IMF-交易成本如何影響匯款

HowDoTransaction

CostsInfluence

Remittances?

KangniKpodarandPatrickAmirImam

WP/22/218

IMFWorkingPapersdescriberesearchin

progressbytheauthor(s)andarepublishedto

elicitcommentsandtoencouragedebate.

TheviewsexpressedinIMFWorkingPapersare

thoseoftheauthor(s)anddonotnecessarily

representtheviewsoftheIMF,itsExecutiveBoard,

orIMFmanagement.

2022

NOV

*Kpodar:InternationalMonetaryFundandFERDI;Imam:InternationalMonetaryFund.ThispaperbenefitedfrominsightfulcommentsandsuggestionsfromCarineMeyimdjui,CedricOkou,BaopingShangandAzarSultanov.ThisresearchispartoftheMacroeconomicResearchinLow-IncomeCountriesproject(ProjectID:60925)supportedbytheUK’sForeign,CommonwealthandDevelopmentOffice(FCDO).TheviewsexpressedinthispaperarethoseoftheauthorsanddonotnecessarilyrepresenttheviewsoftheInternationalMonetaryFund(IMF),orFCDO.

?2022InternationalMonetaryFundWP/22/218

IMFWorkingPaper

Strategy,PolicyandReviewDepartmentandInstituteforCapacityDevelopment

HowDoTransactionCostsInfluenceRemittances?

PreparedbyKangniKpodarandPatrickAmirImam*

AuthorizedfordistributionbyJohannesWiegand

November2022

IMFWorkingPapersdescriberesearchinprogressbytheauthor(s)andarepublishedtoelicitcommentsandtoencouragedebate.TheviewsexpressedinIMFWorkingPapersarethoseoftheauthor(s)anddonotnecessarilyrepresenttheviewsoftheIMF,itsExecutiveBoard,orIMFmanagement.

ABSTRACT:Usinganewquarterlypaneldatabaseonremittances(71countriesovertheperiod2011Q1-2020Q4),thispaperinvestigatestheelasticityofremittancestotransactioncostsinahighfrequencyanddynamicsetting.Itaddstotheliteraturebysystematicallyexploringtheheterogeneityinthecost-elasticityofremittancesalongseveralcountrycharacteristics.Thefindingssuggestthatcostreductionshaveashort-termpositiveimpactonremittances,thatdissipatesbeyondonequarter.Accordingtoourestimates,reducingtransactioncoststotheSustainableDevelopmentGoaltargetof3percentcouldgenerateanadditionalUS$32bninremittances,higherthatthedirectcostsavingsfromlowertransactioncosts,thussuggestinganabsoluteelasticitygreaterthanone.Amongremittancecost-mitigationfactors,highercompetitionintheremittancemarket,adeeperfinancialsector,andadequatecorrespondentbankingrelationshipsareassociatedwithalowerelasticityofremittancetotransactioncosts.Similarly,remittancecost-adaptationfactorssuchasenhancedtransparencyinremittancecosts,improvedfinancialliteraryandhigherICTdevelopmentcoincidewithremittancesbeinglesssensitivetotransactioncosts.Supplementingthepanelanalysis,theuseofmicrodatafromtheUSA-Mexicocorridorconfirmthatmigrantsfacinghighertransactioncoststendtoremitless,andthatthiseffectislesspronouncedforskilledmigrantsandthosethathaveaccesstoabankaccount.

JELClassificationNumbers:D23,F24,F32

Keywords:Remittances;TransactionCosts;Elasticity;Migration

Author’sE-MailAddress:

pimam@

;

kkpodar@

WORKINGPAPERS

HowDoTransactionCosts

InfluenceRemittances?

PreparedbyKangniKpodarandPatrickAmirImam

IMFWORKINGPAPERSHowDoTransactionCostsInfluenceRemittances?

INTERNATIONALMONETARYFUND2

Contents

I.INTRODUCTION

4

II.DATAANDEMPIRICALSTRATEGY

7

A.CostofRemittancesandRemittanceFlowData

7

B.StylizedFactsonRemittanceCosts

8

C.TheModelandEconometricApproach

1

2

III.THERESULTS

1

3

A.HowElasticAreRemittancestoTransactionCosts?

1

3

B.ExplainingHeterogeneitiesintheCostElasticityofRemittances

1

6

C.ExaminingtheResponseofRemittancestoTransactionCost:EvidencefromMicroDatafromthe

US-MexicoCorridor

2

5

IV.Conclusion

2

9

References

38

FIGURES

1.Averagefeeasashareofa$200USDremittance(2011Q1-2020Q4)(percent)

8

2.ChangeinAveragefeefora$200USDremittance(2016Q1-2020Q1)(percent)

9

3.TrendsinAveragefeefora$200USDremittanceandInterestRateSpread(2011-20)(percent)

10

4.Averagefeefora$200USDremittancebyTypeofProvider(2011Q1-2020Q4)(percent)

1

1

5.RemittancesFlowsandAveragefeefora$200USDremittance(2011Q1-2020Q4)

1

1

6.Cost-ElasticityofRemittances:LocalProjections

1

5

7.Cost-ElasticityofRemittanceswithRespecttoCompetitionintheRemittanceMarket

1

8

8.Cost-ElasticityofRemittancesinCountrieswithLowandHighFinancialDevelopment

1

9

9.Cost-ElasticityofRemittancesandCorrespondentbankingrelationships

20

10.Cost-ElasticityofRemittancesandPriceTransparency

21

11.Cost-ElasticityofRemittancesandEducationlevel

2

2

12.Cost-ElasticityofRemittancesandICTDevelopment

2

4

13.Cost-ElasticityofRemittancesandICTAffordability

2

5

TABLES

1.TransactionCostsandRemittances:Fixed-EffectEstimates

1

4

2.TransactionCostsandRemittancesintheUS-MexicoCorridor

2

7

3.TransactionCostsandRemittancesintheUS-MexicoCorridor:TheRoleofFinancialLiteracy

2

8

4.TransactionCostsandRemittancesintheUS-MexicoCorridor:InstrumentalVariableApproach

2

9

ANNEXES

1.SampleComposition

3

1

2.SummaryStatisticsandCorrelationMatrix

32

INTERNATIONALMONETARYFUND3

ANNEXTABLES

1.Cost-ElasticityofRemittances:LocalProjections

33

2.First-StageRegression

34

ANNEXFIGURES

1.ChangeCost-ElasticityofRemittances:InstrumentalVariableLocalProjections

34

2.Cost-ElasticityofRemittances:LowvsHighTransactionCost

35

3.Cost-ElasticityofRemittanceswithrespecttotheGeographicalCoverageofFinancialInstitution

35

4.Cost-ElasticityofRemittanceswithrespecttoAccesstoDepositAccounts

36

5.Cost-ElasticityofRemittancesandICTAccess

36

6.Cost-ElasticityofRemittancesandICTUse

37

7.Cost-ElasticityofRemittancesandICTCapability

37

IMFWORKINGPAPERSHowDoTransactionCostsInfluenceRemittances?

INTERNATIONALMONETARYFUND4

INTRODUCTION

Remittancesmatterfordevelopment.Theyareakeysourceoffundingfordevelopingcountries.Migrantssendpartoftheirearningstofamilymemberstoprovidethemwithbasicsubsistence,buildandinvestintheeconomybackhome.AccordingtoWorldBankdata,remittancestolow-andmiddle-incomecountriesmorethandoubledduringthepast15yearstoreachUS$550billionin2021.Overhalfofitgoestopeopleinruralareas,andabout75percentisusedtocoverbasicssuchasfoodandmedicalorschoolexpenses,whiletheremainingisinvestedinassetsorsaved(IFAD,2021).

Beingamajorsourceofincomeforthepooresthouseholds,policymakersinadvancedanddevelopingcountriesarelookingatavenuestoencourageremittanceflowstogrowevenfurther.However,amajorstumblingblockinsendingmoneyhomeisthehigh,somewouldargueexcessive,transactioncostsinvolved.Despitethecommitmentsfromtheinternationalcommunitytoreducethecostofremittances,andtheinclusionofthe3percentcosttargetintheSustainableDevelopmentGoals(SDGs),progresshasbeenslowinrecentyears.

1

ThestrongresilienceofremittancesduringtheCovid-19pandemichasbroughtbacktotheforefrontthedebatesurroundingthestubbornlyhightransactioncosts(seeKpodaretal.,2021).

Itisacceptedwisdomthatindividualssendingmoneyhomeinmanypartsoftheworld,particularlysub-SaharanAfrica,arepayingaveryhighcost,whichmaydissuadefurtherflows.Consideringfees,exchangeratemarginandothercosts,between5and15percentofremittancesare“l(fā)ost”duetohightransactioncosts,dependingonthecountryandtheamountssendhome(Ratha,2021).Becauseformalremittancesinvolvehighfixedcostsandhenceareexpensivetoprovide,low-incomeindividualsrefrainfromremitting,orareincentivizedtousecheaperinformalalternatives(Gibson,McKenzieandRohorua,2006;Yang,2011).Againstthisbackdrop,thispaperattemptstoaddressthefollowingquestions:howelasticareremittancestochangesintransactioncosts?whatarethefactorsorpolicyinterventionsthatmayhelpexplaincross-countrydifferencesinthecostelasticityofremittances?

Thispaperisrelatedtoseveralstrandsoftheliterature.First,alargebodyofworklooksattheimpactofremittancesatthemacroeconomicandmicroeconomiclevel.Manystudiesfindthatremittancesstimulategrowth,althoughnobroadconsensusexistsyetonthatmatter(seeCazachevicia,Havraneka,andHorvath,2020,forarecentdiscussion).Thisfavorableimpactbuildsonthepremisethatremittancesareanimportantsourceofdevelopmentfinancing,asignificantsourceofinternationalreserves,andoftenarrivecountercyclically.Atthemicro-level,remittancesrespondtoincomeshortfallsand,inthatway,havethepotentialtosmoothhouseholds’income,therebyreducingriskstothefamily(ChoiandYang,2007).Remittancesarealsolinkedtoimprovedeconomic,healthandeducationoutcomesaswellashigherfinancialinclusion.However,therearealsopotentiallysomedeleteriouseffectsobservableatthemacro-level(Chami,Fullenkamp,andJahjah,2003).BesidesproducingDutchDisease-typeeffects,potentiallyreducingthequalityofinstitutions,ordelayingneededfiscaladjustment,remittances

1TheSDGsincludetheobjectiveto“by2030,reducetolessthan3percentthetransactioncostsofmigrantremittancesandeliminateremittancecorridorswithcostshigherthan5percent”(SDG10.C).

IMFWORKINGPAPERSHowDoTransactionCostsInfluenceRemittances?

INTERNATIONALMONETARYFUND5

mayalsocomplicatethemonetarytransmissionmechanism,makingmonetarypolicyharder.Thereisalsoevidencethatremittancesmaybreeddependencybydiscouragingreceivinghouseholdmembersfromworking,possiblytrappingcountriesintoanemiclong-rungrowth(seeAmuedo-Dorantes,2014,forasummary).

Second,anarrowsetofstudiesinvestigatethedeterminantsofremittancetransactioncosts.Forinstance,FreundandSpatafora(2008)usecross-countrydatafor66countriesandfindthatrecordedremittancesdependpositivelyonstocksofmigrantsandnegativelyontransactioncostsandexchangeraterestrictions.Transactioncostsarealsolowerwhenfinancialsystemsaremoredeveloped,andexchangeratesarelessvolatile.Similarly,BeckandMartinezPeria(2011)conclude,basedonadatasetcovering119countrycorridors,thatthesizeofthemigrantpopulation,bankingcompetitionandlowerbarrierstoaccessbankingservicesareconducivetolowerremittancecosts.Morerecently,Beck,Janfils,andKpodar(2022)exploitaricherremittancecostdatasettoshowthatcost-andrisk-basedconstraintsaswellasmarketstructurehinderaffordableremittancetransactioncosts.Corridorandfirm-specificanalysisrevealsthathigherincomepercapita,easiergeographicalaccesstofinancialinstitutions,largerremittancemarket,shorterdistancebetweensendingandreceivingcountries,competitionintheremittancemarket,andpeggedexchangerateregimeareassociatedwithlowertransactioncosts.Inthesamevein,daSilvaFilho(2021)underscoresthatthedriversofremittancecostsaremultipleandcomplex,withregulatoryissues(suchasexclusivityclauses,stringentAML/CFTregulation,andrestrictivelicensing),alackofpricetransparency,highernumberofbanksandathinremittancemarketbeinglikelytohaveadetrimentaleffectonremittancecosts.

Third,thereareseveralselectedstudiesfocusingonthecostelasticityofremittances.Theserangefromcross-countryandpanelstudies(e.g.;FreundandSpatafora,2008,Ahmed,Mughal,andMartínez-Zarzoso,2021)tocountry-specificstudies(FerrianiandOddo,2019,onmigrantsinItaly;KosseandVermeulen,2014,onmigrantsintheNetherlands;Kakhkharova,Akimovb,andRohdeb(2017)onmigrantsinRussia;AhmedandMartínez-Zarzoso,2016,onmigrantsfromPakistan;andGibson,McKenzie,andRohorua(2006)onmigrantsfromTonga).Overall,thesestudiesdocumentasignificantandnegativeeffectoftransactioncostsonremittanceinflows,partlyduetohightransactioncostsencouraginginformalremittances.Nevertheless,thesestudiesdonotgobeyondthecostelasticityestimatetoinvestigatehowthisvariesacrosscountrieswithdifferentcharacteristics.

Inthispaper,weimproveonexistingstudiesalongseveraldimensions.First,weexploitanewquarterlydatabaseonremittances,allowingustoinvestigatetheelasticityofremittancestocostwithhighfrequencydatainadynamicsettingusinglocalprojections(Jorda,2005).Assuch,thispapershedslightontheshort-andmedium-termimpactofashocktotransactioncostsonremittancesandhowpersistentthiseffectmightbe.Second,unlikepreviousstudies,thispaperundertakesasystematicanalysisoffactorsthatcanshapethecostelasticityofremittances.Itisimportanttonotethatwearenottakingaviewonwhetherhavingalow-costelasticitytoremittancesisgood—alowelasticityimplieslessreductioninremittanceflowswhentransactioncostsincrease,butalsolowerincreaseinremittancesoccurswhentransactioncostsarereduced.Butthereasontofocusonelasticitiesisthatalowerelasticitywillimplythatfewerremittancesflowinformally.Twobroadcategoriesoffactors/policiesareconsidered:(i)cost-mitigation

INTERNATIONALMONETARYFUND6

policiesthatdirectlytackletherootcausesofhighremittancecosts(e.g.policiestopromotecompetitionintheremittancemarket,fosterfinancialsectordevelopment,andsafeguardcorrespondentbankingrelationships)and(ii)cost-adaptationpoliciesthatdonotprimarilyaddressthecostofremittances,butpromotemoreefficientremittancechoices(e.g.policiestoenhancepricetransparency,financialliteraryandinformationandcommunicationtechnologies(ICT)development).Third,thispaperusesmicrodatafromtheUSA-Mexicocorridortoascertainthecostelasticityofremittancesanditsheterogeneitywithrespecttofinancialliteracy,thusprovidingmoregranularresultstosupporttheevidencefoundatthemacrolevel.StudyingtheUSA-Mexicocorridorisinterestingfromseveralstandpoints.Itisthelargestcorridorintheword;hasseveralremittanceserviceproviderswithalargearrayofproducts;andhasin-depthpubliclyavailabledata.

Withasamplecovering71countriesovertheperiod2011Q1-2020Q4,thefindingssuggestthata10percentreductionintransactioncostsleadstoa0.9percentincreaseinremittancesinthefirstquarteraftertheshock,withtheimpactbecomingstatisticallyinsignificantfromzeroinsubsequentquarters,implyingthattheresponseofremittancestotransactioncostsisessentiallyofashort-termnature.Thisresultsuggeststhatmovingfromthe2020leveloftransactioncosts(6.3percent)totheSDGtargetof3percentwillgenerateanadditionalUS$32billioninremittances,muchlargerthanthedirectcostsavings.Therefore,migrantswouldnotonlyfullypassonthecostsavingstotheirfamilies,butalsosendmore,implyinganabsoluteelasticityhigherthanone.Theresultsarerobusttoaninstrumentalvariableapproach.

Ininvestigatingthecost-mitigationandadaptationfactors,thispaperfindsthatincountrieswherecompetitionintheremittancemarketishigh,thefinancialsystemisdeveloped,andtieswithcorrespondentbanksholdsup,theelasticityofremittancetotransactioncostsismuchlowerthanotherwise.Thisindicatesthatremittancesarelesssensitivetotransactioncostswherealternativeinformalchannelstotherepatriationofmoneyexist.Likewise,lessopaqueremittancetransactioncosts,improvedfinancialliteraryandhigherICTdevelopment(alongitsmultipledimensions:ICTuse,access,capabilityandaffordability)helpexplainwhysomecountriesmayhavealowercost-elasticityofremittances.Finally,evidencefromtheUSA-Mexicocorridor,usinganannualsurveyoftheBankofMexicoduring2013-2017andcoveringover37,000individuals,confirmsthatmigrantswhofacehighertransactioncoststendtoremitless,evenaftercontrollingforsocio-economiccharacteristics.Moreimportantly,educationleveloraccesstoabankaccount,aproxyoffinancialliteracy,mitigatethecost-elasticityofremittances,consistentwiththefindingsfromthepaneldata.

Thepaperisstructuredasfollows.SectionIIpresentsdataandsomestylizedfactsonremittancesandtransactioncostsandlaysouttheempiricalmodelandmethodologyemployed.SectionIIIdiscussesthekeyresultsofthepaper,withadditionalfindingsrelegatedtotheannexes,whileSectionIVconcludeswithsomepolicyrecommendations.

IMFWORKINGPAPERSHowDoTransactionCostsInfluenceRemittances?

INTERNATIONALMONETARYFUND7

DATAANDEMPIRICALSTRATEGY

A.CostofRemittancesandRemittanceFlowData

Thekeysourceofremittancecostdataisthe“RemittancePricesWorldwide”datasetcompiledbytheWorldBank.Thisdatabaseprovidesinformationfor365corridors(consistingofacombinationof48sendingand105receivingcountries)andbyremittanceserviceproviderregardingthefeepaidbythesenderforsendingtheequivalentsofUS$200andUS$500inlocalcurrency.Italsocontainstheexchangerateappliedtothetransaction,ifavailable,thetypeoftheremittanceserviceprovider(e.g.,aBankorMoneyTransferOperator-MTO),thetimeittakesforthemoneytobecomeavailabletothereceiver,thepaymentinstrumentthatcanbeusedbythesenderandtheformofthepaymenttothereceiver(e.g.cashtocash)aswellastheaccesspointfortheremittanceservice.Thedataareavailableonaquarterlybasisfrom2011Q1onwards.Intheanalysisbelow,theremittancecostasashareoftheamounttransferredaveragedatthecountrylevelisthemainvariableofinterest.

2

Inaddition,theWorldBankalsocompilesadatabaseonremittanceinflowscovering215countriesandterritoriesgoingbacktothe1980s.Thisdatabaseis,however,carriedoutonanannualbasis,andthereforedoesnotallowtoexploitthequarterlyvariationsinthecostofremittances.Moreover,onewouldexpectremittanceflowstorespondsharplytoshort-termvariationsintransactioncosts,whichannualdatacanfailtocaptureormay“oversmooth”.Toaddressthisissue,weconstruct,asdiscussedbelow,anewanduniquedatasetofquarterlyremittanceflowsforasampleof95countries,consistingof18high-incomecountries,62middle-incomecountriesand15low-incomecountries.Thedatarunfrom1971Q1forahandfulofcountriesthrough2020Q4formostcountries.

3

ThenewquarterlyremittancedatabasebuildsonthemonthlydatasetfromKpodaretal.(2021).Theapproachusedinthispaperissimilar,withthemainsourcesbeingthedetailedbalanceofpayments(BOP)andstatisticalnotespublishedbynationalcentralbanksandstatisticalinstitutes.Forsomecountries,dataarereportedinlocalcurrencyoradifferentcurrencythantheUSdollar,inwhichcase,weusethequarterlyaverageexchangeratefromtheIMF'sInternationalFinancialStatistics(IFS)databaseorrelevantcentralbankstoconverttheremittanceflowsintoUSdollars.Thecompilationoftheremittancedatafollowsaninternationallyaccepteddefinitionwherebyremittancesarethesumofpersonaltransfersandcompensationofemployees.

4

ForcountriesthatdonotreportpersonaltransfersintheirBOP,dataonworkers'remittanceshavebeenusedinsteadasaproxy.Whendatacompensationof

2Asimpleaverageisadoptedasdataonthemarketshareofthedifferentremittanceserviceprovidersarenotavailable.

3Twocountrieshavequarterlydatafromthe1970s,andthisfigurerisesquicklyto30countriesintheearly2000s.From2010,thesamplereaches70countriesbeforestabilizingabove90countriesfrom2017onwards.

4Personaltransfersincludeallcurrenttransfersincashorinkindbetweenresidentandnonresidentindividuals,regardlessofthesourceofincomeofthesenderandtherelationshipbetweenhouseholds.Compensationof

employeesreferstotheincomeofcross-border,seasonal,andothershort-termworkerswhoareemployedinaneconomywheretheyarenonresident,orresidentsemployedbynonresidententities.

IMFWORKINGPAPERSHowDoTransactionCostsInfluenceRemittances?

INTERNATIONALMONETARYFUND8

employersismissing,theyarenotaccountedforgiventhattheseflowsaretypicallymarginalcomparedtothesizeofpersonaltransfers.

Thecombinationofthequarterlyremittancesflowsandcostsdataresultsinasampleof71countries,coveringaperiodfrom2011Q1to2020Q4(Annex1providesthesamplecomposition).Theothervariablesinthemodelwillbediscussedinthesubsequentsectionsastheyarebroughtinthemodelspecification.

B.StylizedFactsonRemittanceCosts

Figure1illustratestheaveragefeeasapercentofaremittanceofUS$200bycountryduringtheperiod2011Q1to2020Q4.ThecostsformigrantstosendmoneyacrossbordersareextremelyexpensiveinsouthernAfrica(e.g.Angola,BotswanaandNamibia),wheretheyreachedmorethan2.5timesthesampleaverage(7.5percent),oftenduetohighexchangeratemargins.Smallislandsalsoexhibithighremittancecosts,probablyreflectinglackofeconomyofscaleandlimitedintegrationintotheinternationalfinancialsystem.Attheotherendofthespectrum,thelowestcostsareprevalentineasternEurope.Inbetween,awiderangeofcountriesatdifferentlevelsofdevelopmentexist,withdifferentcorridor-specificcharacteristics.

Figure1.Averagefeeasashareofa$200USDremittance(2011Q1-2020Q4)

(percent)

s

n,

n

0246810121416182022

mibia

ap

R

C

aboVerde

d

blic

L

P

Gua

Mexi

Pana

Com

Hond

Nepal

Ecuad

Nicarag

inea

nds

P

So

erzegovina

G

r

ntilles

S

K

Bulga

Albani

A

Republic

.

K

E

Mo

eg

e

Re

M

C

M

S

S

Romania

Philippine

.

or

ublic

Yeme

ZimbabweambiaThea,zaniai