世界銀行-非洲在數(shù)字時代需要更多的道路嗎？基礎(chǔ)設(shè)施互補性的證據(jù) Does Africa Need More Roads in the Digital Age

PublicDisclosureAuthorizedPublicDisclosureAuthorized

PolicyResearchWorkingPaper10730

DoesAfricaNeedMoreRoadsintheDigitalAge?

EvidenceofComplementaritiesinInfrastructure

MathildeLebrand

ArcadyMongoue

RolandPongou

FanZhang

WORLDBANKGROUP

InfrastructureChiefEconomistOffice

March2024

PolicyResearchWorkingPaper10730

Abstract

Thispaperinvestigateswhethertheexpansionoffastinternetnetworkscomplementsorsubstitutesforthedevelopmentofroadstoimprovemarketaccessandcreatemoreandhigher-skilledjobsinAfrica.Thepapercombinesthegeo-graphiclocationsofhouseholdsandfirmswiththelocationsofmainroadsandoptical-fibernodesin25Sub-SaharanAfricancountries.Usingthedifference-in-differencesandinstrumentalvariablesapproachesandleveragingthehis-toryofpost-independenceroadbuildingandthetimingofthearrivalofsubmarineinternet,thepaperexaminestheimpactsofaccesstothesetwotypesofinfrastructure,

bothinisolationandincombination.Thefindingsshowthatimprovingaccesstobothhaslargeandpositivecom-plementaryeffects.Onaverage,theadditionalimpactsonemploymentfromcombiningaccesstobothtypesofinfra-structureare22percentlargerthanthesumoftheirisolatedeffects.ThefindingssuggestthatabigpushforcombinedinvestmentsinfastinternetandroadaccesscouldenhanceeconomicdevelopmentinAfricaoverall.Firmsandworkersinurbanlocations,femaleworkers,andworkerswithhigherlevelsofeducationgainthemostfromthecomplementar-itiesthatemerge.

ThispaperisaproductoftheInfrastructureChiefEconomistOffice.ItispartofalargereffortbytheWorldBanktoprovideopenaccesstoitsresearchandmakeacontributiontodevelopmentpolicydiscussionsaroundtheworld.PolicyResearchWorkingPapersarealsopostedontheWebat

/prwp.Theauthorsmaybecontacted

atmlebrand@.

ThePolicyResearchWorkingPaperSeriesdisseminatesthefindingsofworkinprogresstoencouragetheexchangeofideasaboutdevelopmentissues.Anobjectiveoftheseriesistogetthefindingsoutquickly,evenifthepresentationsarelessthanfullypolished.Thepaperscarrythenamesoftheauthorsandshouldbecitedaccordingly.Thefindings,interpretations,andconclusionsexpressedinthispaperareentirelythoseoftheauthors.TheydonotnecessarilyrepresenttheviewsoftheInternationalBankforReconstructionandDevelopment/WorldBankanditsaffiliatedorganizations,orthoseoftheExecutiveDirectorsoftheWorldBankorthegovernmentstheyrepresent.

ProducedbytheResearchSupportTeam

DoesAfricaNeedMoreRoadsintheDigitalAge?

EvidenceofComplementaritiesinInfrastructure

MathildeLebrand2,ArcadyMongoue1,RolandPongou1,andFan

Zhang2

1OttawaUniversity

2WorldBank

Keywords:Infrastructure;Employment;Structuraltransformation;Africa

JELClassification:J24,L91,L94,O13,O15,O18

2

1Introduction

Therapidspreadoftheinternethasbeenheraldedasanopportunityfortheeconomiesoflow-incomecountriesto“l(fā)eapfrogintothe21stcentury”andtocre-

atemoreandbetterjobs(HjortandPoulsen,

2019)

.Butitisalsoarguedthatunlessinvestmentsindigitaltechnologiesarecomplementedbypoliciesandmea-surestoimprovetraditionalmeansofmarketaccess,itisunlikelythatthesenew

technologieswillrealizetheirfullpotential(Goldberg,

2019)

.Severalpapersintheeconomicliteraturehaveexaminedtheroleofroadoneconomicgrowthandjobcreation,andafewstudieshaveanalyzedtheimpactofinternetinfrastructureontheseoutcomes.Amongothers,

Dufloetal.

(2015)showthattheproximityto

roadnetworkshascontributedtoeconomicgainswithinregionsinChina.

Hjort

andPoulsen

(2019)showthatfastinternetaccesshashadlarge,positiveeffects

onemploymentratesandongroupsofworkerswithlowerlevelsofeducation.However,mostoftheliteraturehasstudiedtheemploymentimpactsofdifferenttypesofinfrastructureinisolation,leavingopenthequestionofhowtheyworkincombination.Addressingthequestionofwhetherinternetnetworkscomplementorsubstituteforthedevelopmentofroadstoimprovelabormarketoutcomesandeconomicdevelopmentiscrucialfortheoptimaldesignofpoliciestosupplythese

costlyinfrastructures.

Inthispaper,weinvestigatecomplementaritiesbetweentheinternetandroadnetworksinemployment,thestructureofthelabormarket,andfirms’operationsinAfrica.Wealsoanalyzehowthese(complementary)effectsvaryacrossgender,skills,occupations,educationallevels,ruralandurbanlocations,andlevelsofeco-nomicdevelopment.Bothroadsandinternetaccessprovidebettermarketaccesstofirmsandconsumers.Internetconnectivitycansubstituteforaccesstoroads,forexample,byenablingsomepeopletoworkfromhomeratherthantouseroadstocommutetoanoffice.Alternatively,roadsandinternetaccesscanalsocom-plementeachother.Forexample,e-commerceinremoteareasislikelytodevelop

onlyifgoodroadsallowfortradetoandfromtheseregions.Internetaccessmay

3

reduceinformationfrictionsforfirms,allowingthemtofindandusebetterinputs,tofindandemploybetterworkers,andtotraintheirworkersonthejob;atthesametime,goodroadsmayremainnecessarytotakeadvantageoftheseinputs,bothtoallowworkerstoeasilycometoworkplaces,andforfirmstohaveincen-tivestoimproveproductionthroughinvestmentstoincreaselaborproductivity.OurpaperaimstoidentifywhetherroadsandinternetaccessaresubstitutesorcomplementsforjoboutcomesinAfrica.Addressingthisquestionhasimportantimplicationsforoptimizingtheallocationofinfrastructureinvestmentsinpoor

countries.

Toexaminehowtheexpansionoffastinternetnetworkscomplementsorsub-stitutesforthedevelopmentofroadstoencouragejobcreation,wecombinedataoncharacteristicsandgeographiclocationsofhouseholdsandfirms,andonthelo-cationofoptical-fibernodesandroadsinSub-SaharanAfricancountries.WedrawondetaileddemographicdatafromtheDemographicandHealthSurveys(DHS)Program,theWorldBankEnterprisesSurveys(WBES),andtheLivingStandardMeasurementStudy(LSMS).Noveldataonoptical-fibernodescomefromAfricaBandwidthMaps,andsubmarinecablemapsfrom

Mahlknecht

(2014)

.Usingacombinationofdifference-in-differencestechniquesandinstrumental-variableap-proacheswiththesevarieddata,weestimatethejoineffectsonlocallabormarketsandworkersthatstemfromchangesintheproximityofmajorroadsandaccessi-

bilitytohigh-speedinternetinfrastructure.

Wefindlarge,complementaryeffectsofinvestinginthetwotypesofinfras-tructure.Forexample,combinedinvestmentsofthiskindincreasetheprobabilityofhavingbeenemployedinthelastyearbyatleast9percentagepointsfromtheprobabilitythatisthecasewhennoinvestmentsaremade.Bycontrast,roadandinternetinvestmentsundertakeninisolationhavelowerimpacts.Being1kilo-meterclosertoaroadinagivenlocationincreasesthelikelihoodofemploymentby0.8percentagepoint,andhavingfastinternetaccessincreasesthelikelihoodofemploymentby6.7percentagepoints.Thus,theadditionalimpactonem-

ploymentthatarisesfromthecomplementaritiesofcombiningroadandinternet

4

investmentsis,onaverage,22percenthigher.Similarly,complementaritiesthatarisefromthetwoinvestmentsare20percentlargerintermsofboththeshareofagivenlocation’sworkerswhoareskilledandthesharewhoworkinservice-sectorjobs.Betteraccesstoroadsandfastinternetincombinationleadstheshareof

low-skilledworkerstofall.

Investmentsinbothtypesofinfrastructurealsohavepositiveimpactsonfirms.Wefindthatfirms’productivityimprovesandthatlossesdecline.Firmshavemoreopportunitiestoenterintocontractswiththegovernment.Thishasapositiveimpactonthelabordemandoffirms,leadingthemtooperateformorehoursandtohiremorefull-timeemployees.Onaverage,thecombinedimpactofaccessingfastinternetandbringingamainroad1kilometercloserincreasesfirms’hoursofoperationby4.4hoursperweekandleadsthemtoaddonemorefull-timeemployee.Combininginvestmentsinthetwotypesofinfrastructureratherthanmakingisolatedinvestmentsleadstoanincreaseinfirms’hoursofoperationand

inthenumberoffull-timeemployees.

Westudytheheterogeneityinimpactsoftheisolatedandcombinedinvest-mentsinexpandingaccesstofastinternetandmainroads.Theseimpactsdifferbetweenurbanandrurallocations,aswellasbetweenmaleandfemaleworkers.

Isolatedinvestmentsinfastinternetbenefiturbanlocationsandfemaleworkers

themost.Constructionthatbringsamainroad1kilometerclosertoalocalitybenefitsrurallocationsandfemaleworkersthemost.Theadditionalimpactsfromcombiningroadandinternetinvestmentsare72percentlargerinurbanlocationscomparedtolocalitiesthatgainedbetteraccesstoonlyoneofthetypesofin-frastructure;bycontrast,theadditionalimpactsofsuchdualinvestmentsarefarlower,just5percent,inrurallocations.Theadditionalimpactsarearound23percentlargerforbothmaleandfemaleworkersthanthosewithaccesstoonly

oneofthetypesofinfrastructureinurbanareas.

ContributionstotheliteratureOurpaperprovidesnovelfindingsonhow

theexpansionofinternetnetworkscomplementsimportantinfrastructuressuch

asroadstoenhancejobcreationandemploymentindevelopingcountries,anddocumentshowthebenefitsofthiscomplementarityvaryacrossindividualsandsocieties.Itdoessousingbothindividualandfirm-leveldatafromvarioussourcesandemployinganidentificationstrategythatinvolvesanewinstrumentalvari-ablebasedonthearbitraryethniccompositionandpoliticalhistoryofAfricancountries.Ouranalysisaddstotherecentliteratureontheimpactsofinternetaccessoneconomicdevelopment.

HjortandTian

(2021)provideanoverviewof

thenascentyetalreadysizeableempiricalbodyofresearchontheeconomicimpactofinternetconnectivityindevelopingcountries.Theoretically,internetaccesscandriveeconomicdevelopmentthroughitsimpactsonboththesupply-sideandthedemand-sideofaneconomy.Inmanycontexts,internetaccessappearstoimprove

theproductivityofworkersandfirms(Dracaetal.,

2006;

KhannaandSharma,

2018

;

GoldfarbandTucker,

2019;

Houngbononetal.,

2022)

.Researchfindingsshowthatlabormarketoutcomesforfemaleworkersespeciallyimprovewhen

firmsusemoreICT(Dutzetal.,

2017),butthefindingsaremixedonwhetherin

-ternetconnectivityplaysaroleinbuildingskills.Demand-sideforcesfocusontheexpansionofmarketaccessforfirms,workers,and/orconsumers,andthereduc-

tionofinformationfrictions(AllenandArkolakis,

2014;

Fanetal.,

2018;

Couture

etal.

,

2021)

.Internetaccesshasalsobeenshowntoimprovehealth-relatedhu-

mancapitalbyreducingHIV/AIDSprevalence(AbbasiandPongou,

2022)

.Incontributingtothisliterature,ourworkbreaksnewgroundbyanalyzingnotjusttheeffectoftheinternetinisolation,butalsotheadditionalimpactsthatarisefromthecomplementaritybetweeninternetandroadinfrastructureinvestments

onemploymentopportunitiesandoutcomes.

Ourpaperalsoaddstotheliteratureontheimpactsoftransportinfrastructureexpansiononeconomicgrowthanddevelopment.Severalpapershaveshownthat

thenetwelfaregainsofhighwaysarelarge(Aschauer,

1989;

Shatzetal.,

2011;

DurantonandTurner,

2012;

Faber,

2013;

AllenandArkolakis,

2014;

EjazGhani,

2016

;

Heetal.,

2020).Othershavefocusedontheimpactsofbetteraccessibilityto

roadsorrailroadsonGDPpercapita(Atacketal.,

2009;

Donaldson,

2010;

Duflo

5

6

etal.

,

2015),andoneconomicdevelopment(Okoyeetal.,

2019)

.Thesepapers

havefocusedontransportinvestmentsonly.

Veryfewpapershavelookedatcomplementaritiesininfrastructure,especiallyinAfrica.AnalyzingdatafromEthiopia,

Moneke

(2020)examinestheimpacton

employmentandwelfarethatstemsfromcombinedinvestmentsinroadandelec-trification.Usingdatafrom27Sub-SaharanAfricancountries,

Abbasietal.

(2022)

exploretheheterogeneouseffectsofjointinvestmentsinroadandelectrificationonemployment,findingstrongcomplementaritiesbetweentheseinfrastructures.

Gebresilasse

(2023)documentscomplementaritiesbetweenroadconstructionand

extensionservicesinagriculturalproductivityinruralEthiopia.Ourpaperaddstothisnewliteraturebyexaminingthecomplementaritiesofinternetandroad

investments.

Ourpaperalsomakesamethodologicalcontribution.Itaddstotheempiricalliteratureoninfrastructurebyusingtwooriginalinstrumentalvariablestoaddressthepossibleendogeneityofinfrastructurelocations.Weexploittheleast-cost-pathproceduretoconstructtheoreticalroadnetworksconnectingthecentroidsofdifferentethnicterritoriesinSub-SaharanAfrica.Weinteracttheresultingexogenousdistancetothenearestmainroadwiththeaccessibilityofsubmarine

internet.

Theremainderofthispaperisorganizedasfollows:Section2presentstheconceptualframework.Section3describesthecontext,data,andourempirical

strategy.Sections4and5presentthemainfindings.Section6concludes.

2TheGainsfromInternetConnectivityandRoads

Following

HjortandTian

(2021),wepresentanoverviewofpotentialmecha

-nismsdrivingthejobandeconomicimpactsofthecombinedprovisionofinternetconnectivityandroadnetworks.Internetandroadaccesscandriveeconomic

developmentthroughbothsupply-anddemand-sideimpacts.

7

2.1Supply-Side

Thepresenceorabsenceofaccesstotheinternetandroadnetworkscandirectlyaffecttheproductivityoffirms,workers,andotherinputsintheproductionpro-cess.Findingsfrom

Lietal.

(2017)suggestthatroadinvestmentsinChinahad

significantandpositivereturns,increasingmanufacturingfirms’productivity.Asimilarstudyby

GordonandLi

(1995)findsthatroadinvestmentsledtoa4.6

percentincreaseintheproductivityoffirmsthatcontractwithgovernmentand

toa3.6percentincreaseinannualtotalfactorproductivity(Zhu,

2012)

.

Gib-

bonsetal.

(2016)similarlyfindthatnewroadinfrastructureintheUKgreatly

impactedtheproductivityoffirmsthere.Accesstotheinternetcanalsoaffectfirmproductivitybydirectlyaffectingworkers’on-the-jobproductivity,inducing

humancapitalaccumulation,andenhancingfirm-workermatching.

Severalempiricalpaperspointtowardadirectimpactofaccessibilitytoroads

onwagesandemployment(Gibbonsetal.,

2016;

Yamauchi,

2016;

Matasetal.,

2015

;

Yamauchietal.,

2011;

FrishandTsur,

2010;

Reardonetal.,

2001);otherem

-piricalstudiespointtowardadirectinfluenceofhigh-speedinternetconnectivityonwagesandemployment,too.

Chenetal.

(2020)documentsignificantincreases

inworkers’wagesandfirmproductivityinresponsetoaninternet-upgradingpro-gramthatincreasedinternetspeedsinChina.InthecontextofBrazil,

Almeida

etal.

(2017),

Poliquin

(2020),and

Tian

(2021)findthatnewlyestablishedbroad

-bandaccessincreasedworkers’wagesonaverage;bycontrast,

Dutzetal.

(2017)

findanegativecorrelationbetweenincreasedinternetaccessinBrazilandaver-agewages.InNigeria,

Bahiaetal.

(2020)examinehowtherolloutofmobile

broadbandaffectedlabor-marketoutcomes,householdconsumption,andpovertyinNigeria.Usingadifference-in-differencesapproachtotrackimpactsonin-dividualhouseholds,theyshowthatinternetconnectivityincreasedlabor-force

participationandemployment.

Anotherstrandoftheliteraturepointstowardadirectimpactofinternetcon-nectivityonhumancapitaldevelopment.Internetconnectivitycanfacilitateon-

the-jobtraining.

HjortandPoulsen

(2019)findevidencethatinternet-connected

8

firmsinsixAfricancountriesinvestedmoreintheirworkers’humancapital;

Mouelhi

(2009)alsodocumentsanadditionalhumancapitalinvestmentamong

internet-connectedfirmsinTunisia.Internetconnectivitycanaffecthumancap-italdevelopmentbothathomeandinschool.

Bianchietal.

(2020)showthat

connectinghigh-qualityteachersinurbanareasofChinawithmillionsofstudentsinruralprimaryandmiddleschoolsimprovedstudents’long-runacademicachieve-ment,labor-marketoutcomes,andinternetuse.However,theoverallevidenceontheimpactsofconnectedschoolsontestscoresismixedthusfar.Positiveimpacts

havebeendocumentedinPeru(Khoetal.,

2018)andMalawi(Derksenetal.,

2019);bycontrast,workinPeru(Malamudetal.,

2019)andBrazil(Bessoneand

Dahis

,

2020)hasfoundnoimpactontestscores

.

Internetaccesscanalsoaffectlaborproductivitybyimprovingfirm-workermatching,especiallyindevelopingeconomieswherefrictionsinthelabormarketarelarge.Empiricalevidenceislargelyfromhigh-incomecountries,suchasthe

US(KuhnandMansour

,

2011)andNorway(Bhulleretal.,

2020).

Ledermanand

Zouaidi

(2020)findarobust,negativerelationshipbetweeninternetusageand

long-term,frictionalunemploymentacrosscountries.Internetaccesscanalsoin-creasefirm-levelproductivitybyfacilitatingtheadoptiontangibleinputssuchasmachines,newtechnologies,andintermediatematerials,andtheadoptionofintan-gibleinputssuchasmanagement,organizationalpractices,andservices.Evidenceonthetake-upoftangibleinputsinresponsetointernetconnectivityislimited,butsuchhigh-speedconnectivityhasbeenshowntoaffecttheorganizationofproductionandtradeindevelopingcountries.

Houngbononetal.

(2022)findthat

individualfirmsinAfricaare20percentagepointsmorelikelytoundertakeprocessinnovationand12percentagepointsmorelikelytoundertakeproductinnovationwhenfastinternetbecomesavailable.

Tian

(2021)showsthatinternetaccessal

-lowsfirmsinurbanareastoreorganizeproductiontoenhancecollaborationand

facilitatethedivisionoflabor.

Usingenterprise-levelpaneldata,

Gibbonsetal.

(2016)examinetheeffectsof

investmentsinroadconstructiononemploymentandlaborproductivityinthe

9

UK.Theyfindthatimprovementsinroadaccesssubstantiallyincreasedworkersproductivityattheindividuallevel.InChina,

Lietal.

(2017)findthatinvestments

inroadinfrastructureyieldedanannualreturnof11.6percentfromproductivitygains.

LiandLi

(2013)findthatinvestinginroadsinChinaincreasedfirms’

productivity,withanannualrateofreturnofcloseto10percent.

2.2DemandSide

Roadscanimpactemploymentandfirms’expansionthroughexportsandtrade

(Mbekeani

,

2010;

Martincusetal.,

2017),regionalintegration(Mbekeani,

2010),

andnewmarket-entryenhancement(Meloetal.,

2010;

Holl,

2004).Internetcon

-nectivitycanalsoaffecteconomicactivitybothbydirectlyexpandingmarketaccessforfirms,workers,andconsumers,andbyaddressinginformationfrictions.Forexample,e-commercemayallowfirmstomaketheirproductsaccessibletomoreconsumers,especiallyinruralandremoteregions.Internetconnectivityap-pearstoenablefirmstoexpandtheirsalesbetweenregionswithincountries(see

Fanetal.

(2018)forChina)andthroughexportingandimporting(see

Hjortand

Poulsen

(2019)forAfrica).Internetaccessappearstolowerpricesandexpandthe

varietyofchoicesavailabletoconsumers(Coutureetal.

(2021),forChina).Inter

-netaccesscanhelpreduceinformationfrictionsthatarepervasiveindevelopingcountries.Internetaccesscanleadto(1)areductioninpricedispersion,(2)higherlocalpriceswhenbuyershaveadegreeofmonopsonypower(see

Goyal

(2010)for

India,and

GuerreroBarretoandRitterBurga

(2014)forPeru),(3)areductionin

uncertaintyaboutproductquality,and(4)improvementincommunicationwith

tradepartners(see

Leuvenetal.

(2018)and

Fernandesetal.

(2019)forChina)

.

2.3Heterogeneities

Theliteraturehasuncoveredimportantheterogeneityintheeffectsofaccesstohigh-speedinternetonwagesandproductivity.Manypapershavefoundthatinternetconnectivityappearstoespeciallybenefitfemaleworkers.InVietnam,

ChunandTang

(2018)findsuggestiveevidencethatfirmsthatincreasedtheir

10

ICTusealsoincreasedtheshareoffemalelabor.InBrazil,

Dutzetal.

(2017)

findthatemploymentgrowthfrominternetaccessestablishedindifferentareasofBrazilwasgreateramonglow-skilled,female-filledjobs.Similarresultshavebeen

foundinMexico(Juhnetal.,

2013)andNigeria(Bahiaetal.,

2020)

.

Theevidenceoninternettechnologyasasourceofskillbiasindevelopingcountriesismixed.

KhannaandSharma

(2018)showdescriptiveevidenceofcom

-plementaritybetweenusingICTandundertakingnon-routinetasks.

Chenetal.

(2020)findthattheadoptionofhigh-speedinternetprovidedgreaterbenefits

toChinesefirmsthatwereinmoreskill-intensiveindustriesandthathadmoreeducatedworkers.

Dutzetal.

(2017)alsofindevidencethat,withinthemanufac

-turingsectorinBrazil,internetaccessappearstohaveraisedwagesinmedium-andhigh-skilljobs,butnotinlow-skilljobs.InTanzania,

Bahiaetal.

(2021)show

thatbroadbandavailabilityincreasedbothlaborforceparticipationandwageem-ploymentamongyoung,educatedmen.

HjortandPoulsen

(2019)showthatthe

gradualarrivaloffastinternetinfrastructureinAfricaincreasedemploymentrates,evenforless-educatedworkergroups,althoughtheestimatesarethattheeffects

areconsiderablylargerformore-educatedworkers.

Theevidenceondifferencesacrossurbanandruralareasindevelopingcountriesisalsomixed.

Masakietal.

(2020)findthatthelabormarketimpactoffast

internetinSenegalislargerforhouseholdsinurbanareasthanthoseinruralareas;bycontrast,

Bahiaetal.

(2020)findespeciallybeneficialimpactsinrural

areasinNigeria.

Theeffectsofroadconstructionhavealsobeenshowntovary,dependingonlocation,industrytype,andquality.

Baum-Snowetal.

(2017)provideevidence

thatnew,regionalhighwayspromotedtheconcentrationofoutputsbutalsotheconcentrationoftheworking-agepopulationintoregionalprefecturesinChina.

Moneke

(2020)findthatinvestmentsinroadexpansionsthatbroughttheroad

networkclosertoisolatedareasimprovedproductivity,withsomemanufacturingsectorsgrowingandexpandingtoexporttheirproducts.Thestudyalsofound

thatincreasingroadaccessalonecausedalargeshiftofemploymentfromthe

11

manufacturingsectortosectorsinvolvinghigh-skillactivities,and,atthesame

time,causedserviceemploymenttorise,largelyininformal,smallretailbusinesses.

3DataandEmpiricalStrategy

3.1DataDescription

Wecombinegeo-codedsurveydatawithspatialdataoninternetandroadnet-

works

.1

3.1.1DemographicandHealthSurveys(DHS)

TheDHSarecross-sectionalsurveysthathavebeenconductedinthemajorityofdevelopingandmiddle-incomecountriessincethe1980s.Theyarerepresentativeatthenationalandsubnationallevels.Theyprovideinformationondemographiccharacteristicsandsocioeconomicstatus(e.g.,age,gender,education,occupation)ofallhouseholdmembersandrecentsurveyshavebeengeoreferenced.RecentstudieshaveusedDHStoexaminetheeconomicandsocialimpactsofinfrastruc-

tureindevelopingcountries(Okoyeetal.,

2019;

HjortandPoulsen,

2019;

Moneke,

2020

;

Canningetal.,

2020;

HerreraDappeandLebrand,

2021;

Lebrand,

2022;

Ab-

basietal.

,

2022).Werelyondatafrom56surveysconductedin20Sub-Saharan

Africancountries.Thetotalsamplesizeis806,378,with552,022womenand254,356men.Gridcells,20kilometersby20kilometersinarea,areconstructedaroundeachlocationtoaccountforanytime-invariantfactorthatmightaffectourinfrastructurelocationandoutcomevariables.Wegroupthosewhoareem-ployedintotwocategories:high-skilledworkersandlow-skilledworkers.Wealsodifferentiatebetweenagriculturalworkersandnon-agriculturalworkers.Skilledworkersincludeemployedworkersinthefollowingoccupationcategories:profes-sional,managerial,skilledmanuallabor,services,oragriculture.Inoursample,

67.7percentofrespondentsclaimedtobeemployed,while32.3reportedbeing

1Table(

A1

)intheOnlineAppendixprovidesdetailsonthecountriesanalyzedandyearsinwhichthesurveyswereconducted.

12

unemployed

.2

3.1.2WorldBankEnterpriseSurveys(WBES)

TheWBESarecountrywide,representativesamplesofbusinessesandfirmsfromallsectorsofactivitywithatleastfiveworkers.Theyincludeinformationonenterprises’establishments,assets,operations,sourcesoffunding,structureoftheworkforce,andtypeofactivities.Weexploitdatathatwerecollectedoverthe2005-2018periodforcoastalnationsthathadsurveycyclesbeforeandafterthedatewhensubmarineinternetwasacquired.ThesecountriesareGhana,Kenya,Mauritania,Nigeria,Senegal,andTanzania.Welimitoursampletosmall-andmedium-sizeenterprises(SMEs)thatemployamaximumof99individuals.Itisalsoimportanttonotethatwhiledigitalizedroadmapsarefortheperiodbefo