IMF-比特幣跨境流動(dòng)入門：測(cè)量和驅(qū)動(dòng)因素（英）

APrimeronBitcoinCross-BorderFlows

MeasurementandDrivers

EugenioCerutti,JiaqianChen,andMartinaHengge

WP/24/85

IMFWorkingPapersdescribe

researchinprogressbythe

author(s)andarepublishedto

elicitcommentsandtoencourage

debate.

TheviewsexpressedinIMFWorking

Papersarethoseoftheauthor(s)

anddonotnecessarilyrepresentthe

viewsoftheIMF,itsExecutive

Board,orIMFmanagement.

N

Ar

2024

APR

R

*WethankPamelaCardozo,AndrésFernández,ClemensGrafvonLuckner,MaggieHu,KennethKang,AdrianLee,and

conferenceandseminarparticipantsatthe4thCryptoAssetLabConference,theIMFMacrofinancialSeminar,andtheIMFEmergingMarketForumforusefulcommentsandsuggestions.

?2024InternationalMonetaryFund

WP/24/85

IMFWorkingPaper

Strategy,Policy,andReviewDepartment

APrimeronBitcoinCross-BorderFlows:MeasurementandDriversPreparedbyEugenioCerutti,JiaqianChen,andMartinaHengge*

AuthorizedfordistributionbyKennethKang

April2024

IMFWorkingPapersdescriberesearchinprogressbytheauthor(s)andarepublishedtoelicit

commentsandtoencouragedebate.TheviewsexpressedinIMFWorkingPapersarethoseofthe

author(s)anddonotnecessarilyrepresenttheviewsoftheIMF,itsExecutiveBoard,orIMFmanagement.

ABSTRACT:Therapidgrowthofcryptoassetsraisesimportantquestionsabouttheircross-borderusage.Togainabetterunderstandingofcross-borderBitcoinflows,weuserawdatacoveringbothon-chain(onthe

Bitcoinblockchain)andoff-chain(outsidetheBitcoinblockchain)transactionsglobally.Weprovideadetaileddescriptionofavailablemethodologiesanddatasets,anddiscussthecrucialassumptionsbehindthe

quantificationofcross-borderflows.WethenpresentnovelstylizedfactsaboutBitcoincross-borderflowsandstudytheirglobalanddomesticdrivers.Bitcoincross-borderflowsresponddifferentlythancapitalflowsto

traditionaldriversofcapitalflows,anddifferencesappearbetweenon-chainandoff-chainBitcoincross-borderflows.Off-chaincross-borderflowsseemcorrelatedwithincentivestoavoidcapitalflowrestrictions.

JELClassificationNumbers:

E42;E51;E58;F21;F32;F38

Keywords:

Cryptoassets;Bitcoin;Cross-borderflows;Capitalflows

Author’sE-MailAddress:

ecerutti@,

jchen@,

mhengge@

2

1Introduction

TherapidgrowthofBitcoinsinceitslaunchin2009hasincreaseditspotentialmacroeconomicimplications.Bitcoinistheunitofaccountofadecentralizedglobaldigitalecosystemwithpublicaccess.Thetechnologicalinnovationbehindit—asustainableblockchainconsistingofadistributedledgerthatoperatesandexistswithoutanytrustedparties(Halaburdaetal.,

2022

)—hasfacilitatedaverylargenumberoftransactionsandreceivedtheattentionofprivateandpublicagents.Despiteitspricevolatilityandthefactthatitisnotbackedbyanyrealassetoranygovernmentalclaim,thepriceofBitcoinandthenumberofactiveusershasincreasedmarkedlyoverthepastdecade(Aueretal.,

2023

).Theglobalnatureoftheunderlyingtechnologiesimpliesthatalargeproportionofthetransactionsarelikelyperformedacrossnationalborders.Nonetheless,identifyingBitcoin

cross-bordertransactionsisfarfromstraightforward.

WhileaccesstotheBitcoinnetworkispublic,itprovidesalargedegreeofpseudonymitybe-causetheindividualusersarenotwellidentified.Thispseudonymityoftheusers,combinedwiththesegmentationoftheBitcoinmarketintoon-chain(recordedontheBitcoinblockchain)andoff-chain(notrecordedontheBitcoinblockchain)transactions,haslimitedacomprehensiveun-derstandingofglobalBitcoincross-borderflows.

1

Inthispaper,wetacklethischallengebyusingthreecomplementaryandworkabledatasetstostudyBitcointransactionsandcross-borderflows.WediscusstheprosandconsofthedifferentapproachestoestimatingBitcoincross-borderflows,presentnovelstylizedfactsusingrawdatathatcoverbothon-chainandoff-chaintransactions,andanalysethedriversofBitcoincross-borderflowsincomparisontocapitalflowsforalargepanelofcountries.OurfindingssuggestthatBitcoincross-borderflowsresponddifferentlythancapitalflowstotraditionalglobaldrivers.Moreover,differencesappearbetweenon-chainandoff-chain

Bitcoincross-borderflows.

WestartbyconstructingdatasetsonBitcoinflows.First,webuildadatasetofon-chainflows

1Anon-chainBitcoinuser(i.e.,anypersontransactingBitcoinsthroughtheBitcoinnetwork)doesnothaveanaccountbutawallet,whichisthecombinationofaBitcoinaddress(i.e.,“anaccountnumber”,alsocalledapublickey)andaprivatekey(i.e.,apassword).Ashighlightedintheliterature,walletsarerelativelyeasytocreate,usersoftenhavemorethanonewallet,andthewaytheBitcoinsystemdefinesanaddressimpliesthatthereispotentiallyanearlyinfinitenumberofwallets.InthewordsofHalaburdaetal.(

2022

),whilethereisroomfor2160differentaddressesintheBitcoinsystem,theestimatednumberofgrainsofsandonEarthis“only”263.Pseudonymityoftheoff-chainBitcoinuser(i.e.,anypersontransactingBitcoinsoutsidetheBitcoinnetwork)isgivenbytheprivacyandcharacteristicsoftheavailableBitcoinexchanges.SeeSection

2.1.

3

acrosscryptoexchanges.Second,wecomplementthisdatasetwithon-chaincross-countryflowsestimatedbyChainalysis.Thisdatasetusesinformationonwebtrafficonthecryptoexchangesthroughwhichthetransactionsoccurtoidentifytheresidencyoftheusers.Third,weconstructadatasetofoff-chaincross-borderflowsrecordedintheoperationsoftheexchangeLocalBitcoinsfollowingGrafvonLuckneretal.(

2023

).Werelyoninformationonthefiatcurrencyusedinthetransactionsasaproxyfortheresidencyoftheusers.Ourdetaileddescriptionofthesethreeap-

proacheshighlightstheirprosandconsandillustratesthecomplementarynatureofthesedatasets.

AkeyquestionistowhatextentBitcoinisusedforcross-bordertransactions.ThefindingsinourpaperputintoperspectivenotonlytherelativeimportanceandthecharacteristicsofBitcoincross-borderflowsbutalsotheirheterogeneity,especiallyintermsofon-chainandoff-chaintransactions.TheuseofBitcoinforcross-bordertransactionsisgeographicallywidespread,withrelativelyhighintensitiesacrossregionsbothforon-chainandforoff-chainflows,andsomepunctualdifferencesdrivenbydatacoverageandtheunderlyingestimationassumptions.Ourusageofrawon-chainandoff-chaindataallowsustoestablishthaton-chaintransactionsare,onaverage,substantiallylargerthanoff-chaintransactions.ThispatternlikelyreflectsthesecurityfeaturesthattheblockchainprovidesandtheBitcoinblockchainfeestructure.ThemagnitudesoftheestimatedBitcoincross-borderflowsaresizeablewithrespecttoseveralcountries’GDP,especiallyinthosewhichexperience

relativelysmallcapitalflows.

Ananalysisofthedriversofcross-borderflowsindicatesthatBitcoincross-borderusageisnotalignedwiththesamefactorswhichdrivecapitalflows,andthattherecouldbesomedifferencesbetweenon-chainandoff-chainBitcoincross-borderflows.WhilethelevelsofcapitalflowsandestimatedBitcoincross-borderflowsarenotdirectlycomparableduetomethodologicaldifferences(grossvs.netconcepts)

2

,wegainfourkeyinsights:i)Bitcoinon-chaincross-borderflowsresponddifferentlytotraditionalinflowdriversthancapitalflows.Theyseemtobenegativelyrelatedtobroaddollarappreciationevents,butunlikecapitalflows,reactpositivelytochangesinriskaversionascapturedbytheVIX;ii)On-chaincross-borderflowsarepositivelycorrelatedwithimprovements

incrypto-specificsentiment(cryptofear&greed)

3

;iii)Yet,neithertraditional,globaldriversof

2CapitalinflowsintheBalanceofPaymentssensearemeasuredasnetchangesinnon-residents’holdingsofdomesticassets—–suchaspurchasesnetofsalesofadomesticbond–—insteadofthegrossinflowsresultingfromanon-residentpurchasingBitcoinfromaresident.SeeSection

3.1

3The

cryptofear&greed

indexcapturescrypto-specificsentiment.SeeSection

3.2.

4

capitalflowsnorsentimenttowardsBitcoinaffectoff-chainBitcoincross-borderflows;andiv)WhiledomesticfactorsplayalessprominentroleinoursamplebothforBitcoincross-borderflowsandforcapitalflows,inflationandaBitcoin-basedmeasureoftheparallelratepremiumarepositively

correlatedwithon-chaincross-borderflowsandoff-chaincross-borderflows,respectively.

Thecontributionsofourpapertotheliteraturearethreefold.First,weextendeffortstoidentifyandanalyzeBitcoincross-borderflowsbyexploringcomplementarydatasetsonglobalBitcoincross-borderflowsbasedonBitcointransactions.Huetal.(

2021

)developamethodtoidentifyBitcoinon-chaincross-borderflowsfromChina.Theyhighlightthaton-chain“uneconomical”flowsfromCNYtoUSDviaBitcoinweredrivenbycapitalflightmotives.Takingadvantageofrawtransactionoff-chaindatafromLocalBitcoinsandPaxful,GrafvonLuckneretal.(

2023

)developamethodologytoidentifycross-borderflowsbasedonmatchingtransactionsandtheunderlyingusedfiatcurrencies.TheyfindthatBitcoinhasbecomeanincreasinglyimportantchanneltosendremittancesandevadecapitalcontrolsinemergingmarkets.ContemporaneouswithourstudyandwithamainfocusonBrazil,Cardozoetal.(

2024

)compareauthorities’estimatesofcryptocross-borderflowsbasedonaggregateFXtransactiondatawithourestimatedLocalBitcoinsoff-chaindataaswellason-chaincross-borderdatafromcommercialvendorsCrystalandChainalysis.TheystressthatBrazilianBitcoincross-borderflowsarenotcorrelatedwithregularcapitalflowsnorremittances,andthattheyareasvolatileasregularBrazilianportfolioandFDIflows.OurstudyofthedriversofBitcoincross-borderflowsdiffersfromtheirsinthatweexploreacross-countrypanelsetup.Ourdirectusageofbothrawon-chainandoff-chainBitcoindataallowsustoidentifynotonlytheprosandconsofdifferentapproachestoestimatingBitcoincross-borderflowsbutalso

toprovideinsightfulstylizedfacts.

Second,wecomplementthesmallbutgrowingliteratureanalyzingthedriversofcryptoassetsandtheirrelationshipwithtraditionalfinancialassets.BenignoandRosa(

2023

)findthattheBitcoinpriceisorthogonaltoallintradaymacronewsthattheyconsider(excepttheCPI)aswellasunresponsivetounexpectedchangesintheshort-termratewhileitsreactiontonewsaboutthefuturepathofpolicyisnotrobust.Bourietal.(

2017

)investigatewhetherBitcoincanhedgeglobaluncertainty,measuredbythefirstprincipalcomponentoftheVIXindexesof14developedanddevelopingequitymarkets.TheyshowthatBitcoinpricesreactpositivelytouncertaintyatboth

higherquantilesandshorterfrequencymovementsofBitcoinreturns.Incontrast,Adrianetal.

5

(2022)andIyerandPopescu(

2023

)findstronginterconnectednessbetweencryptoassetpricesandequitypricesaswellastheVIX,suggestingthatcryptoassetsbehaveakintoriskyassets.Ourcross-borderresultssuggestthatBitcoinon-chainflowsarepositivelyrelatedwiththeVIX(theoppositeresponseofcapitalflows).ThisresultalsoconcurswithDiCasolaetal.(

2023

)whoshow

thatanincreaseintheVIXresultsinhigherBitcoin-fiatcurrencytradingvolumes.

Third,weprovideanuancedperspectiveonBitcoinandcapitalflowmanagementbydifferenti-atingbetweenon-chainandoff-chainflows.Analysesofthecross-borderusageofBitcoinhighlightdifferentactivitiesfrome-commerce(Polasiketal.,

2015

)tomoneylinkedtoillicitundertakings(RonandShamir,

2014

).

4

Nonetheless,independentoftheunderlyingactivity,cross-bordertrans-fersthroughBitcoinarethoughttobemotivatedbyhighcostsorgovernmentcontrolshinderingtransfersviatraditionalfinancialinstitutions(Biaisetal.,

2023

).Inthiscontext,theroleofcapitalcontrolshasbeenwidelyhighlightedintheliterature.Huetal.(

2021

)showhowcapitalflightmotivescanexplainon-chainBitcoinflowsinChina.GrafvonLuckneretal.(

2023

)stressthatoff-chainBitcointransactionscanfacilitatecross-bordercapitalflightandenablethecircumven-tionofcapitalflowmanagementmeasuresinseveralemergingmarkets.Inarecentstudy,GrafvonLuckneretal.(

2024

)arguethatunderlyingcapitalmovementsstilloccurviatraditionalchannelsandserveasliquidityprovidersforcryptoexchanges.Theyalsoproposethattherelativepriceofcryptoassetsinaneconomyprovidesinsightfulinformationaboutthestrengthofthedemandforcapitalflight.Wecontributetothisdebatebyprovidingcross-countrypanelevidencethatoff-chaincross-borderoutflowsarepositivelycorrelatedwithaBitcoinparallelpremiumwhichweinterpret

asabroaderproxyforexchangeratepressures,reflectingmacroeconomicimbalances.

ThepolicyimplicationsoftheusageofBitcoinforcross-bordertransactionsarepotentiallyamplebutnotdefinitiveatthisstage

.5

Thevolumeofbothon-chainandoff-chainBitcoinflowsappearsimportant.Nonetheless,Bitcoincross-borderflowsresponddifferentlythancapitalflowstotraditionaldrivers.Overall,thereisnotmuchevidencethatBitcoincross-borderflowshave

replacedexistingcapitalflowsatthisstage.Thisfeaturecouldbereadaspositive,inthesense

4InNovember2023,theU.S.TreasuryDepartmentfoundthatBinancefailedtoreporttransactionsassociatedwithillicitactivitiesandviolatedU.S.Anti-MoneyLaunderinglaws.

5Moregenerally,IMF

(2023a

)recommendssoundmonetarypolicyframeworks,notgrantingcryptolegaltenderstatus,integratingcryptoswithexistingcapitalflowmanagementregulations,andclarifyingtaxtreatmentofcryptostoavoidadverseimplicationsofcryptoassetsontheeconomy.FSBandIMF(

2023

)highlightaddressingdatagapsandenhancinginternationalcooperationandinformationsharingaskeystepstowardsaddressingpotentialmacroeconomicandfinancialstabilityriskfromcryptoassets.

6

thatcapitalflowsremainthekeyquantitativetransmissionchannelofglobalspikesinriskaversionand/orflighttosafetytriggers.Nonetheless,cryptomarketsareevolvingfast(e.g.,therecentauthorizationforBitcoinETFs)andbetterdataarekeytoevaluatingtheneedfortargetedpolicy

responsesinthefuture.

Theremainderofthepaperisorganizedasfollows.Section

2

discussesdifferentapproachesandunderlyingassumptionstomeasureBitcoincross-borderflows.Section

3

describesthedataandpresentsasetofstylizedfacts.Section

4

introducesourempiricalstrategyandpresentstheresultsoftheempiricalanalysis.Section

5

recapitulatesandprovidesafewadditionalremarkson

datalimitationsandevaluatingpolicyimplications.

2MeasuringBitcoinCross-BorderFlows

Cryptoassettransactionsoccuronmarketplaceswhicharedifferentfromthetraditionalfinancialsystemforcapitalflows.Inthissection,weprovideanoverviewofthemodalitiesofBitcointransactions,inparticularthedifferencesbetweenon-chainandoff-chaintransactions.Inaddition,wediscussthreealternativeapproachestomeasuringBitcoincross-borderflowsandzoominon

theunderlyingassumptionstoidentifytheresidencyoftheusers.

2.1ModalitiesofBitcoinCross-BorderFlows

TherearetwotypesofBitcointransactions:on-chaintransactions,whicharepermanentlyrecordedontheBitcoinblockchain(adistributedledgerwhichispublicasitisvisibletoallnetworkpar-ticipants)andareimmutable,andoff-chaintransactions,whichoccuroutsidetheblockchain.Itischallengingtocomparethevolumeofon-chainandoff-chaintransactions,partlyduetodiffi-cultiesinaccountingfordecentralizedtransactions.Nonetheless,basedonasampleof17cryptoexchanges,MakarovandSchoar(

2022

)findthattheBitcoinoff-chainvolumeissomewhatlarger

thantheon-chainvolume.

Figure

1

showsanillustrativeexampleofanon-chainBitcointransaction.Thesenderfirstini-

tiatesatransactionfromtheirdigitalwallet.Thetransactionrequiresinformationonthesender’s

7

address,therecipient’saddress,andtheamountofBitcoinbeingsent.Thesenderthensignsthetransactionusingthewallet’sprivatekey(a256-bitrandomnumberakintoadigitalpassword).Oncesigned,thetransactionisbroadcasttotheBitcoinnetworkforvalidation.Next,itisenteredintothemempool—thequeueofpendingtransactions.Minersthenvalidatetransactionsbysolv-ingcomplexcryptographicpuzzlesinexchangeforafee

.6

Validatedtransactionsarestoredinablockandconfirmedbyotherminersinthenetwork.Finally,thetransactionisdisseminatedtotheblockchainandtherebycompleted.Becauseoftheimmutabilityoftheblockchain,on-chain

transactionsofferahighdegreeofsecurity.

Figure1:Anillustrativeexampleofanon-chainBitcointransaction

Off-chaintransactionsoccuroutsidetheblockchainnetwork.Thesetransactionstakeplace,forexample,throughathirdparty,suchasacryptoexchangewhichverifiesthetransaction’slegitimacyandfacilitatesitscompletion.Sinceoff-chaintransactionsoccurwithoutarecordon

theblockchain,minersarenotrequiredtovalidatetransactions.Thisfeaturecanlowerfeesand

6Feesaredeterminedbytransactionsize(i.e.,datavolume)andblockspace(i.e.,thedesiredspeed.

8

speedupprocessingtimerelativetoon-chaintransactions.However,off-chaintransactionsprovidealowerdegreeofsecurity,reflectingthelackofapublicrecordofthetransactionandexchanges’

vulnerabilitytocryptohacks.

Therearethreemajortypesofcryptoexchanges:centralizedexchanges,peer-to-peer(P2P)exchanges,anddecentralizedexchanges.Centralizedexchangesaresimilartotraditionaltradingplatforms.Theymanageacentralorderbookandareresponsibleforthecustodyofusers’cryptoassets(usersdonothavedirectcontrolovertheirprivatekeys),security,maintenance,functional-ities,andtransactionapproval.CentralizedexchangesaretypicallyregulatedwhichimpliesthattheymustcomplywithKnowYourCustomerandAnti-MonetaryLaunderingregulations.Theseregulationsrequirethecollectionandstorageofcustomerinformation,includingidentitydocu-mentsandaddresses.Centralizedexchangesusuallyprovideauser-friendlyinterface,supportabroadersetofcryptoassetsandproductsfortrading,andaremoreliquidastheyhaveaccesstoanextensivepoolofbuyersandsellerssuchasregulatedmarketmakers.WhiletheratiooftradingvolumeonP2Panddecentralizedexchangestocentralizedexchangeshasincreased,centralizedexchangesstillaccountforasubstantialshareoftradingactivity(seeMakarovandSchoar(

2022

)and

CryptoSlate

).P2Pexchangesaredecentralizedexchangeswithadegreeofcentralization.Theyfacilitatedirecttradingbetweenclientswhocreatepubliclistingstomatchforatrade.P2Pexchangesfrequentlyprovideanescrowservicetosecurethetransaction.Decentralizedexchanges,incontrast,usesmartcontractstofacilitatedirectP2Ptradingwithoutanyintermediary.This

featurepreservesahighlevelofprivacyandreducestheriskofhacks.

2.2ApproachestoMeasuringBitcoinCross-BorderFlows

Approach1:MeasuringBitcoinCross-ExchangeFlowsbasedonExchangeWallets

WebeginouranalysisbymeasuringBitcoinflowsbasedonpubliclyavailableinformationfromon-chaintransactionsandwallets.Whilethisapproachonlyprovidescross-exchangeflows,ratherthancross-borderflows,itisusefultogainadeeperunderstandingofthetransaction-leveldataandthemethodologiesusedbycommercialdataproviders.Weobtaindataonon-chaintransactions

fromKondoretal.(

2021

)whoextracttheBitcointransactiondatafromtheblockchainwithan

9

open-sourceclient.Thedatacontaininformationonsendingandreceivingaddresses,theamount

ofeachtransaction,andatimestamp,amongotherdetails(Table

1

).

Figure

2

illustratesthematchingofon-chaintransactionswithexchangewallets.Thechallengeliesinidentifyingthelocationofthemarketparticipantsastheirwalletaddressesarepseudonymous.Nevertheless,forsomeaddressesinourdataset,wecanuse

WalletExplorer

whichcollectsdataonexchangewalletsfrompublicwebsitesandinternaltransactions.WethuscombineinformationfromtheblockchainandWalletExplorertoidentifytheexchangetowhichthewalletaddressbelongs.OurapproachissimilartoHuetal.(

2021

),althoughbroaderaswelookatexchangesavailableglobally.DuetothevastsizeoftheBitcoinuniverse,WalletExplorercannotprovideinformationonallavailablewalletsandexchanges.Itallowsustoidentifyabout5percentoftheaddressesinthedataset

.7

Theresultingexchange-leveldataset(Table

2

)contains1.6milliontransactionson

80differentexchanges.Aboutone-thirdofthesetransactionsoccuracrossexchanges.

Table1:Blockchain:rawdata

Thistableshowsthestructureoftheblockchainrawdata.

ID

Time

Amount(BTC))

Sendingaddress

Receivingaddress

746604

15Jun201119:12:17

50

1Q2TWHE3GMdB6BZKafqwxXtWAWgFt5Jvm3

12cbQLTFMXRnSzktFkuoG3eHoMeFtpTu3S

183

15Jun201121:30:05

40

12cbQLTFMXRnSzktFkuoG3eHoMeFtpTu3S

13HtsYzne8xVPdGDnmJX8gHgBZerAfJGEf

...

307112382

28Mar201807:51:51

1.583

1EiMLeHTKWHVWkUQVHkAc1mJsFR5zHoyBn

17XsVcCLzeRcx2TSyhX7QUvFX8FPrBmQBm

...

500663152

07Feb202007:45:17

0.0046

3Lbpq35oqmrQjrkiRX7FAQKL3NEwpgBwLt

3KGKozt3CE5xgbzwrnqxfPtJHF85WQn42D

Onesignificantbenefitofworkingwiththerawdatafromtheblockchainisthatitallowsustoexploreinmoredetailthenatureofon-chaintransactions.Additionally,thedatasetcapturestransactionsonmanyexchangesandassuchislikelymorerepresentativethantransactionsobtainedfromindividualexchanges.However,itischallengingtoassignexchangestocountriestoobtainBitcoincross-borderflowsratherthancross-exchangeflows.Thelocationofregistrationdoesinmanycasesnotprovidereliableinformationontheusersoftheexchangesbecause(i)manyexchangesareregisteredindomicileswithcrypto-friendlyregulationsorfavorabletaxregimesbutareglobalinnature

8

and(ii)usersmaytransactonexchangesincountriesotherthanwherethey

reside,especiallygiventhelargeconcentrationofexchanges.

7ThisisinlinewithLiangetal.(

2019

)whoshowthat,asof2018,WalletExplorerdetects4.3percentofaddressesintheBitcoinblockchain.

8Forexample,manyexchangesareregisteredintheSeychelleswherenoregulatoryframeworkiscurrentlyinplace.

10

Therearecurrentlytwotypesofdatasetswhichaddanadditionallayerlinkingexchangestocountries.Huetal.,

2021

userawon-chaindataandcategorizeexchangesintoChineseandnon-ChineseexchangestoidentifycapitalflightthroughBitcoinfromChina.Thefocusonasinglecountryfacilitatesassigningtheexchange.Thecommercialdataprovider

CrystalBlockchain

assignsexchangestocountriesbasedonwheretheexchangeisregistered,withalimitingfocusonexchangeswhichthedataproviderexpectstooperateinasinglecountryonly(foranapplicationofthisdataset,seeCardozoetal.,

2024

).Giventheglobalnatureofmanyofthelargeexchanges

inourdataset,suchasBinanceandKraken,westopshortofassigningexchangestocountries.

Table2:Blockchain:matchedcross-exchangetransactions

Thistableshowsthestructureoftheblockchainmatchedtransactions.Thedatasetcontains592,218cross-exchangetransactionsoverAug2nd,2011–Feb7th,2020.

Time

Amount(BTC))

Sendingaddress

Receivingaddress

2Aug201111:08:39

0.1405

Cavirtex

VirWoX

11Sep201107:25:02

7.9610

Cavirtex

BTC-e

...

28Mar201806:25:41

147.0000

MercadoBitcoin

Poloniex

...

07Feb202005:17:26

0.7613

Kraken

Poloniex

Approach2:MeasuringBitcoinCross-BorderFlowsbasedonExchangeWalletsandWebTraffic

AnalternativemethodtomeasureBitcoincross-borderflowslinkswalletaddressestoexchangesandadditionallyuseswebtrafficinformationtoinferthelocationofthemarketparticipants.Figure

3

illustrateshowthecommercialdataprovider

Chainalysis

implementsthisapproachtomeasureon-chaincross-borderflows.Inthefirststep,ChainalysisidentifiestheexchangestowhichwalletsbelongbasedonpublicdatafromtheblockchainandinformationfromexchangeswhichsubmittheirtransactiondatatoChainalysistooutsourcethemonitoringofcounterpartyrisk(ChainalysisKnowYourTransaction).Thisstepisconceptuallyequivalenttoourapproach1butbroaderinscope.Sinceexchangessubmitthedatausingpseudonymstoconcealtheirusers’identities,Chainalysiscannotdirectlyassignalocationtoeachwallet.Instead,inthesecondstep,theyassignflowstocountriesbasedonmonthlywebtrafficpatternsforeachexchange.Imagineastylizedexamplewithtwoexchanges,threecountries,andahypotheticaltransactionvolumeof100Bitcoinfromexchange1toexchange2onagivenday.BasedonthewebtrafficpatternshowninFigure

3,

Chainalysiswoulddistributethisdailytransactionvolumeasfollows:35BitcoinfromcountryX

11

Figure2:Matchingon-chaintransactionswithexchangewallets

tocountryX,35BitcoinfromcountryXtocountryZ,15BitcoinfromcountryYtocountryX,

and15BitcoinfromcountryYtocountryZ

.9

WeobtainBitcoincross-borderflowsfromChainalysisoverMarch2019–March2023.Theidentifiedflowsrelyontwokeyassumptions:(i)usersdonotmaskonlineactivitybyemployingvirtualprivatenetworks(VPNs)and(ii)transactionamountsare,onaverage,broadlyequalacrossusersindifferentcountries

.10

Thisapproachimprovesuponthecross-exchangeapproachbyusingdetailedexchangewalletinformation.Thus,itcapturesalargershareofthetotalBitcoinon-chain

transactionvolumeforala