements inin thethe housinghousing iinfrastructurenfrastructure thatthat couldcould leadlead toto moremore eentrenchedntrenched tenancytenancy rights.rights. InIn areasareas wwherehere cchiefshiefs areare notnot ableable toto enforceenforce theirtheir authority,authority, gangsgangs sometimessometimes fi llll thethe governancegovernance

6 Across the developing world, poor households tend to spend a large share of their income on food, as discussed in this journal by Banerjee and Dufl o (2007). 7 Owning and renting out structures in slums can be immensely profi table to landowners: in Nairobi, Amis (1984) reported that annual returns on the housing capital stock in slums could be as high as 131 percent. Benjamin Marx, Thomas Stoker, and Tavneet Suri 197

sspacepace ttoo enforceenforce rulesrules ofof theirtheir own,own, levylevy taxes,taxes, andand controlcontrol expenditureexpenditure andand invest-invest- mmentsents inin theirtheir neighborhoodsneighborhoods (Marx,(Marx, Stoker,Stoker, andand SuriSuri 2013).2013). InIn otherother areas,areas, thethe fformalormal ggovernanceovernance systemsystem isis eentirelyntirely aabsentbsent andand hhasas bbeeneen rreplacedeplaced byby thesethese ootherther iinterests—annterests—an exampleexample isis thethe rolerole ofof drugdrug cartelscartels inin thethe ooff RioRio dede JaneiroJaneiro ((FerrazFerraz aandnd OOttonittoni 22013).013).8 A fi fthfth potentialpotential contributorcontributor toto lowlow investmentinvestment trapstraps inin slumsslums comescomes fromfrom thethe wwell-knownell-known TTodaroodaro paradoxparadox (1976):(1976): slumslum livingliving standardsstandards cannotcannot bebe improvedimproved wwithoutithout ggeneratingenerating aann aadditionaldditional iinflnfl uxux ofof ruralrural migrants,migrants, whichwhich inin turnturn depressesdepresses ppublicublic andand privateprivate investmentsinvestments inin thethe existingexisting settlements.settlements. ThisThis maymay givegive littlelittle iincentivencentive fforor thethe publicpublic sectorsector toto investinvest inin infrastructureinfrastructure andand publicpublic goodsgoods inin slums.slums. HHowever,owever, therethere isis littlelittle rigorousrigorous evidenceevidence onon thethe linklink betweenbetween thesethese “pull”“pull” factorsfactors andand sslumlum growthgrowth inin thethe developingdeveloping world,world, whilewhile “push”“push” factorsfactors (such(such asas overcrowdingovercrowding onon tthehe ffertileertile llandsands aandnd aann ooverallverall ddeclineecline iinn aagriculturalgricultural ooutput)utput) hhaveave bbeeneen bbetteretter ddocu-ocu- mmented.ented. FForor instance,instance, LiptonLipton (1977)(1977) andand BatesBates (1981)(1981) arguedargued thatthat thethe “urban“urban bias”bias” ooff policypolicy andand tax-basedtax-based incomeincome transferstransfers betweenbetween peasantspeasants andand citycity dwellersdwellers untiluntil thethe 11990s990s waswas a chiefchief causecause ofof rural–urbanrural–urban migration.migration. TheThe seminalseminal modelmodel onon thethe issueissue ooff rural–urbanrural–urban migrationmigration waswas thatthat ofof HarrisHarris andand TodaroTodaro (1970),(1970), whowho modeledmodeled thethe rrural–urbanural–urban wagewage gapgap asas a drivingdriving forceforce behindbehind migrationmigration decisions.decisions. However,However, thisthis wworkork hadhad littlelittle toto saysay aboutabout locationslocations decisionsdecisions ofof migrantsmigrants withinwithin cities,cities, andand wewe areare nnotot aawareware ooff aanyny mmoreore rrecentecent ttheoreticalheoretical aattemptttempt ttoo mmodelodel tthosehose llocationocation cchoices.hoices.9

The Policy Trap IInn aadditionddition ttoo bbeingeing ttrappedrapped iinn llow-human-capitalow-human-capital aandnd llow-investmentow-investment eequilib-quilib- rriums,iums, sslumlum aareasreas aarere ggenerallyenerally pplaceslaces ooff eextremextreme ppolicyolicy nneglect,eglect, wellwell bbeyondeyond thethe llackack ofof publicpublic goodsgoods provisionprovision discusseddiscussed above.above. ThisThis “policy“policy trap”trap” stemsstems fromfrom polit-polit- iicalcal eeconomyconomy ffactorsactors tthathat aarere ddifferentifferent ffromrom tthehe mmarketarket ffailuresailures wwee jjustust ddiscussed.iscussed. FFirst,irst, tthehe informalinformal naturenature ofof slumslum neighborhoodsneighborhoods impliesimplies thatthat thesethese areasareas areare uusuallysually cconsideredonsidered notnot eeligibleligible fforor uurbanrban pplanninglanning oorr ppublicublic uupgradingpgrading pprojects.rojects. TThishis mmayay bbee fforor ppurelyurely aadministrativedministrative rreasonseasons oorr bbecauseecause ppublicublic iinvestmentsnvestments ccouldould aamountmount ttoo mmoreore eentrenchedntrenched ooccupancyccupancy rrightsights fforor tthehe sslumlum rresidentsesidents — aann ooutcomeutcome tthathat ggovernmentsovernments ggenerallyenerally ddoo nnotot ffavoravor ((FoxFox 22013).013). OOverver tthehe ppastast ttwo decades,wo decades, tthehe ffewew countriescountries thatthat mademade signifisignifi cantcant advancesadvances inin thethe strugglestruggle againstagainst slumslum growthgrowth wwereere thosethose wherewhere politicalpolitical supportsupport waswas widespreadwidespread forfor reducingreducing thethe prevalenceprevalence ofof sslumslums andand wherewhere a genuinegenuine politicalpolitical commitmentcommitment waswas expressedexpressed forfor curbingcurbing slumslum eexpansion—forxpansion—for eexample,xample, iinn EEgyptgypt aandnd MMexicoexico ((UN-HabitatUN-Habitat 22006).006). SSecond,econd, enumerationenumeration problemsproblems andand thethe factfact thatthat slumslum populationspopulations areare oftenoften ((deliberatelydeliberately oror mmistakenly)istakenly) uundercountedndercounted distortsdistorts thethe weightweight assignedassigned toto sslumlum aareasreas

8 Ferraz and Ottoni (2013) study the effects of a pacifi cation program in the Rio favelas where military interventions were necessary to re-establish a police presence in the slums. 9 For US cities, one canonical modeling approach to looking at location decisions in urban areas is to think of a city with a central business district and how housing, commerce, and schools are organized relative to that city center, along the lines of Alonso (1964), Muth (1969), and Mills (1972). But these sorts of models do not seem an apt description of the location decisions and outcomes in the cities of low-income countries. 198 Journal of Economic Perspectives

iinn thethe politicalpolitical process.process. ForFor example,example, SabrySabry ((2010)2010) showedshowed howhow thethe undercountingundercounting ooff ashwa’iyyat ((informalinformal settlement)settlement) populationspopulations ofof GreaterGreater CairoCairo ledled toto anan under-under- ssamplingampling ooff sslumlum hhouseholdsouseholds iinn hhouseholdousehold ssurveys.urveys. IInn IIndia,ndia, sslumlum ppopulationsopulations wwereere ccomprehensivelyomprehensively enumeratedenumerated forfor tthehe fi rstrst timetime inin 2001,2001, butbut discrepanciesdiscrepancies inin thethe sstate-leveltate-level ddefiefi nitionsnitions ofof slumsslums andand thethe refusalrefusal ofof somesome statesstates toto validatevalidate thethe slumslum sstatisticstatistics rresultedesulted iinn ““grossgross uunder-estimation/under-coveragender-estimation/under-coverage ooff sslumlum ppopulationsopulations iinn tthehe ccountry”ountry” ((GovernmentGovernment ooff IIndiandia 22011).011). LLackack ooff rrepresentationepresentation ccanan hhaveave ddramaticramatic cconsequencesonsequences whenwhen issuesissues ofof evictioneviction areare atat stake.stake. FForor eexample,xample, tthehe MMakokoakoko nneigh-eigh- bborhoodorhood inin Lagos,Lagos, Nigeria,Nigeria, oneone ofof thethe oldestoldest slumsslums iinn tthehe wworldorld untiluntil itsits ppartialartial ddemolitionemolition inin 2012,2012, hadhad notnot beenbeen coveredcovered byby thethe country’scountry’s lastlast nationalnational censuscensus inin 22007007 ((BabalolaBabalola 2009).2009). NotNot havinghaving accurateaccurate censuscensus datadata onon slumsslums isis problematicproblematic forfor mmanyany rreasons.easons. TThehe ttruerue ppopulationopulation ooff tthehe sslumlum iiss uunknown—fornknown—for eexample,xample, ttherehere hhasas bbeeneen controversycontroversy overover thethe populationpopulation ofof thethe KiberaKibera slum,slum, withwith estimatesestimates rangingranging ffromrom aaboutbout 1170,00070,000 ttoo ooverver oone million—andne million—and ppolicyolicy iinterventionsnterventions aarere iimpossiblempossible ttoo pplanlan wwithoutithout aaccurateccurate ppopulationopulation numbers.numbers.1100 TThird,hird, cateringcatering toto thethe interestsinterests ofof thethe silentsilent majoritymajority ooff sslumlum ddwellerswellers mmightight nnotot eevenven bbee iinn tthehe bbestest iinterestnterest ooff tthehe ppeopleeople iinn cchargeharge iinn tthehe sslum.lum. AAss ddiscussediscussed aabove,bove, pplanninglanning oror regulatoryregulatory ppowersowers iinn sslumslums ooftenften ddoo nnotot bbelongelong ttoo llegitimateegitimate governancegovernance bbodies,odies, butbut areare usuallyusually splitsplit betweenbetween a galaxygalaxy ofof privateprivate actors,actors, landlords,landlords, chiefschiefs aandnd bbureaucrats,ureaucrats, andand gangs.gangs. ConflConfl ictingicting interestsinterests betweenbetween thesethese actors,actors, andand policypolicy cconflonfl ictsicts betweenbetween centralcentral governmentgovernment andand municipalmunicipal authoritiesauthorities couldcould explainexplain whywhy ““statusstatus qquo”uo” iinterestsnterests havehave oftenoften pprevailedrevailed inin slumsslums ((FoxFox 22013).013). IInn ootherther wwords,ords, mmaintainingaintaining highhigh ttransactionransaction costscosts aandnd oopaquepaque ggovernanceovernance mmechanismsechanisms ccanan bbee vveryery bbenefienefi cialcial toto a minorityminority ofof bureaucraticbureaucratic entrepreneursentrepreneurs willingwilling toto garnergarner supportsupport inin sslumlum ppatronageatronage ppoliticsolitics oorr ttoo eextractxtract rrentsents ffromrom ttheirheir iinformalnformal ccontrolontrol ooverver tthehe llandand ((AmisAmis 1984;1984; Marx,Marx, Stoker,Stoker, andand SuriSuri 2013).2013). InIn lineline withwith this,this, FoxFox (2013)(2013) documentsdocuments hhowow mmembersembers ooff TTanzania’sanzania’s rrulinguling Chama Cha Mapinduzi ((CMM)CMM) PPartyarty aarere iinvolvednvolved iinn ttransactionsransactions onon slumslum landland markets.markets. Syagga,Syagga, Mitullah,Mitullah, andand Karirah-GitauKarirah-Gitau (2002)(2002) ffoundound tthathat 557 percent7 percent ooff llandlordsandlords iinn NNairobiairobi slumsslums wwereere ppublicublic eemployees.mployees.

Slums and Economic Development

Slum Growth in a Cross-Country Perspective TThehe cconceptualizationonceptualization ofof slumsslums asas pplaceslaces ofof povertypoverty trapstraps isis aatt ooddsdds wwithith a ““modernization”modernization” view,view, wwhichhich aassumesssumes tthathat tthehe pprevalencerevalence ooff sslumslums aandnd uurbanrban ppovertyoverty sshouldhould decreasedecrease asas marketsmarkets developdevelop andand thethe forcesforces ofof economiceconomic developmentdevelopment ccomeome uundernder way.way. HereHere wewe presentpresent somesome simplesimple empiricalempirical factsfacts onon thisthis hypothesis.hypothesis. IInn particular,particular, isis therethere a functionalfunctional relationshiprelationship betweenbetween economiceconomic growth,growth, urbanurban

10 Aside from the policy constraints that lack of data poses, the poor census implies that the sampling frame for any survey (national or specifi cally of the slum) cannot draw a representative random sample. It is therefore much harder to accurately collect socioeconomic indicators of conditions in a slum and understand how these conditions evolve over time. The Economics of Slums in the Developing World 199

Figure 1 Patterns in Urban and Slum Growth, 1990 –2007

120

100

80

60

40

20

0

–20 Egypt Mexico India Indonesia China Bangladesh Pakistan Nigeria

–40

–60

–80 % increase in total urban population % increase in slum population % increase in per capita GDP divided by 10

Source: Figure compiled by the authors using data from UN-Habitat’s Global Urban Indicators online database (UN-Habitat 2013). The data on GDP per capita come from the World Bank’s World Development Indicators online database. Notes: Figure 1 compares slum growth and urban growth in the ten countries that had more than 10 million slum households in 1990. The countries are ranked by the percentage growth in their slum population from 1990 to 2007, shown by the bars. The dotted line shows for each country in the sequence the percentage increase in GDP per capita from 1990 –2007 divided by 10 (to fi t on the same scale). ggrowth,rowth, andand thethe prevalenceprevalence ofof slums?slums? IsIs therethere evidenceevidence thatthat standardsstandards ofof livingliving areare iimprovingmproving wwithinithin sslums,lums, aand/ornd/or aacrosscross ggenerationsenerations ooff sslumlum ddwellers?wellers? OOverver thethe pastpast 20 years,20 years, countriescountries thatthat experiencedexperienced fastfast economiceconomic growthgrowth areare aalsolso thethe onesones thatthat achievedachieved thethe mostmost signifisignifi cantcant reductionsreductions inin thethe proportionproportion ofof uurbanrban householdshouseholds livingliving inin slums.slums. IInn a ccross-countryross-country rregressionegression fframework,ramework, AArimahrimah ((2010)2010) foundfound thatthat thethe prevalenceprevalence ofof slumsslums inin anyany givengiven countrycountry waswas signifisignifi cantlycantly ccorrelatedorrelated withwith a varietyvariety ofof aggregateaggregate economiceconomic indicators,indicators, includingincluding GDPGDP perper ccapitaapita ((negatively),negatively), thethe ddebtebt sstocktock aandnd ddebtebt sservice,ervice, aandnd iinequalitynequality mmeasuredeasured bbyy tthehe GGiniini ccoeffioeffi cientcient (positively).(positively). However,However, cross-countrycross-country correlationscorrelations overlookoverlook widelywidely hheterogeneouseterogeneous situations,situations, asas rapidrapid urbanizationurbanization ratesrates inin developingdeveloping countriescountries areare ooftenften notnot aassociatedssociated withwith fastfast economiceconomic growth.growth. InIn fact,fact, a nnumberumber ooff tthehe lleasteast ddevel-evel- oopedped countriescountries havehave experiencedexperienced a rapidrapid growthgrowth ofof theirtheir urbanurban populationpopulation withoutwithout eexperiencingxperiencing muchmuch economiceconomic growthgrowth atat aall.ll. EExtremextreme rruralural ppoverty,overty, nnaturalatural ddisasters,isasters, aandnd ccivilivil wwarsars hhaveave bbeeneen tthehe mmainain ddriversrivers ooff tthishis ““urbanizationurbanization wwithoutithout ggrowth.”rowth.” AAnn eexamplexample iiss ggiveniven bbyy tthehe DDemocraticemocratic RRepublicepublic ofof thethe Congo,Congo, wherewhere thethe populationpopulation ooff tthehe ccountry’sountry’s ccapitalapital KinshasaKinshasa moremore thanthan tripledtripled inin sizesize betweenbetween thethe beginningbeginning ofof tthehe MMobutuobutu rregimeegime ((1965)1965) aandnd tthehe eendnd ooff tthehe SSecondecond CCongoongo WWarar ((2002).2002). FFigure 1igure 1 ccomparesompares sslumlum ggrowthrowth aandnd uurbanrban ggrowthrowth iinn tthehe ttenen ccountriesountries tthathat hhadad moremore tthanhan 110 million0 million sslumlum hhouseholdsouseholds iinn 11990.990. TThehe ccountriesountries aarere rrankedanked 200 Journal of Economic Perspectives

bbyy thethe percentagepercentage ggrowthrowth iinn ttheirheir sslumlum ppopulationopulation ffromrom 11990990 ttoo 22007,007, sshownhown bbyy tthehe bars.bars. TThesehese bbarsars ccanan bbee ccomparedompared ttoo ooverallverall uurbanrban ppopulationopulation ggrowth,rowth, sshownhown bbyy tthehe ddarkark ssolidolid lline.ine. A ffewew ccountriesountries eexperiencedxperienced eexplosivexplosive uurbanrban ggrowthrowth wwithith limitedlimited oror nono slumslum expansionexpansion (for(for instance,instance, India,India, Indonesia,Indonesia, andand Brazil),Brazil), wwhereashereas iinn oothersthers ((PakistanPakistan aandnd NNigeria),igeria), sslumlum ggrowthrowth aaccountsccounts fforor mmostost ooff uurbanrban growth.growth. TheThe countriescountries wherewhere urbanurban populationpopulation ggrowthrowth ooutstrippedutstripped sslumlum ppopulationopulation ggrowthrowth hhaveave a ddecliningeclining ssharehare ooff uurbanrban ppopulationopulation llivingiving iinn sslums:lums: bbetweenetween 11990990 aandnd 22009,009, tthishis pproportionroportion ddecreasedecreased ffromrom 5555 ttoo 229 9 p percentercent iinn IIndia,ndia, fromfrom 4444 ttoo 229 percent9 percent inin China,China, andand fromfrom 5151 ttoo 23 percent23 percent inin IndonesiaIndonesia ((UN-Habitat 2012a).UN-Habitat 2012a). TThehe dotteddotted lineline inin Figure Figure 1 showsshows forfor eacheach countrycountry inin thethe sequence,sequence, thethe ppercentageercentage increaseincrease inin GDPGDP perper capitacapita overover 1990–20071990–2007 divided by 10 ((toto fi t onon tthehe samesame sscale).cale). VVeryery roughly,roughly, perper capitacapita growthgrowth inin GDPGDP waswas similarsimilar betweenbetween thethe setset ooff countriescountries wherewhere slumslum populationspopulations fellfell (Egypt,(Egypt, Mexico,Mexico, andand Indonesia),Indonesia), andand thethe ssetet ooff ccountriesountries wwherehere mmostost ooff uurbanrban ggrowthrowth wwasas sslumlum ppopulationsopulations ((thethe PPhilippines,hilippines, PPakistan,akistan, andand Nigeria).Nigeria). InIn fact,fact, itit appearsappears thatthat thethe connectionconnection betweenbetween economiceconomic ggrowthrowth andand slumslum growthgrowth acrossacross countriescountries isis quitequite diverse,diverse, withoutwithout a uniformuniform ppattern.attern. Below,Below, wewe discussdiscuss howhow differentdifferent policypolicy choiceschoices maymay havehave contributedcontributed toto tthesehese hheterogeneouseterogeneous eexperiences.xperiences.

An Intergenerational Perspective IInn mmanyany sslumslums iinn llow-incomeow-income ccountries,ountries, livingliving standardsstandards dodo notnot seemseem toto bebe iimprovingmproving ooverver ttime.ime. IInn KKibera,ibera, KKenya,enya, ccensusensus ddataata ssuggestuggest tthathat llivingiving cconditionsonditions hhaveave eithereither deteriorateddeteriorated oror atat bestbest stagnatedstagnated overover thethe 1999–20091999–2009 period,period, a periodperiod dduringuring whichwhich thethe economyeconomy asas a wholewhole grewgrew atat 3.5 percent3.5 percent perper annum.annum. TheThe shareshare ooff householdhousehold headsheads withwith a primaryprimary educationeducation fellfell fromfrom 4747 toto 40 percent40 percent overover thisthis ttime;ime; tthehe nnumberumber ooff rroomsooms pperer ccapitaapita hheldeld eessentiallyssentially tthehe ssameame aatt 00.68.68 iinn 11999,999, aandnd 00.67.67 inin 2009;2009; andand thethe shareshare ofof thosethose usingusing a pitpit latrinelatrine (rather(rather thanthan thethe mainmain sewersewer ssystem)ystem) ffellell oonlynly sslightly,lightly, ffromrom 882 percent2 percent iinn 11999999 ttoo 777 percent7 percent bbyy 22009.009. OOff ccourse,ourse, tthishis ssimpleimple aanalysisnalysis ooff llivingiving sstandardstandards ooverver a ggiveniven pperioderiod oover-ver- llooksooks a ffundamentalundamental sselectionelection pproblem:roblem: hhouseholdsouseholds tthathat iimprovedmproved ttheirheir cconditionondition ooverver tthehe pperioderiod mmayay nnoo llongeronger lliveive iinn tthehe sslum,lum, wwhilehile ootherther ppooroor hhouseholdsouseholds mmayay hhaveave mmigratedigrated iintonto tthehe sslum.lum. TThehe aavailablevailable eevidence,vidence, ooverall,verall, ddoesoes nnotot pproviderovide prima facie eevidencevidence ofof rapidrapid changeschanges inin thethe compositioncomposition ofof slums.slums. InIn ourour KiberaKibera ddata,ata, respondentrespondent householdshouseholds havehave livedlived inin thethe slumslum forfor 16 16 yearsyears onon average,average, aandnd incomesincomes dodo notnot increaseincrease (nor(nor decrease)decrease) withwith thethe durationduration ofof residency,residency, asas iillustratedllustrated inin Figure 2Figure 2 (panel A).(panel A). InIn thethe BangladeshiBangladeshi settlementssettlements studied,studied, incomeincome pperer capitacapita correlatescorrelates negativelynegatively andand signifisignifi ccantlyantly withwith thethe totaltotal numbernumber ofof yearsyears tthathat thatthat householdhousehold hashas spentspent inin thethe slumslum andand withwith thethe numbernumber ofof yearsyears sincesince thethe hhouseholdsouseholds fi rstrst leftleft the countrysidethe countryside (Figure (Figure 2,2, panel panel B).B). Similarly,Similarly, UN-HabitatUN-Habitat ((2003)2003) reportedreported thatthat 41 percent41 percent ofof KolkataKolkata slumslum dwellersdwellers hadhad livedlived inin slumsslums forfor ooverver 30 years,30 years, andand moremore thanthan 70 percent70 percent hadhad livedlived inin slumsslums forfor overover 15 years.15 years. InIn a ccasease studystudy ofof BangkokBangkok slums,slums, 60 percent60 percent ofof individualsindividuals werewere reportedlyreportedly bornborn inin thethe ssameame slumslum (UN-Habitat(UN-Habitat 2003).2003). Benjamin Marx, Thomas Stoker, and Tavneet Suri 201

Figure 2 Living Standards and Duration of Residency in Slums of Bangladesh and Kenya

A: Kenya

300

200

100 Monthly consumption per capita in US$ (2013)

0 0 20406080 Years of residency in the Kibera slum B: Bangladesh

25

20

15 Monthly income

per capita in US$ (2000) 10

5 0 20406080 Number of years since the household left rural area

Notes: The fi rst scatterplot shows the result of a single variable regression of monthly consumption per capita on the number of years the household has lived in the slum from our Kibera survey data. The estimated slope is – 0.017 (SE 0.15). In this dataset the average monthly consumption is USD 64. The second scatterplot presents the result of a regression of monthly income per capita (defi ned as the total household income in US$ (non PPP) divided by the household size) on the number of years since the household fi rst left rural areas for Tongi and Jessore slums in Bangladesh, controlling for the age and education of the household head, the number of adults in the household, and district fi xed effects. The average monthly consumption is US$15 (US$21 in 2012 equivalent). The slope of the fi tted line is the coeffi cient of interest in that regression, and the estimated value of this coeffi cient is – 0.055 (SE: 0.019). The negative sign of the slope is robust to removing these controls and to the inclusion of more controls. 202 Journal of Economic Perspectives

TToo oourur knowledge,knowledge, PerlmanPerlman (2010)(2010) isis oneone ofof thethe fewfew studiesstudies toto havehave attemptedattempted ttoo aaddressddress sselectionelection iissuesssues iinn ssurveysurveys ooff sslumslums bbyy ttracingracing sslumlum rrespondentsespondents aandnd ttheirheir ddescendantsescendants overover anan extendedextended periodperiod ofof time:time: 11969–2001,969–2001, inin thethe favelas ooff RRioio dede JJaneiro.aneiro. HerHer fi ndingsndings werewere somewhatsomewhat ambiguous.ambiguous. SheShe reportedreported thatthat a majoritymajority ofof iindividualsndividuals interviewedinterviewed inin 19691969 andand foundfound againagain inin 20012001 wwereere nnoo llongeronger llivingiving iinn tthehe favelas (63 percent(63 percent ofof thethe originaloriginal interviewees,interviewees, asas wellwell asas 64 percent64 percent ofof theirtheir cchildrenhildren andand 68 68 percentpercent ofof theirtheir grandchildren),grandchildren), andand thatthat thethe fractionfraction thatthat hadhad rremainedemained inin thethe favelasfavelas hadhad betterbetter accessaccess toto publicpublic servicesservices (including(including education)education) aass wellwell asas improvedimproved householdhousehold amenities.amenities. However,However, onlyonly a nonrandomnonrandom 41 percent41 percent ooff tthehe ooriginalriginal samplesample (307(307 outout ofof 750)750) couldcould bebe accuratelyaccurately relocated,relocated, ofof whichwhich 222 percent2 percent werewere stillstill alivealive andand 19 percent19 percent werewere deceased.deceased. InIn addition,addition, fromfrom a newnew rrandomandom ssampleample ooff 4425 individuals25 individuals iinn tthehe ssameame ssurveyurvey aareas,reas, sshehe ffoundound cconsiderablyonsiderably hhigherigher unemploymentunemployment ratesrates (51 percent,(51 percent, upup fromfrom 332 percent2 percent iinn 11969)969) aandnd a hhigherigher ffractionraction ooff rrespondentsespondents wwithith nnoo iincomencome ((23 percent,23 percent, uupp ffromrom 117 percent7 percent iinn 11969).969). OOnene qquestionuestion thatthat remainsremains isis whywhy householdshouseholds thatthat initiallyinitially migratedmigrated toto slumsslums bbutut ddidid notnot experienceexperience welfarewelfare gainsgains dodo notnot returnreturn toto theirtheir provinceprovince ofof origin.origin. A ddatasetataset oonn ttwo slumswo slums iinn NNairobiairobi ((APHRCAPHRC 22013)013)1111 tthathat collectscollects informationinformation onon wherewhere ooutmigrantsutmigrants gogo onceonce theythey leaveleave thethe slumslum providesprovides somesome answersanswers toto thatthat question.question. BBetweenetween 20032003 andand 2007,2007, 15 percent15 percent ofof slumslum residentsresidents movedmoved locations.locations. OfOf these,these, 226 percent6 percent movedmoved toto anotheranother slum,slum, 4 percent4 percent movedmoved toto anotheranother locationlocation inin thethe ssameame sslums,lums, 222 percent2 percent mmovedoved ttoo a nnonslumonslum aarearea iinn NNairobi,airobi, aandnd 441 percent1 percent mmovedoved ttoo ruralrural Kenya.Kenya. TheThe notionnotion ofof slumsslums beingbeing a povertypoverty traptrap impliesimplies thatthat householdshouseholds ddoo nnotot nnecessarilyecessarily movemove ttherehere pplanninglanning ttoo sstay,tay, bbutut iinsteadnstead ggetet ccaughtaught iinn a llow-levelow-level eequilibrium.quilibrium. TheThe APHRCAPHRC datadata doesdoes speakspeak toto this—verythis—very fewfew slumslum residentsresidents moved,moved, oonlynly aaboutbout 115 percent,5 percent, aandnd oone-thirdne-third ooff tthesehese sstayedtayed inin slumslum aareas.reas. EEmpirically,mpirically, llittleittle iiss kknownnown aboutabout thethe outsideoutside optionsoptions thatthat slumslum householdshouseholds havehave oorr ccontemplateontemplate atat anyany givengiven time.time. AlthoughAlthough therethere isis a smallsmall probabilityprobability ofof successsuccess ((forfor example,example, ofof fi ndingnding consistentconsistent employment),employment), perhapsperhaps thethe expectedexpected gainsgains areare llargearge enoughenough toto makemake thisthis decisiondecision individuallyindividually rational.rational. A relatedrelated literatureliterature isis tthathat oonn eentrepreneurship—manyntrepreneurship—many sstudiestudies fi ndnd thatthat entrepreneurshipentrepreneurship doesdoes notnot paypay ((HamiltonHamilton 2000;2000; MMoskowitzoskowitz andand Vissing-JørgensenVissing-Jørgensen 2002),2002), butbut somesome ofof thisthis maymay bebe ddueue toto nonpecuniarynonpecuniary benefibenefi ttss ooff bbeingeing sself-employedelf-employed (Hamilton(Hamilton 2000)2000) oror perhapsperhaps ttoo thethe overconfioverconfi dencedence thatthat entrepreneursentrepreneurs havehave inin theirtheir ownown skillsskills (Bernardo(Bernardo andand WWelchelch 2001).2001). InIn thethe literatureliterature onon education,education, researchersresearchers alsoalso foundfound thatthat informa-informa- ttionion pprovidedrovided ttoo pparentsarents oonn tthehe rreturnseturns ttoo eeducationducation ccanan hhaveave eeffectsffects oonn eenrollmentnrollment rratesates ooff ttheirheir cchildrenhildren ( JensenJensen 2010;2010; NguyenNguyen 2008).2008). IssuesIssues ofof overconfioverconfi dencedence aandnd misinformationmisinformation aboutabout individualindividual successsuccess probabilitiesprobabilities maymay alsoalso bebe atat stakestake inin sslums—therelums—there isis rroomoom fforor rresearchesearch oonn eexpectationsxpectations aandnd ttheirheir rroleole iinn mmigrationigration ddeci-eci- ssions,ions, eespeciallyspecially iinn tthehe ccontextontext ooff sslumlum ppopulations.opulations.

11 This data covers two slums, and Viwandani, which form a Demographic Surveillance Site in Nairobi. The data is collected by the African Population and Health Research Center. A subset of the data is available at http://www.aphrc.org/. The Economics of Slums in the Developing World 203

Limitations of Past Approaches in Slum Policy

RRiddingidding ccitiesities ooff sslumslums mmayay bbee cconsideredonsidered aann eessentialssential ppartart ooff tthehe ddevelopmentevelopment pprocess;rocess; yyetet tthishis hhasas pprovedroved nnearlyearly impossibleimpossible forfor policymakerspolicymakers inin mostmost emergingemerging aandnd developingdeveloping economies.economies. InIn a recentrecent testimonytestimony ofof thethe problemsproblems encounteredencountered bbyy aambitiousmbitious slumslum ppolicy,olicy, tthehe ggovernmentovernment ofof IndiaIndia announcedannounced inin 20092009 thethe imple-imple- mmentationentation ofof thethe Rajiv Awas Yojana ((RAY)RAY) schemescheme toto makemake thethe countrycountry “slum-free”“slum-free” wwithinithin fi vve years.e years. Three monthsThree months later,later, thethe timeframetimeframe ofof thethe programprogram waswas extendedextended ttoo sseven even y yearsears (Aggarwal(Aggarwal 2009).2009). InIn 2011,2011, India’sIndia’s CommitteeCommittee onon SlumSlum StatisticsStatistics eestimatedstimated thatthat thethe totaltotal slumslum populationpopulation inin thethe countrycountry wouldwould stillstill increaseincrease byby 112 percent2 percent bbetweenetween 20112011 aandnd 22017017 ((GovernmentGovernment ooff IIndiandia 22011).011). BByy tthehe eendnd ooff 22012,012, tthehe sschemecheme wwasas sstilltill iinn iitsts iinfancy,nfancy, aandnd eearlyarly iimplementationmplementation hhadad bbeeneen hhamperedampered bbyy pproblemsroblems wwithith llandand rrecords,ecords, ssiteite sselection,election, aandnd tthehe aallocationllocation ooff llandand ((KunduKundu 22012).012). IInn tthishis ssection,ection, wwee ddiscussiscuss wwhathat ppolicyolicy aapproachespproaches towardstowards slumsslums hhaveave bbeeneen ttakenaken aandnd wwhyhy tthesehese aapproachespproaches havehave beenbeen largelylargely unsuccessful.unsuccessful.

From “Benign Neglect” to “Aided Self-help” HHistorically,istorically, outrightoutright evictionsevictions havehave beenbeen usedused asas a primaryprimary policypolicy instrumentinstrument ttoo reducereduce slumslum populations.populations. A highlyhighly publicizedpublicized exampleexample ofof a large-scalelarge-scale slumslum cclearancelearance schemescheme waswas “Operation“Operation CleanClean Up”Up” implementedimplemented inin 20052005 inin Harare,Harare, ZZimbabwe,imbabwe, wherewhere 700,000 individuals700,000 individuals lostlost theirtheir homeshomes (Tibaijuka(Tibaijuka 2005).2005). WhileWhile littlelittle ddataata hashas beenbeen collectedcollected onon slumslum householdshouseholds evictedevicted asas partpart ofof thesethese policies,policies, itit isis qquiteuite clearclear thatthat slumslum cclearancelearance doesdoes notnot addressaddress thethe rootsroots ofof thethe slumslum problem.problem. HHenceence a popularpopular alternativealternative toto clearance,clearance, widelywidely adoptedadopted inin thethe 1960s1960s andand 1970s,1970s, wwasas tthehe ddeliberateeliberate neglectneglect ofof expandingexpanding slumslum areas.areas. AsAs partpart ofof tthishis aapproach,pproach, therethere wwereere nnoo ppolicingolicing ofof squatters,squatters, butbut alsoalso nono provisionprovision ofof publicpublic servicesservices inin informalinformal ssettlements.ettlements. InIn tthehe ““benign”benign” iinterpretation,nterpretation, policymakerspolicymakers assumedassumed thatthat thethe mmarketarket wwouldould “take“take carecare ofof it”:it”: slumsslums providedprovided much-neededmuch-needed low-costlow-cost housinghousing toto urbanurban ddwellers,wellers, wwhoho wwouldould eventuallyeventually movemove iintonto fformalormal hhousing.ousing. BByy tthehe 11970s,970s, hhowever,owever, iitt bbecameecame cclearlear tthathat nneithereither sslumlum cclearancelearance nnoror ““benignbenign neglect”neglect” wouldwould addressaddress thethe continuouscontinuous expansionexpansion ofof slums.slums. SlumSlum inhab-inhab- iitantstants wwereere notnot beingbeing pulledpulled awayaway byby improvedimproved economiceconomic prospectsprospects nornor beingbeing ppushedushed awayaway fromfrom thethe less-desirableless-desirable aaspectsspects ooff uunregulatednregulated sslumlum lliving.iving. A nnewew tthinkinghinking eemergedmerged tthathat tthehe uurbanrban ppooroor wwouldould fi nndd ccreativereative ssolutionsolutions ttoo iimprovemprove ttheirheir livelihoodslivelihoods asas longlong asas basicbasic improvementsimprovements ttoo thethe locallocal environmentenvironment ccouldould bbee performedperformed byby thethe government.government. RatherRather thanthan resettlingresettling thethe squatters,squatters, govern-govern- mmentsents shouldshould focusfocus theirtheir effortsefforts onon providingproviding basicbasic infrastructureinfrastructure andand improvingimproving ssanitaryanitary cconditions—likeonditions—like ssupplyingupplying ssafeafe wwaterater aandnd ffacilitatingacilitating wwasteaste ddisposal.isposal. UUponpon ccompletionompletion ooff tthesehese basicbasic improvements,improvements, tthehe sslumlum rresidentsesidents wwouldould sstarttart iinvestingnvesting inin theirtheir ownown dwellings,dwellings, andand improvementsimprovements inin livingliving standardsstandards wouldwould ffollowollow suit.suit. ThisThis “aided“aided self-help”self-help” paradigm,paradigm, inspiredinspired byby earlierearlier experiencesexperiences inin EEuropeurope aandnd tthehe UUSSRSSR ((HarrisHarris 11999)999) aandnd bbyy tthehe iinflnfl uentialuential workwork ooff BBritishritish aarchi-rchi- ttectect JohnJohn TTurnerurner ((TurnerTurner aandnd FFichterichter 11972),972), cconvincedonvinced tthehe WWorldorld BBankank aandnd ootherther ppolicyolicy plannersplanners toto reorientreorient theirtheir policiespolicies towardstowards a “slum“”upgrading” approach.approach. 204 Journal of Economic Perspectives

CComparedompared ttoo eevictionsvictions oorr ““benignbenign nneglect,”eglect,” sslumlum uupgradingpgrading sseemedeemed aatt tthehe ttimeime ttoo presentpresent ggreatreat aadvantages.dvantages. FFirst,irst, iitt wwasas vveryery ccheap:heap: a llargearge uupgradingpgrading pprojectroject iinn JJakartaakarta costcost UUS$38S$38 ttoo UUS$120S$120 pperer hhousehold,ousehold, aass oopposedpposed ttoo tthehe ccostost ooff bbuildinguilding eentirelyntirely newnew hhousingousing uunitsnits fforor tthehe sslumlum ddwellerswellers ((WerlinWerlin 11999).999). SSecond,econd, iitt wwouldould bbee endorsedendorsed bbyy tthehe iinhabitants,nhabitants, aass llocalocal sstakeholderstakeholders aandnd ccommunitiesommunities wwouldould vviewiew tthemselveshemselves asas receivingreceiving aann iimprovedmproved sstandardtandard ooff llivingiving aandnd wwouldould bbee iinvolvednvolved iinn tthehe maintenancemaintenance ooff tthehe nnewew iinfrastructure.nfrastructure. BByy tthehe eearlyarly 11980s,980s, sslumlum uupgradingpgrading hhadad bbeeneen iincludedncluded iinn nnumerousumerous ppovertyoverty aalleviationlleviation programsprograms acrossacross thethe ddevelopingeveloping world.world. EarlyEarly eevaluationvaluation resultsresults fromfrom uupgradingpgrading projectsprojects conductedconducted inin Kolkata,Kolkata, Jakarta,Jakarta, andand ManilaManila seemedseemed promising:promising: fforor instance,instance, mortalitymortality causedcaused byby waterbornewaterborne diseasedisease waswas halvedhalved amongamong benefibenefi cia-cia- rriesies inin KKolkata,olkata, andand investmentsinvestments inin homehome iimprovementsmprovements wwereere ddoubledoubled inin JJakartaakarta ((WerlinWerlin 1999).1999). UpgradingUpgrading programsprograms alsoalso seemedseemed toto increaseincrease thethe housinghousing supplysupply aandnd thethe supplysupply ofof laborlabor byby householdshouseholds (Keare(Keare andand ParrisParris 1982).1982). However,However, thethe earlyearly eenthusiasmnthusiasm fforor tthehe uupgradingpgrading aapproachpproach bbeganegan toto ddwindlewindle bbyy tthehe llateate 11980s980s aass sslumlum aareasreas ccontinuedontinued ttoo eexpand,xpand, aandnd tthehe bbasicasic iinfrastructurenfrastructure iimprovementsmprovements aappearedppeared ttoo bbee uunsustainable.nsustainable. TheThe mmaintenanceaintenance ssystemsystems ooff tthehe uupgradedpgraded ppublicublic ggoodsoods ccollapsedollapsed aandnd eenvironmentalnvironmental andand healthhealth issuesissues werewere onon thethe rise.rise. InIn Jakarta,Jakarta, wherewhere allall slumslum aareasreas hadhad benefibenefi tedted fromfrom thethe upgradingupgrading program,program, 93 percent93 percent ofof thethe city’scity’s wellswells werewere ccontaminatedontaminated withwith fecesfeces (Werlin(Werlin 1999).1999). Overall,Overall, eveneven thoughthough upgradingupgrading programsprograms wwereere rarelyrarely submittedsubmitted ttoo rrigorousigorous eevaluation,valuation,1122 thethe availableavailable evidenceevidence suggestedsuggested thatthat tthehe upgradingupgrading projectsprojects ofof anan “aided“aided self-help”self-help” approachapproach diddid notnot seemseem anan adequateadequate ppolicyolicy lleverever ttoo ttransformransform sslumlum cconditionsonditions iinn a mmeaningfuleaningful wway.ay.

The Land Titling Paradigm TThehe writingswritings ofof HernandoHernando dede SotoSoto (2000),(2000), despitedespite somesome earlyearly criticismcriticism fromfrom tthehe eeconomicsconomics professionprofession (Woodruff(Woodruff 2001)2001) werewere inflinfl uentialuential inin refocusingrefocusing thethe slumslum ddebateebate ttowardsowards iissuesssues ooff llandand ttenureenure aandnd llandand rrights.ights. DDee SSotooto aarguedrgued tthathat ggivingiving tthehe ppooroor propertyproperty titlestitles overover theirtheir landland wouldwould provideprovide collateralcollateral forfor millionsmillions ofof poorpoor uurbanrban householdshouseholds acrossacross thethe developingdeveloping world.world. InIn dede Soto’sSoto’s argument,argument, propertyproperty rrightsights wwereere thethe panacea,panacea, andand encouragingencouraging slumslum dwellersdwellers toto investinvest inin improvingimproving ttheirheir homeshomes waswas stillstill viewedviewed asas thethe HolyHoly GrailGrail ofof slumslum policy.policy. TheThe keykey driverdriver ofof thisthis eencouragementncouragement wouldwould nono longerlonger bebe thethe provisionprovision ofof publicpublic goodsgoods byby thethe govern-govern- mment,ent, bbutut rratherather a rreductioneduction inin thethe riskrisk ofof eviction,eviction, resultingresulting fromfrom thethe titlingtitling ofof landland ooccupiedccupied bbyy ssquatters.quatters. TThus,hus, hhomeome iinvestmentsnvestments wwouldould bbecomeecome ssaferafer fforor ppooroor hhouse-ouse- hholds,olds, andand sslumlum hhouseholdsouseholds wwouldould becomebecome ableable toto accessaccess creditcredit marketsmarkets toto fi nancenance iinvestmentsnvestments toto ccreatereate smallsmall bbusinessesusinesses aandnd eeducateducate ttheirheir cchildren—thehildren—the uultimateltimate eenginesngines ofof ppovertyoverty rreduction.eduction. LandLand ttitlingitling wwouldould cconcomitantlyoncomitantly increaseincrease thethe llocalocal ttaxax bbasease aandnd eenablenable mmunicipalitiesunicipalities ttoo iimprovemprove tthehe pprovisionrovision ooff bbasicasic ppublicublic ggoods.oods.

12 Slum upgrading programs until now rarely provided a natural experiment framework that could isolate causal effects (Field and Kremer 2006). A recent exception is the work of Galiani et al. (2013), who provide causal estimates of the impact of housing upgrading projects in slums of three Latin American countries. Benjamin Marx, Thomas Stoker, and Tavneet Suri 205

IInn tthehe ppastast 220 years,0 years, nnationalational ggovernmentsovernments aandnd tthehe WWorldorld BBankank iimplementedmplemented uurbanrban llandand ttitlingitling projectsprojects inin moremore thanthan 18 African,18 African, Asian,Asian, andand LatinLatin AmericanAmerican countriescountries ((Durand-Lasserve,Durand-Lasserve, FFernandes,ernandes, PPayne,ayne, aandnd RakodiRakodi 2007).2007). TThehe positivepositive relationshiprelationship betweenbetween tenuretenure securitysecurity andand investmentsinvestments inin landland hashas bbeeneen wellwell documenteddocumented inin thethe empiricalempirical literatureliterature forfor ruralrural settingssettings (for(for example,example, BBanerjee,anerjee, Gertler,Gertler, andand GhatakGhatak 2002;2002; BesleyBesley 1995;1995; GoldsteinGoldstein andand UdryUdry 2008;2008; HHornbeckornbeck 2010).2010). However,However, fewfew academicacademic studiesstudies havehave lookedlooked atat thethe impactimpact ofof urbanurban ttitlingitling programsprograms onon thethe investmentinvestment decisionsdecisions mademade bbyy hhouseholdsouseholds inin slums.slums. FForor uurbanrban householdshouseholds inin Peru,Peru, FieldField (2005)(2005) foundfound thatthat landland titlingtitling increasesincreases thethe raterate ooff housinghousing renovationsrenovations (by(by aboutabout two-thirds),two-thirds), andand FieldField (2007)(2007) showedshowed thatthat titlingtitling iincreasedncreased householdhousehold laborlabor supplysupply (by(by 1010 toto 15 percent)15 percent) byby freeingfreeing resourcesresources thatthat wwereere ppreviouslyreviously uusedsed ttoo pprotectrotect hhouseholdousehold aassets.ssets. GGalianialiani aandnd SSchargrodskychargrodsky (2010)(2010) sshowedhowed thatthat thethe allocationallocation ofof formalformal landland titlestitles inin BuenosBuenos AiresAires ledled toto increasedincreased iinvestmentsnvestments aandnd eeducationducation aamongstmongst hhouseholdsouseholds wwhoho bbenefienefi tedted fromfrom thethe titling.titling. OnOn tthehe policypolicy side,side, however,however, byby 20052005 concernsconcerns beganbegan toto emergeemerge aboutabout howhow urbanurban landland ttitlingitling pprogramsrograms wwereere bbeingeing wwidelyidely ppromotedromoted wwithoutithout mmuchuch iiff aanyny eevidencevidence oonn ttheirheir eeffectivenessffectiveness ((Durand-Lasserve,Durand-Lasserve, Fernandes,Fernandes, Payne,Payne, andand RakodiRakodi 2007).2007). TThehe majormajor limitationlimitation ofof thethe landland titlingtitling argumentargument isis thatthat itit assumesassumes thatthat a lacklack ooff fformalormal titlestitles impliesimplies weakweak oror nonexistentnonexistent propertyproperty rights.rights. However,However, therethere isis nono ssystematicystematic evidenceevidence thatthat landland rrightsights aarere aalwayslways wweaklyeakly enforcedenforced inin slums.slums. AsAs oonene ccounterexample,ounterexample, LanjouwLanjouw andand LevyLevy (2002)(2002) havehave arguedargued thatthat informalinformal rightsrights cancan eeffectivelyffectively substitutesubstitute forfor formalformal titlestitles inin slumslum settlementssettlements inin Ecuador.Ecuador. InIn Kibera,Kibera, wwherehere allall vacantvacant landland waswas formallyformally rreclaimedeclaimed byby thethe KenyanKenyan governmentgovernment inin 1969,1969, tthehe mmembersembers ooff oonene eethnicthnic ggrouproup sstilltill cclaimlaim llandand rrightsights bbasedased oonn ttenancyenancy ppermitsermits aallocatedllocated byby thethe BritishBritish colonialcolonial authoritiesauthorities inin thethe earlyearly twentiethtwentieth century.century. TheThe ffactact thatthat mostmost individualsindividuals recognizedrecognized asas landlordslandlords livelive outsideoutside thethe slumslum (Syagga,(Syagga, MMitullah,itullah, andand Karirah-GitauKarirah-Gitau 2002)2002) alsoalso impliesimplies thatthat theirtheir informalinformal rightsrights overover thethe rrentersenters aarere sstronglytrongly eenforced.nforced. IInn a similarsimilar vein,vein, analyzinganalyzing thethe impactimpact ofof twotwo titlingtitling projectsprojects inin SenegalSenegal andand SSouthouth Africa,Africa, Payne,Payne, Durand-Lasserve,Durand-Lasserve, andand RakodiRakodi (2008)(2008) foundfound thatthat “residents“residents iinn mmostost iinformalnformal ssettlementsettlements iinn bbothoth [[SenegalSenegal aandnd SSouthouth AAfrica]frica] aalreadylready eenjoynjoy de facto ttenureenure security.”security.” HHenceence thethe impactimpact ofof titlingtitling projectsprojects onon individualsindividuals whowho aalreadylready enjoyenjoy tenuretenure securitysecurity is,is, fromfrom thethe start,start, uncertain:uncertain: PalmerPalmer (1998)(1998) pointedpointed ooutut howhow thethe effectivenesseffectiveness ofof landland titlingtitling programsprograms actuallyactually dependeddepended onon whetherwhether ttheyhey couldcould achieveachieve anan increaseincrease inin thethe totaltotal securitysecurity thatthat poorpoor householdshouseholds enjoy.enjoy. DDurand-Lasserve,urand-Lasserve, Fernandes,Fernandes, Payne,Payne, andand RakodiRakodi (2007)(2007) reviewedreviewed evidenceevidence fromfrom llandand titlingtitling projectsprojects thatthat maymay hhaveave aactuallyctually decreased ttenureenure securitysecurity (because(because theythey aallowedllowed fforor llawfullyawfully eenforcednforced eevictions)victions) inin AAfghanistan,fghanistan, Cambodia,Cambodia, EEgypt,gypt, IIndia,ndia, aandnd Rwanda.Rwanda. TitlingTitling pprogramsrograms aalonelone ccannotannot bbee eexpectedxpected toto liftlift householdshouseholds outout ofof ppovertyoverty andand toto overhauloverhaul existingexisting socialsocial andand eeconomicconomic dynamicsdynamics withinwithin thethe slum,slum, bbecauseecause existingexisting systemssystems ofof ownershipownership actact toto preservepreserve thesethese dynamics.dynamics. InIn fact,fact, landland ttitlingitling isis moremore llikelyikely ttoo bbenefienefi t thethe “slumlords”“slumlords” (whose(whose informalinformal ownershipownership rightsrights aarere oftenoften well-recognizedwell-recognized locally)locally) andand hurt,hurt, atat thethe bottombottom ofof thethe pyramid,pyramid, thethe slumslum rrenters,enters, eithereither inin thethe fformorm ooff ooutrightutright eevictionsvictions oorr iincreasedncreased rrentsents iinn tthehe ttitleditled aarea.rea. 206 Journal of Economic Perspectives

TThesehese featuresfeatures havehave beenbeen documenteddocumented inin Senegal,Senegal, forfor instanceinstance (Payne,(Payne, Durand-Durand- LLasserve,asserve, aandnd RRakodi 2008).akodi 2008). UUltimately,ltimately, iindividualndividual aapproachespproaches suchsuch asas upgradingupgrading andand formalformal titlingtitling havehave llargelyargely failedfailed toto iimprovemprove llivelihoodsivelihoods inin sslums.lums. RRecognizingecognizing tthathat tthehe eeffectffect ooff ttitlingitling pprojectsrojects andand conventionalconventional landland administrationadministration systemssystems hadhad beenbeen limited,limited, UUN-HabitatN-Habitat (2012c)(2012c) recentlyrecently advocatedadvocated a “continuum“continuum ofof landland rights”rights” relyingrelying onon pparticipatoryarticipatory llandand eenumerationnumeration andand record-keepingrecord-keeping toto improveimprove tenuretenure securitysecurity fforor thethe urbanurban poor.poor. ThisThis wouldwould ideallyideally bebe includedincluded inin a moremore holisticholistic approachapproach ofof sslumlum ppolicy.olicy. CCountriesountries thatthat managedmanaged toto curbcurb thethe growthgrowth ofof slums,slums, suchsuch asas BrazilBrazil oorr Egypt,Egypt, indeedindeed appearappear toto bebe thosethose wherewhere slumslum policypolicy reliedrelied onon a combinationcombination ofof iinstrumentsnstruments ——includingincluding eeffortsfforts ttoo increaseincrease thethe transparencytransparency andand effieffi ciencyciency ofof landland mmarkets,arkets, ttoo iimprovemprove llocalocal ggovernance,overnance, ttoo iincreasencrease ppublicublic iinvestmentsnvestments mmassively,assively, aandnd ttoo iincreasencrease tthehe ssupplyupply ooff ccheapheap hhousingousing ((UN-HabitatUN-Habitat 22010a).010a).

Conclusion

SSlumslums representrepresent a majormajor policypolicy challengechallenge forfor developingdeveloping economieseconomies inin thethe ttwenty-fiwenty-fi rstrst century.century. AddingAdding migrantsmigrants toto thethe existingexisting slumslum populations,populations, UN-HabitatUN-Habitat ((2012b)2012b) eestimatedstimated tthathat 4450 million50 million nnewew hhousingousing uunitsnits wwouldould bbee nneededeeded iinn tthehe nnextext 220 years0 years justjust toto accommodateaccommodate householdshouseholds inin urgenturgent needneed ofof housing.housing. YetYet thethe chal-chal- llengeenge ofof slumsslums isis notnot simplysimply oneone ofof housinghousing policy:policy: a holisticholistic approachapproach isis neededneeded toto aaddressddress hhousingousing nneedseeds forfor rruralural mmigrants,igrants, hhealthealth aandnd ssanitationanitation issues,issues, locallocal ggover-over- nnance,ance, privateprivate savingssavings andand investments,investments, andand landland marketmarket institutions.institutions. BothBoth formalformal aandnd informalinformal systemssystems ofof propertyproperty rightsrights maymay bebe necessarynecessary toto curbcurb thethe rapidrapid growthgrowth ooff slumslum aareasreas wworldwide.orldwide. InIn thethe absenceabsence ofof strongstrong policypolicy agendasagendas similarsimilar toto thosethose aadopteddopted inin SingaporeSingapore or,or, moremore recently,recently, inin Brazil,Brazil, itit seemsseems unlikelyunlikely thatthat slumsslums willwill ddisappearisappear inin thethe foreseeableforeseeable future,future, asas implicitlyimplicitly assumedassumed byby a modernizationmodernization viewview ooff tthehe iissue.ssue. OOverall,verall, ttherehere hhasas bbeeneen vveryery llittleittle ttheoreticalheoretical aandnd eempiricalmpirical eeconomicconomic rresearchesearch aaboutbout hhowow tthehe ppublicublic ppolicyolicy cchallengeshallenges pposedosed bbyy sslumslums iinn llow-incomeow-income eeconomiesconomies sshouldhould bbee aaddressed.ddressed. A rresearchesearch aagendagenda oonn sslumslums ccouldould ffocusocus oonn tthree hree d distinctistinct ssetsets ooff mmethodologicalethodological aandnd ppolicyolicy qquestions.uestions. FFirst,irst, tthehe mmethodologicalethodological pproblemsroblems tthathat hhamperamper fi eldeld rresearchesearch iinn sslumslums sshouldhould bbee aaddressed.ddressed. IInn pparticular,articular, eeffortsfforts ttoo eenumeratenumerate sslumlum ppopulationsopulations aandnd ttoo ttrackrack ppanelanel rrespondentsespondents ooverver sseveraleveral ggenera-enera- ttionsions ooff sslumlum ddwellerswellers ccouldould bbee ssteppedtepped uup,p, aandnd eempiricalmpirical mmethodsethods uusedsed ttoo ddealeal wwithith ssurveyurvey aattritionttrition iinn ootherther ccontextsontexts ((Fitzgerald,Fitzgerald, GGottschalk,ottschalk, aandnd MMoffioffi tttt 11998)998) sshouldhould bbee mmoreore cconsistentlyonsistently aapplied.pplied. TThishis wwouldould aallowllow fforor a bbetteretter uunderstandingnderstanding ooff tthehe mmostost ppressingressing iissuesssues ffacedaced bbyy sslumlum ddwellerswellers aandnd a bbetteretter iintegrationntegration ooff tthesehese ddwellerswellers iinn nnationalational ppoliticalolitical pprocesses.rocesses. RRelatedly,elatedly, ttherehere hhasas bbeeneen llittleittle eeffortffort ttoo ssystematicallyystematically sstudytudy tthehe mmovementsovements iintonto aandnd ooutut ooff sslums,lums, oorr ttoo ccollectollect ddataata ttoo ttrackrack tthehe iindividualsndividuals wwhoho ddoo mmove,ove, eevenven iiff tthosehose eexitingxiting aarere ffewew aandnd ffarar bbetween.etween. SSimilarly,imilarly, uunderstandingnderstanding tthehe iintergenerationalntergenerational ccorrelationorrelation iinn iincomesncomes aandnd ootherther ssocioeconomicocioeconomic ooutcomesutcomes wwouldould bbee aann iimportantmportant ccontribution.ontribution. The Economics of Slums in the Developing World 207

SSecond,econd, thethe returnsreturns toto upgradingupgrading differentdifferent typestypes ofof publicpublic servicesservices shouldshould bebe iidentifidentifi eded andand quantifiquantifi ed,ed, soso thatthat cost-effectivecost-effective projectsprojects andand programsprograms thatthat provideprovide ddiscernableiscernable welfarewelfare gainsgains forfor slumslum residentsresidents cancan bebe moremore consistentlyconsistently applied.applied. TheThe AAbdulbdul LutifLutif JameelJameel PPovertyoverty ActionAction LabLab (J-PAL)(J-PAL) UrbanUrban ServicesServices InitiativeInitiative (USI)(USI) hhasas bbegunegun ttoo ppromoteromote ssuchuch aann aagendagenda ((DuflDufl o,o, Galiani,Galiani, andand MobarakMobarak 2012).2012). TThird,hird, ggiveniven tthehe sstabilitytability iinn sslumslums ddocumentedocumented aabovebove aandnd tthehe nnotionotion tthathat sslumslums mmayay bbee povertypoverty trapstraps ofof somesome sort,sort, perhapsperhaps oneone policypolicy directiondirection wouldwould bebe thethe “big“big ppush.”ush.” SSeveraleveral pprogramsrograms wwithith mmassass iinvestmentsnvestments oorr wwholesaleholesale rrelocationelocation ooff sslumlum hhouse-ouse- hholdsolds iintonto hhousingousing eestatesstates aappearppear ttoo hhaveave bbeeneen successful.successful. HHowever,owever, iinn mmanyany currentcurrent sslumslums iinn thethe developingdeveloping world,world, thisthis cannotcannot bebe donedone withoutwithout a politicalpolitical willingnesswillingness toto cchangehange ggovernanceovernance ddynamicsynamics iinn sslumlum aareasreas oor,r, iinn pparticular,articular, ttoo ddealeal wwithith tthehe aactorsctors wwhoho hhaveave ttakenaken ooverver tthehe ggovernanceovernance ooff tthesehese aareasreas iinn tthehe aabsencebsence ooff tthehe ggovernment.overnment. TThehe ggovernanceovernance iissuesssues mmayay pperhapserhaps bbee tthehe mmostost ppressingressing ssinceince tthesehese iinformalnformal aactorsctors hhaveave a sstrongtrong ppresenceresence iinn tthehe sslumslums aandnd hhaveave llargearge rrentsents aatt sstaketake sshouldhould ttherehere bbee aanyny cchange.hange. TTheyhey hhaveave sstrongtrong iincentivesncentives ttoo mmaintainaintain tthehe sstatustatus qquo.uo. WWithoutithout cchangeshanges ttoo tthesehese iinsti-nsti- ttutionsutions aandnd a rreversaleversal ooff tthehe llackack ooff ggovernance,overnance, iitt iiss uunlikelynlikely tthathat aanyny aattemptsttempts aatt aanyny fformorm ooff bbigig ppushush oorr ccoordinatedoordinated iinvestmentnvestment wwillill hhaveave tthehe ddesiredesired eeffects.ffects.

■ The authors would like to thank David Autor, Chang-Tai Hsieh, Ulrike Malmendier, and Timothy Taylor for comments that greatly improved the paper. We thank the MIT Sloan School of Management for funding the Kibera surveys used in this paper. In addition, the authors acknowledge use of publicly available data for Bangladesh from the Supporting Household Activities for Health, Assets and Revenue Surveys; for India, from the National Family Health Survey (NFSH-3); and for Hyderabad, from the data repository of the Abdul Latif Jameel Poverty Action Lab ( J-PAL). The authors are extremely grateful to the Kenya National Bureau of Statistics for access to their 1999 and 2009 Population Censuses and UNICEF for access to the 2010 UNICEF SMART Survey for Sierra Leone.

References

Aggarwal, Smita. 2009. “Slum-free India Deadline and Demographic Surveillance Site) dataset. http:// Now 7 Years.” The Indian Express, August 28. http:// www.aphrc.org/ (accessed February 15, 2013). www.indianexpress.com/news/slumfree-india Arimah, Ben C. 2010. “The Face of Urban -deadline-now-7-yrs/508138/. Poverty. Explaining the Prevalence of Slums in Alonso, William. 1964. Location and Land Use: Developing Countries.” UNU-WIDER Working Toward a General Theory of Land Rent. Cambridge, Paper no. 2010/30. MA: Harvard University Press. Azariadis, Costas, and Allan Drazen. 1990. Amis, Philip. 1984. “Squatters or Tenants: The “Threshold Externalities in Economic Develop- Commercialization of Unauthorized Housing in ment.” Quarterly Journal of Economics 105(2): 501–26. Nairobi.” World Development 12(1): 87– 96. Babalola, Yemi. “Makoko Residents and Their APHRC. 2013. NUHDSS (Nairobi Urban Health Unwanted Guest.” Africa News Service, May 1, 2009. 208 Journal of Economic Perspectives

Banerjee, Abhijit V. and Esther Dufl o. 2007. Dufl o, Esther, Sebastian Galiani, and Mushfi q “The Economic Lives of the Poor.” Journal of Mobarak. 2012. “Improving Access to Urban Economic Perspectives 21(1): 141– 68. Services for the Poor: Open Issues and a Frame- Banerjee, Abhijit, Esther Dufl o, Rachel work for Future Research Agenda.” J-PAL Urban Glennerster, and Cynthia Kinnan. 2010. “The Services Review Paper. Cambridge, MA: Abdul Miracle of Microfi nance? Evidence from a Random- Latif Jameel Poverty Action Lab. http://www ized Evaluation.” BREAD Working Paper 278. .povertyactionlab.org/publication/improving Banerjee, Abhijit, Paul Gertler, and Maitreesh -access-urban-services-poor. Ghatak. 2002. “Empowerment and Effi ciency: Durand-Lasserve, Alain, Edesio Fernandes, Tenancy Reform in West Bengal.” Journal of Political Geoffrey Payne, and Carole Rakodi. 2007. Economy 110(2): 239 – 80. “Social and Economic Impacts of Land Titling Banerjee, Abhijit, Rohini Pande, Yashas Vaidya, Programmes in Urban and Peri-urban Areas: Michael Walton, and Jeff Weaver. 2011. “Delhi’s A Review of the Literature.” Paper presented Slum-Dwellers: Deprivation, Preferences and at the World Bank Urban Research Sympo- Political Engagement among the Urban Poor.” sium, Washington DC, May 2007. http://www International Growth Centre Conference proceed- .bhamlive3.bham.ac.uk/Documents/college ings (Growth Week 2011). -social-sciences/government-society/idd Basu, Kaushik. 2003. Analytical Development /research/social-economic-impacts/social Economics: The Less Developed Economy Revisited. -economic-impacts-literature-review.pdf. Cambridge, MA: MIT Press. El-Zanaty, Fatma, and Ann A. Way. 2004. Greater Bates, Robert H. 1981. Markets and States in Cairo Slums: A Profi le Based on the 2003 Egypt Demo- Tropical Africa: The Political Basis of Agricultural graphic and Health Survey. http://www.measuredhs Policies. Berkeley, CA: University of California .com/pubs/pdf/FA43/FA43.pdf. Press. Ferraz, Claudio, and Bruno Ottoni. 2013. “State Bernardo, Antonio E., and Ivo Welch. 2001. “On Presence and Urban Violence: Evidence from the the Evolution of Overconfi dence and Entrepre- Pacifi cation of Rio’s Favelas.” Unpublished paper. neurs.” Journal of Economics & Management Strategy Field, Erica. 2005. “Property Rights and Invest- 10(3): 301–330. ment in Urban Slums.” Journal of the European Besley, Timothy. 1995. “Property Rights and Economic Association 3(2–3): 279 –90. Investment Incentives: Theory and Evidence Field, Erica. 2007. “Entitled to Work: Urban from Ghana.” Journal of Political Economy 103(5): Tenure Security and Labor Supply in Peru.” Quar- 903 –937. terly Journal of Economics 122(4): 1561–1602. Bleakley, Hoyt. 2010. “Health, Human Capital, Field, Erica, and Michael Kremer. 2006. “Impact and Development.” Annual Review of Economics 2: Evaluation for Slum Upgrading Interventions.” 283 –310. World Bank, Doing Impact Evaluation Series No. 3. Bowles, Samuel, Steven N. Durlauf, and Karla Fitzgerald, John, Peter Gottschalk, and Robert Hoff, eds. 2006. Poverty Traps. Princeton University Moffi tt. 1998. “An Analysis of Sample Attrition in Press. Panel Data: The Michigan Panel Study of Income Bradley, David, Carolyn Stephens, Trudy Dynamics.” Journal of Human Resources 33(2): 251– 99. Harpham, and Sandy Cairncross. 1992. “A Fox, Sean. 2013. “The Political Economy of Review of the Environmental Health Impacts in Slums: Theory and Evidence from Sub-Saharan Developing Country Cities.” World Bank Urban Africa.” LSE International Development Work- Management Program Discussion Paper 11101. ing Paper No. 13.146. http://www.lse.ac.uk Chowdhury, Shyamal, Ahmed Mushfi q /internationalDevelopment/pdf/WP/WP146.pdf. Mobarak, and Gharad Bryan. 2009. “Migrating Frankenhoff, Charles A. 1967. “Elements of Away from a Seasonal Famine: A Randomized an Economic Model for Slums in a Developing Intervention in Bangladesh.” Human Develop- Economy.” Economic Development and Cultural ment Research Paper 2009/41. http://hdr.undp Change 16(1): 27–36. .org/en/reports/global/hdr2009/papers Galiani, Sebastian, Paul Gertler, Ryan Cooper, /HDRP_2009_41.pdf Sebastian Martinez, Adam Ross, and Raimundo Currie, Janet, and Tom Vogl. 2013. “Early-Life Undurraga. 2013. “Shelter from the Storm: Health and Adult Circumstance in Developing Upgrading Housing Infrastructure in Latin Countries.” Annual Review of Economics 5: 1– 36. American Slums.” NBER Working Paper 19322. Davis, Mike. 2006. Planet of Slums. Verso. Galiani, Sebastian, and Ernesto Schargrodsky. de Soto, Hernando. 2000. The Mystery of Capital: 2010. “Property Rights for the Poor: Effects of Why Capitalism Triumphs in the West and Fails Every- Land Titling.” Journal of Public Economics 94(9–10): where Else. Basic Books. 700–729. Benjamin Marx, Thomas Stoker, and Tavneet Suri 209

Glaeser, Edward. 2011. Triumph of the City: How Joireman, Sandra F. 2011. Where There is No Our Greatest Invention Makes Us Richer, Smarter, Government: Enforcing Property Rights in Common Greener, Healthier, and Happier. New York, NY: Law Africa. New York, NY: Oxford University Press. Penguin Press. Keare, Douglas H., and Scott Parris. 1982. Goldstein, Markus, and Christopher Udry. “Evaluation of Shelter Programs for the Urban 2008. “The Profi ts of Power: Land Rights and Agri- Poor: Principal Findings.” World Bank Staff cultural Investment in Ghana.” Journal of Political Working Papers 547. Economy 116(6): 981–1022. Kundu, Amitabh. 2012. “The Challenges of Golgher, André Braz. 2012. “The Selectivity Making India Slum-free.” India at LSE blog. of Migration and Poverty Traps in Rural Brazil.” http://blogs.lse.ac.uk/indiaatlse/2012/10/24 Population Review 51(1): 9–27. /the-challenges-of-making-indian-cities-slum-free Government of India, Ministry of Housing and -part-1/. Urban Poverty Alleviation. 2011. “Report of the Lanjouw, Jean O., and Philip I. Levy. 2002. Committee on Slum Statistics/Census.” http:// “Untitled: A Study of Formal and Informal Prop- mhupa.gov.in/W_new/Slum_Report_NBO.pdf. erty Rights in Urban Ecuador.” Economic Journal Gulyani, Sumila, Ellen M. Bassett, and 112(482): 986 –1019. Debabrata Talukdar. 2012. “Living Conditions, Lipton, Michael. 1977. Why Poor People Stay Poor: Rents, and Their Determinants in the Slums of Urban Bias in World Development. Cambridge, MA: Nairobi and Dakar.” Land Economics 88(2): 251–74. Harvard University Press. Gupta, Kamla, Fred Arnold, and H. Lhungdim. Lopez, Nestor. 2007. “Urban and Rural Dispari- 2009. Health and Living Conditions in Eight Indian ties in Latin America: Their Implications for Cities. National Family Health Survey (NFHS-3), Education Access.” Background paper prepared India, 2005–06 Mumbai: International Institute for the Education for All Global Monitoring for Population Sciences. http://pdf.usaid.gov/pdf Report 2008 titled Education for All by 2015: _docs/PNADQ634.pdf. Will We Make It? http://unesdoc.unesco.org Hamilton, Barton H. 2000. “Does Entrepre- /images/0015/001555/155540e.pdf. neurship Pay? An Empirical Analysis of the Returns López-Casasnovas, Guillem, Berta Rivera, to Self-Employment.” Journal of Political Economy and Luis Currais, eds. 2005. Health and Economic 108(3): 604 – 631. Growth: Findings and Policy Implications. Cambridge, Hannum, Emily, Meiyan Wang, and Jennifer MA: MIT Press. Adams. 2008. “Urban–Rural Disparities in Access to Marx, Benjamin, Thomas M. Stoker, and Primary and Secondary Education Under Market Tavneet Suri. 2013. “There is No Free House: Reform.” In One Country, Two Societies? Rural-Urban Ethnic Patronage and Property Rights in a Inequality in Contemporary China, edited by Martin Kenyan Slum.” http://www.mit.edu/~tavneet King Whyte. Cambridge, MA: Harvard University /Marx_Stoker_Suri.pdf. Press. Matsuyama, Kiminori. 2005. “Poverty Traps.” Harris, John R., and Michael P. Todaro. 1970. In The New Palgrave Dictionary of Economics, 2nd “Migration, Unemployment and Development: Edition, edited by L. Blume and S. Durlauf. A Two-Sector Analysis.” American Economic Review Palgrave Macmillan. 60(1): 126 –142. Miguel, Edward, and Michael Kremer. 2004. Harris, Richard. 1999. “Slipping through the “Worms: Identifying Impacts on Education and Cracks: The Origins of Aided Self-help Housing, Health in the Presence of Treatment Externali- 1918 – 53.” Housing Studies 14(3): 281–309. ties.” Econometrica 72(1): 159 –217. Hornbeck, Richard. 2010. “Barbed Wire: Mills, Edwin S. 1972. Studies in the Structure of Property Rights and Agricultural Development.” the Urban Economy. Baltimore, MD: Johns Hopkins Quarterly Journal of Economics 125(2): 767– 810. Press. Jack, William, and Tavneet Suri. Forthcoming. Moskowitz, Tobias, and Annette Vissing- “Risk Sharing and Transactions Costs: Evidence Jørgensen. 2002. “The Returns to Entrepreneurial from Kenya’s Mobile Money Revolution.” American Investment: A Private Equity Premium Puzzle?” Economic Review. American Economic Review 92(4): 745 –78. Jalan, Jyotsna, and Martin Ravallion. 2002. Murphy, Kevin M., Andrei Shleifer, and “Geographic Poverty Traps? A Micro Model of Robert W. Vishny. 1989. “Industrialization and Consumption Growth in Rural China.” Journal of the Big Push.” Journal of Political Economy 97(5): Applied Econometrics 17(4): 329 – 46. 1003 –1026. Jensen, Robert. 2010. “The (Perceived) Returns Muth, Richard F. 1969. Cities and Housing: to Education and the Demand for Schooling.” The Spatial Pattern of Urban Residential Land Use. Quarterly Journal of Economics 125(2): 515 – 48. University of Chicago Press. 210 Journal of Economic Perspectives

Nguyen, Trang. 2008. “Information, Role Models UN-Habitat. 2003. The Challenge of Slums: Global and Perceived Returns to Education: Experimental Report on Human Settlements 2003. London and Evidence from Madagascar.” January 23. http:// Sterling, VA: Earthscan Publications. www.povertyactionlab.org/sites/default/files UN-Habitat. 2006. State of the World’s Cities /documents/Nguyen%202008.pdf . 2006/2007. Nairobi: Human Palmer, David. 1998. “Security, Risk and Regis- Settlements Programme. tration.” Land Use Policy 15(1): 83 – 94. UN-Habitat. 2010a. State of the World’s Cities Payne, Geoffrey, Alain Durand-Lasserve, and 2010/2011. Nairobi: United Nations Human Settle- Carole Rakodi. 2008. “Social and Economic Impacts ments Programme. of Land Titling Programmes in Urban and Peri- UN-Habitat. 2010b. The Challenge of Slums: urban Areas: International Experience and Case Global Report on Human Settlements 2003. Revised Studies of Senegal and South Africa: Final Report.” and updated version (April 2010). http://www http://www.unrol.org/fi les/synthesis_report.pdf. .unhabitat.org/downloads/docs/GRHS_2003 Perlman, Janice. 2010. : Four Decades of _Chapter_01_Revised_2010.pdf. Living on the Edge of Rio de Janeiro. New York, NY: UN-Habitat. 2012a. State of the World’s Cities Oxford University Press. 2012/2013: Prosperity of Cities. Nairobi: United Sabry, Sarah. 2010. “How Poverty is Underesti- Nations Human Settlements Programme. mated in Greater Cairo, Egypt.” Environment and UN-Habitat. 2012b. Sustainable Housing for Urbanization 22(2): 523–41. Sustainable Cities: A Policy Framework for Developing Syagga, Paul, Winnie Mitullah, and Sarah Countries. Nairobi: United Nations Human Settle- Karirah-Gitau. 2002. “Nairobi Situation Analysis ments Programme. Supplementary Study: A Rapid Economic Appraisal UN-Habitat. 2012c. “Land Records for the Poor.” of Rents in Slums and Informal Settlements” Brief 1/2012. Nairobi: United Nations Human (draft). Contribution to the Preparatory Phase Settlements Programme. ( January–November 2002) of the Government UN-Habitat. 2013. Global Urban Indicators of Kenya & UN-HABITAT Collaborative Nairobi database. http://www.unhabitat.org/stats/Default Slum Upgrading Initiative. .aspx (accessed February 15, 2013). Tibaijuka, Anna Kajumulo. 2005. “Report of the US Census Bureau. 2013. “New York County Fact-Finding Mission to Zimbabwe to Assess the (Manhattan Borough), New York.” State & County Scope and Impact of Operation Murambatsvina QuickFacts. http://quickfacts.census.gov/qfd/states by the UN Special Envoy on Human Settle- /36/36061.html. ments Issues in Zimbabwe Mrs. Anna Kajumulo Werlin, Herbert. 1999. “The Slum Upgrading Tibaijuka.” United Nations. http://ww2.unhabitat Myth.” Urban Studies 36(9): 1523 – 34. .org/documents/ZimbabweReport.pdf. World Bank. 2009. The World Development Report Todaro, Michael P. 1976. “Urban Job Expansion, 2009: Reshaping Economic Geography. Washington Induced Migration, and Rising Unemployment: DC: The World Bank. A Formulation and Simplifi ed Empirical Test World Bank. 2013. “World Development for LDCs.” Journal of Development Economics 3(3): Indicators.” http://data.worldbank.org/data 211–25. -catalog/world-development-indicators (accessed Turner, John. 1969. “Uncontrolled Urban February 15, 2013). Settlement: Problems and Policies.” In The City in World Health Organization. 2013. “Global Newly Developing Countries: Readings on Urbanism Database on Child Growth and Malnutrition.” and Urbanization, edited by Gerald William Breese, http://www.who.int/nutgrowthdb/en/ (accessed 507– 534. Englewood Cliffs, NJ: Prentice Hall. February 15, 2013). Turner, John F. C., and Robert Fichter, eds. Woodruff, Christopher. 2001. “Review of de 1972. Freedom to Build: Dweller Control of the Housing Soto’s The Mystery of Capital.” Journal of Economic Process. Collier Macmillan Ltd. Literature 39(4): 1215 –23.