世界银行 -中低收入国家的国内贸易成本_第1页
世界银行 -中低收入国家的国内贸易成本_第2页
世界银行 -中低收入国家的国内贸易成本_第3页
世界银行 -中低收入国家的国内贸易成本_第4页
世界银行 -中低收入国家的国内贸易成本_第5页
已阅读5页,还剩61页未读 继续免费阅读

下载本文档

版权说明:本文档由用户提供并上传,收益归属内容提供方,若内容存在侵权,请进行举报或认领

文档简介

PublicDisclosureAuthorizedPublicDisclosureAuthorized

PolicyResearchWorkingPaper10789

Intra-nationalTradeCostsinLow-andMiddle-IncomeCountries

BernardoDíazdeAstarloaNinoPkhikidze

WORLDBANKGROUP

TransportGlobalPracticeJune2024

PolicyResearchWorkingPaper10789

Abstract

Casualobservationsuggeststhatintra-nationaltradecostsremainhighinlow-andmiddle-incomecountries.Preciselyestimatingthemiscrucialforguidingpoliciesaimedatoptimizingeconomicefficiencywithinacountry’sborders.Thispaperestimatesintra-nationaltradecostsforsixlow-andmiddle-incomecountriesinAfricaandEasternEurope:Kenya,Madagascar,Nigeria,Rwanda,Tanzania,andGeor-gia.Theanalysisexploitsunit-levelpricedatacollectedbycountries’nationalstatisticalofficesforconsumerprice

indexcalculationpurposes.Itappliesthepricedifferentialmethodology,whichaimsatestimatingtradecostswhileaccountingforthepossibilityofimperfectcompetitionamongintermediaries,controllingforspatialvariationinmarkups.Thefindingsshowthattheintra-nationaltradecostsinthesampleofcountriesarebetween2.5and14timeslargerthanpreviousestimatesfortheUnitedStatesusingthesamemethodology.

ThispaperisaproductoftheTransportGlobalPractice.ItispartofalargereffortbytheWorldBanktoprovideopenaccesstoitsresearchandmakeacontributiontodevelopmentpolicydiscussionsaroundtheworld.PolicyResearchWorkingPapersarealsopostedontheWebat

/prwp.Theauthorsmaybecontactedatbernardo

.diazastarloa@economicas.uba.arandnpkhikidze@.

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

ProducedbytheResearchSupportTeam

Intra-nationalTradeCostsinLow-andMiddle-IncomeCountries*

BernardoD´ıazdeAstarloatNinoPkhikidze‡

Keywords:Tradecosts,Pricedifferentialmethodology,Traveldistance,Kenya,Madagascar,Nigeria,Rwanda,Tanzania,Georgia.

JELclassiication:R12,F1,O1,D02.

*ThispaperwaspreparedasbackgroundresearchfortheWorldBank’sShrinkingEconomicDistanceFlagshipReport.IgnacioCaroSol´ısprovidedsuperbresearchassistance.Forhelpandassistanceinaccessingpricedata,wethanktheWorldBank’scountryteamsandthenationalstatisticsofficesoftheresearchcountries.WehavebenefitedgreatlyfromdiscussionswithMatiasHerreraDappe,RomanZarate,TheophileBougna,AlejandroMolnar,MathildeLebrand,AigaStokenberga,andtheparticipantsofWorldBank’sWorkshoponShrinkingEconomicDistance.Thefindings,interpretations,andconclusions

exp-ierilley,UniversityofBuenosAires.Email:

bernardo.diazastarloa@economicas.uba.ar.

‡WorldBank.Email:

npkhikidze@.

2

1Introduction

Internationaltradecostshavefallendrasticallyinthelastdecades.However,intra-national(i.e.within-country)tradecostsstillremainhighinlow-andmiddle-incomecountries.Quantifyingtradecostsandtheircomponentsiscrucialsincelowertradecostscanincreaseefficiency,improvethewelfareofhouseholdsandfirms,anddrivestructuraltransforma-

tionthroughincreasedtradeopportunities(Sotelo,

2019;

FajgelbaumandRedding,

2014;

CostinotandDonaldson,

2016)

.Byunderstandingthecostsassociatedwithdomestictrade,policymakerscanidentifybarriershinderingmarketaccessandformulatetargetedpoliciestoreducetheseobstacles.

Low-andmiddle-incomecountriesareoftencharacterizedbymanybarriersaffectingthefreeflowofgoodswithintheirborders,suchaspoorroadconditions,oldtransportfleet,andpoorlogistics.Theseaffecttradeflowsandlimitthegainsofreducedinterna-tionaltradecostsandincreasedregionalintegration.Remoteregionstendtobeespeciallyisolatedoftenduetohightravelcosts,includinghightraveltime.Inturn,highertrans-portationcostsareoftenpassedontoconsumersintheformofincreasedprices.Therefore,householdsandfirmsintheseareasfacehigherspatialpricegaps,whichlimitsthevariety

ofgoodstheycanchoosefromandreducesdisposableincome(Aggarwal,

2018;

Martin,

MaynerisandTheophile,

2020)

.

Despitetherecentisolatedattemptstostudytheextentofintra-nationaltradecosts,1

thereisagapintheliteratureonthecross-countrycomparisonofwithin-countrytradecosts.

Inthispaperweestimateintra-nationaltradecostsforsixlow-andmiddle-incomecountriesinAfricaandEasternEurope:Kenya,Madagascar,Nigeria,Rwanda,Tanza-niaandGeorgia.Weexploitmicrodatacollectedbycountries’nationalstatisticaloffices(NSO)forconsumerpriceindex(CPI)calculationpurposes.

Therearetwomainestimation-basedapproachesintheliteraturetoestimatetrade

costs,thegravityapproachandthepricedifferentialapproach(Co¸sar,

2022)

.Thegravityapproachlooksattheimpactofvaryingtraveldistanceortraveltimeontradeflows,whichareinverselyproportionaltoeachother

.2

UsinggravityestimationtoestimatetheimpactoftraveldistanceontradebetweenColombiancities,

Duranton

(2015)estimates

thatincreasingthetraveldistancebetweenthemby10%decreasestradeflowsby7%andweightby6%.

Thepricedifferentialmethodcombinesdataonpricesfromdifferentmarketsorcitieswithvariablesthatarethoughttoaffectthecostofshippinggoodsbetweenlocations.Usingthesedata,onecancomputepricedifferencesacrossspace(spatialpricegaps)and

assesshowdifferentcostshiftersaffectthesedifferences.Distancebetweenwherethe

1See,forexample,

AtkinandDonaldson

(2015)ontheestimatesforEthiopia,Nigeria,andtheU.S

.,and

Bougna,Ewane,JonesandKondylis

(2020)onRwanda

.

2HeadandMayer

(2014)summarizetheliteratureongravityequationsandprovideanextensive

overviewofbest-practicemethodsandastep-by-stepcookbook.

3

productisproducedorsourcedfrom(originlocations)andwhereitisfinallysoldbyretailers(destinationlocations)typicallysummarizesthecontributionoftransportcoststooveralltradecosts.However,theremaybeothervariablesthatcanbecorrelatedwithdistancewhichcanalsoaffecttradecosts,suchastheextentofcompetitionbetween

logisticsoperators,distributorsorretailers(AtkinandDonaldson,

2015)orwithin-country

“border-effects”(Borraz,Cavallo,RigobonandZipitria,

2016)

.

Inthispaper,wefollowtheprice-differentialapproachandapplythemethodologydevelopedby

AtkinandDonaldson

(2015),whichaimsatestimatingtradecostswhile

takingintoaccountthepossibilityofimperfectcompetitionamongintermediaries,con-trollingforspatialvariationinmarkups.Wefindthatintra-nationaltradecostsinoursampleofcountriesarebetween2.5and14timeslargerthanpreviousestimatesfortheU.S.by

AtkinandDonaldson

(2015)usingthesamemethodology.

Thispaperrelatestoagrowingliteraturethatattemptstoestimateintra-nationaltradecosts.Givenitshistoricallyhightransportcosts,Africa,andinparticularSub-SaharanAfrica,hasbeenthefocusofseveralpapersstudyingtradeandtransportprices.

Bougnaetal.

(2020)estimateintra-nationaltradecostsinRwandausingdataonrural

marketsandfindthemtobe10timeshigherthanintheU.S.Theyalsofindthatmarkupsarehigherinlocationswithfewertraders.

Mayneris,MartinandTheophile

(2020)doc

-umentspatialdifferencesinthecostoflivingacrossEthiopiancitiesandfindthatmoreremotecitiesshowsignificantlyhigherprices,lessproductavailability,andahighercostofliving.

Porteous

(2019)takesadifferentmethodologicalapproachto

AtkinandDonaldson

(2015)toestimateintra-nationaltradecostsinSub-SaharanAfrica

.Thepaperfocusesonstaplecerealgrainstradedbetween230largeregionalhubsin42countries,andquantifiestradecostsbyestimatingastructuralmodelthatfeaturescompetitivetraders(i.e.,ab-stractingfrommarkups),grainconsumption,storage,andtrade.Theresultsin

Porteous

(2019)suggesttradecoststhatarequalitativelysimilarbuthigherinmagnitudetothose

estimatedin

AtkinandDonaldson

(2015),althoughtheformerpotentiallyincludenon

-distancedependentfactors.Specifically,themediantradecostisover5timeshigherthaninternationalbenchmarkfreightrates.Inacounterfactualwheretradecostsonoverlandlinksarereducedtolevelssimilartotherestoftheworld,grainpricesfall30%onaverageandaggregatewelfareincreasesby1.6%ofGDP.

TeravaninthornandRaballand

(2009)

estimatetradecostsusingtruckersurveysintheU.S.andalongseveralmajortransportcorridorsinSub-SaharanAfricaandfind1.88to3.28timeshighertradecostscomparedtotheU.S.,whicharelowerthantheresultsof

AtkinandDonaldson

(2015)

.Intermsofthetradecostsinhigh-incomecountries,

Agnosteva,AndersonandYotov

(2019)use

gravitymodeltechniquestounderstandtheeffectsofintra-regional,inter-regionalandinternationalfrictionsontradeflowsinCanadianprovinces,andfindthatfurtherandless-developedregionstendtohaverelativelylowinternalfrictionsandlargeborderef-fects,whileeconomicallymoredevelopedandcentralregionsshowrelativelylowborderbarriers.

4

2Data

Oursamplecovers6low-andmiddle-incomecountriesinAfricaandEasternEurope:

Kenya,Madagascar,Nigeria,Rwanda,TanzaniaandGeorgia.Inordertoapplythemethodologydevelopedby

AtkinandDonaldson

(2015),werequiretwosetsofdatafor

eachcountry.First,weneedpricedataonnarrowly-definedproductvarieties,withob-servationsspanningmultiplelocationswithineachcountryatamonthlyfrequencyoverseveralyears.Second,weneedthelocationsinwhichproductvarietieswereproducedorfromwhichtheywereimported.Thelatterallowsfortheidentificationoftradingpairsandthecorrespondingdirectionoftradefromorigintodestinationlocations.Inthissec-tion,webrieflydescribehowweconstructthesedata.Appendix

A

includesadditionaldetailsaboutthedatapreparationprocess.

2.1Pricedata

OurpricedatacomefromCPImicrodataprovidedtousbyNSOsineachcountryin

oursample.NSOscollectpricedatafromapredefinedlistofproductvarieties(includinggoodsandservices),designedtocapturetheconsumptionbasketofatypicalhousehold,atspecifiedlocations.Apartfromthemonthlypricequote,theinformationprovidedtousbyNSOsincludesaproductdescription,insomecasesincludingthebrandandpresentation(e.g.,“Coca-Cola,canned,500ml”or“Bodylotion,Nivea,200ml”),thenameofthelocationwherethepricewascollected,and,occasionally,thenameofaproducerorthecountryfromwheretheproductwasimported.

Werestrictthechoiceofproductstonarrowlydefinedvarieties,i.e.,thoseforwhichwecanobserveadetailedproductdescriptionandabrand.Formostcountriesinoursample,pricesareatthetownorcitylevel,withtheexceptionofRwanda,wherepricesarecollectedatthe(moreaggregated)districtlevel.

Theanalysisinsection

4

usescleansamplesofpricedata,obtainedbyapplyingacleaningalgorithmtotherawdata.Moreover,asin

AtkinandDonaldson

(2015),we

convertallpricestoconstantU.S.dollarsbydeflatingthemusinginflationratesatoriginlocationsandbilateralnominalexchangeratesbetweenlocalcurrenciesandtheU.S.dollar.ThecleaninganddeflatingproceduresaredescribedinAppendix

A.

2.2Products’originlocations

Theestimationstrategyproposedby

AtkinandDonaldson

(2015)dependsonprecisely

identifyinglocationswhereaproductisproducedorimported.Todothis,wesearchovertheInternettodeterminewhichmanufacturersproduceeachbrandandwheretheirfactoriesarelocated.Forimportedgoods,weidentifythemainportsofentryineach

5

countryandtheirlocationstoassignimportedproductstothem

.3

Forlandlockedcoun-tries,weidentifyportsatneighboringcountriesthroughwhichmostimportscome(e.g.,insomecases,countriesareinconflictwithsomeborderingcountriesandimportsdonotcomethroughthem).Then,weidentifythebordercrossingsthatareclosesttothoseports.Finally,wegeocodealllocationstodeterminetheirlatitudeandlongitude.Foreachproduct,wedefinea“tradingpair”asanoriginandadestinationlocation.

Conditionalonbeingnarrowlydefined,weconsiderproductsthatarebroadlyavailableacrosslocationsandperiodsinoursample.Specifically,werestrictsamplestoproductsforwhichpricesareobservedintheproducts’sourceanddestinationlocationsinmorethansixmonths

.4

Table

1

describesthemainfeaturesoftheresultingsamplesweconsiderforeachcountry.Thenumberofproductsrangesfrom9inMadagascarto43inRwanda.Thefulllistofproductsforeachcountry,includingtheirbrands,presentations,andmanufacturers,isincludedintheAppendix

A.2.

Thenumberofdestinationlocations(markets,cities,ordistricts)rangesfrom6inGeorgiato38inNigeria.Figure

1

showsthelocationsforeachcountryinoursample,aswellasthemajorroadsconnectingthem,andindicatesthelocationwhereaproductismanufacturedorimported,onwhichweelaborateinsection

2.2.5

Table

1

alsoshowstheperiodscoveredbyeachcountrysample.WiththeexceptionoftheNigeriasample,whichstartsinJanuary2001,allsamplesstartafterJanuary2010.ThelongestpanelcorrespondstoMadagascarandcoverstheperiodfromJanuary2010toApril2021.TheshortestpanelistheoneforKenya,coveringfromOctober2018toJanuary2022.

Averageoriginprices,pricegapsbetweenoriginanddestinationlocations,anddis-tances,arereportedinTable

2.

Table1:Mainfeaturesandcoverageofthedata

Country

No.ofproducts

Origins

Destinations

Typeoflocations

Startperiod

Endperiod

Kenya

20

10

32

Cities

Oct2018

Jan2022

Madagascar

9

3

7

Majorcities

Jan2010

Apr2021

Nigeria

25

5

38

Cities

Jan2001

Jul2010

Rwanda

43

4

13

Districts

Jan2013

Dec2020

Tanzania

32

5

20

Cities

Jan2012

Apr2021

Georgia

21

4

6

Cities

Jan2012

Dec2020

Notes:thistabledescribesthemainfeaturesofthedatasetsforeachcountryinoursample.Source:

authors’elaborationbasedonCPIdatafromcountries’NSO.

3OurmainsourcesareLogcluster,awikiwithdetailedlogisticsinformationabouteachcountry,andWorldPortSource.

4Asdiscussedinsection

3.2,estimationofpass-throughratesreliesonvariationofpricesovertimefor

eachproduct-destinationpair.

5ThedataunderlyingthemapsaretakenfromtheGlobalRoadsInventoryProject(GRIP),availableat

/download-grip-dataset.

6

(a)Kenya(b)Madagascar

(c)Nigeria(d)Rwanda

(e)Tanzania(f)GeorgiaFigure1:Mapsofmarketlocations

Notes:thisfigureshowsmarketlocationsforeachcountryinoursample,aswellasmajorroadsconnectinglocations,indicatingoriginlocations(productionlocationsorportsofentry).Source:authors’elaborationbasedontheGRIPdataset.

7

2.3Distancebetweenlocations

Thebaselinemeasureofdistanceweemployisgeodesicdistance,thatis,theshortestdis-tancebetweentwolocations.Table

2

reportsthemean,minimumandmaximumgeodesic

distancebetweenoriginanddestinationlocations.

Alternatively,following

AtkinandDonaldson

(2015),weusetwoadditionalmeasures

tocontrolforquantityandqualitydifferencesincountries’roadnetworks.Thefirst,whichaimsatcapturingquantitydifferences,isthedistancebetweentwolocationsalongthequickestroute,ascalculatedbyGoogleMaps.Figure

1

showstheroadnetworkineachcountry.Thesecond,whichshouldcapturequalitydifferences,isthetimeittakestocompleteatripbetweentwolocationsfollowingthequickestroute,alsofromGoogleMaps.

Table2:Descriptivestatistics

KE

MG

NG

RW

TZ

GE

Meanoriginprice

2.306

0.922

0.902

2.328

0.898

2.577

(1.833)

(0.451)

(1.223)

(2.993)

(0.840)

(2.775)

Meandestinationprice

2.357

0.918

0.960

2.302

0.910

2.584

(1.984)

(0.520)

(1.386)

(2.980)

(1.001)

(2.729)

Meanpricegap(tradingpairs)

0.052

-0.004

0.058

-0.025

0.013

0.007

(0.454)

(0.264)

(0.433)

(0.480)

(0.662)

(0.454)

Geodesicdistance(km)

279

414

597

80

493

170

(208)

(224)

(292)

(35)

(276)

(79)

Logdistancetooriginlocation

5.305

5.858

6.214

4.273

5.943

5.001

(0.870)

(0.605)

(0.691)

(0.499)

(0.927)

(0.551)

Min.distance(tradingpairs,km)

20.6

114.3

34.2

7.9

2.2

50.9

Max.distance(tradingpairs,km)

915.8

874.9

1,254.8

152.9

1,362.6

318.5

Numberoftradingpairs

134

18

153

30

85

20

Observations

9,280

5,997

48,379

26,215

50,583

6,173

Notes:Thistablepresentsdescriptivestatisticsforoursampleofcountries.Thefirstandsecondrowsreportthemeanpriceatthefactorylocation(orlocationoftheportofentry)andthemeanpriceatdestinations.Thethirdrowreportspricegapsusingdataformtradingpairs.AllpricesaredeflatedbyinflationattheoriginlocationandthenconvertedtoU.S.dollarsusingtheexchangerateprevailingduringthebaseperiod.Thefourthrowreportsmeangeodesicdistancebetweentradingpairsmeasuredinkilometers.Thefifthrowreportstheaveragedistancetotheoriginlocation(inlogkilometers).Thefifthandsixthrowsreporttheminimumandmaximumdistancesbetweentradingpairs,respectively.Theseventhrowreportsthenumberoftradingpairs.Thelastrowreportsthenumberofobservations.Source:Authors’estimatesbasedonnationalCPIdataanddistancescalculatedusingGoogle.

3Methodologicalframework

Thetheoreticalframeworkunderlyingourestimationstrategyisbasedonthemodelde-velopedby

AtkinandDonaldson

(2015)

.Inthissection,weoutlinethemainelements

8

andintuitionofthetheory.Interestedreaderscanfindadditionaldetailsandproofsinthatpaper.

3.1Theory

Theeconomyischaracterizedbymultiplemarkets(orlocations)indexedbydandasingleconsumptiongood.TheinversedemandforthisproductineachmarketisgivenbyP(Qd,Dd),whereQdisquantitydemandedandDdcharacterizesdemandconditionsinlocationd.Productionofthegoodcantakeplaceathomeorabroadatasinglefactory.Whatmattersforthemodelisthedomesticoriginoftheproduct,denotedo.Ifproducedathome,oindicatesthefactorylocation;ifimported,oindicatestheportorbordercrossingthroughwhichtheproductenteredthecountry.

IdenticaldomesticintermediariesbuytheproductinwholesalemarketsattheoriginlocationatpricePoandsellittoconsumersind,effectivelyactingasretailersaswell.Intermediaries’tradingtechnologyischaracterizedbytheirtotalcosts,Cd(qd),whereqdisthequantitytradedfromotod.Intheextensionofthemodeltomanyproducts,wefurtherassumethateachproductissourcedfromonelocationonly,sothatocanbeomittedfromthesubscriptinqd.TotalcostsincludeafixedcostFdandamarginalcostcd.Themarginalcostincludesthecostofbuyingthegoodattheorigin,givenbyitspricePo,andthecostoftradingthefoodfromotod,givenbyτ(Xd),whereXdisavectorofcostshifters,includingdistanceand,potentially,otherfactorssuchasroadqualityanddestination-specificretailcosts.Totalcostsarethengivenby

C(qd)=Fd+[Po+τ(Xod)].

Anintermediarychoosequantitiesqdtomaximizeprofitsinanimperfectlycompetitivemarket,subjecttotheperceivedresponseofotherintermediaries.Thereisnoentryofintermediariesand,undercertainassumptions,thefirst-orderconditionsoftheintermedi-ary’soptimizationproblemimplythatpricegapsbetweentheoriginandthedestinationlocationsaregivenby:

∆Pod≡Pd−Po=τ(Xod)+µd(τod,Dd,ϕd)(1)

whereµdisthemark-upchargedbytheintermediaryatlocationd,whichdependsontradecosts,demandconditionsatd,andacompetitivenessindexϕd,whichinturndependsonthenumberofintermediariesandtheextentofstrategicinteractionsamongthem.

From(1),variationintradecostsshiftersimplies:

whereρod≡1+(∂µ/∂τod)isthepass-throughrate,thatis,theeffectoftradecostson

9

prices.Itcanbeshownthat,inthisframework,thepass-throughratedependsontheelasticityoftheslopeofthedemandscheduleandcompetitiveconditions(ϕd).InourempiricalanalysisinSection

4,themaintradecostshifterofinterest

xodisdistance.Tosimplifytheexposition,inwhatfollowswesetXod=xod.

Equation(2)illustratesthreesourcesofbiasthatcanarisewhenaimingatidentifying

tradecostsfromspatialpricegaps:(1)incompletepass-through(ρod1),(2)variationindemandconditionsacrosslocations,and(3)variationintheextentofcompetitionacrosslocations.Underadditionalassumptions,thepass-throughratecancontrolforthissourcesofbias.Inparticular,undertheassumptionthatthepass-throughrateisindependentofquantities

qd,equation(2)simplifiesto

∆Pod=ρodτ(xd)+(1−ρod)(ad−Po),(3)

whereadisalocaldemandshifter.Underthisspecification,ρodandthelocaldemandshifteradaresufficienttocontrolforlocalcompetitiveanddemandconditions.

Thefollowingsubsectionexplainshowthistheoreticalframeworkcanguidethespec-ificationofempiricalmodelstoestimatetheeffectofdistanceontradecostsexploitingorigin-destinationpricegaps,controllingforotherfactorsthatcorrelatewithdistanceandcouldaffectpricedifferentialsthroughvariationofmarkupsacrosslocations.Section

4

describesthemethodsusedtoestimatethesemodels.

3.2Estimationstrategy

Theestimationstrategyexploitspricedataonmultipleproductsksourcedfromlocationsoandsellingatlocationsdatdifferenttimeperiodst.Withvariationacrossproducts,

destinations,originlocations,andtime,equation(3)impliesthefollowingrelationship

betweenproductpricesattheoriginandthedestination:

Pt=ρdP+ρdτ(xdt)+(1−ρd)at,(4)

wherewehavefurtherassumedthatthepass-throughrateremainsfixedovertime

.6

Thelatterassumptionholdsiflocaldemandelasticitiesandcompetitiveconditions(e.g.,entryofnewintermediaries)areconstantovertime.

Inequation(4),tradecosts

τ(xdt),pass-throughratesρdanddemandshiftersat

arenotobservable.Thestrategyaimsatidentifyingτ(xdt)andρdbasedonvariation

inpricesovertimeandspace,anddistancebetweenlocations,whiletreatingatasun-

observedheterogeneityandcapturingitthroughasetoffixedeffects.Forthisreason,preciselyidentifyingoriginlocationsiskeyfortheaccuracyoftheestimates.

Theestimationprocedureinvolvestwosteps,asfollows:

6Itisinfeasibletoestimateseparatepass-throughratesforeachperiodsincethenumberofparameterstoestimatewouldequalthenumberofobservations.

10

1.Pass-throughrates.Inthefirststep,weproduceestimatesofpass-throughrates

dbyregressingdestinationpricesonoriginprices,controllingfortradecostsand

demandshifters.Specifically,weestimatetheparametersofthefollowingmodel:

Pt=ρdP+γ+γt+εdt,(5)

whereγareproduct-destinationfixedeffects,γtisaproduct-destinationlinear

timetrend,andεdtisanunobservederrortermthatcapturesshockstotradecosts

andlocaldemandshifters.

2.Tradecosts.Inthesecondstep,werecoverestimatesoftradecosts(xodt)using

pass-throughestimatesfromthefirststep.Given(unbiased)estimatesd,wecan

manipulateequation(4)toderiveandexpressionforan“adjustedpricegap”:

Computingthis“adjustedpricegap”allowsfortheestimationofτ(xdt)when

markupsarepositiveandvaryacrosslocations,i.e.,ρd1.Localdemandcon-

ditionsarecontrolledforbyincludingfixedeffectsinteractedwiththeadjustment

factor,andtradecostsaredecomposedasτ(xdt)=f(xod)+χdt,whereχdtcaptures

unobservabletime-varyingfactorsthataffecttradecosts.Theestimatingmodelthen

becomes(7)

whereαareproduct-timefixedeffects,αdaredestinationfixedeffects,andξdt

capturesunobservedshockstolocaldemandandtradecosts.InSection

4.3,we

specifythefunctionf(xod)thatwetaketothedatatorecovertheeffectofdistanceontradecosts.

4Estimatesofintra-nationaltradecosts

Thissectionpresentsestimationresultstoquantifytherelationshipbetweendistanceandintra-nationaltradecostsinoursampleoflow-andmiddle-incomecountries.Usingthenotationintroducedintheprevioussection,weattempttoestimate∂τ(xod)/∂xod,wherexodisoneofthedifferentmeasuresofdistancedescribedinsection

2.3.

Ultimately,weapplythestrategydescribedinsection

3.2,whichaimsatestimatingthisrelationship

usingdataaboutpricegapsacrosslocations,controllingforvariationinmarkupsacrosslocations.Beforeturningtothatexercise,forbenchmarkingpurposes,wefirstpresentestimatesthatabstractfromvariationinmarkups.

4.1Spatialpricegaps

Underthetheoreticalframeworkdescribedinsection

3.1,ifm

温馨提示

  • 1. 本站所有资源如无特殊说明,都需要本地电脑安装OFFICE2007和PDF阅读器。图纸软件为CAD,CAXA,PROE,UG,SolidWorks等.压缩文件请下载最新的WinRAR软件解压。
  • 2. 本站的文档不包含任何第三方提供的附件图纸等,如果需要附件,请联系上传者。文件的所有权益归上传用户所有。
  • 3. 本站RAR压缩包中若带图纸,网页内容里面会有图纸预览,若没有图纸预览就没有图纸。
  • 4. 未经权益所有人同意不得将文件中的内容挪作商业或盈利用途。
  • 5. 人人文库网仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对用户上传分享的文档内容本身不做任何修改或编辑,并不能对任何下载内容负责。
  • 6. 下载文件中如有侵权或不适当内容,请与我们联系,我们立即纠正。
  • 7. 本站不保证下载资源的准确性、安全性和完整性, 同时也不承担用户因使用这些下载资源对自己和他人造成任何形式的伤害或损失。

评论

0/150

提交评论