12341–5236–7896510 9 30521 31 5 algorithmic trading in the...
TRANSCRIPT
– 1 –
sveriges riksbank economic review 2013:1
* TheauthorswouldparticularlyliketothankSusannaGrufman,MiaHolmfeldt,KajKvaavik,AnnaLidberg,JennyMannent,KjellNordin,MattiasPersson,KasperRoszbachandJonasSöderberg.
AlgorithmictradingintheforeignexchangemarketMaria Bergsten and Johannes Forss sandahl*TheauthorsworkattheFinancialStabilityDepartmentoftheRiksbank.
The foreign exchange market is an important part of the financial system and performs
important functions for the real economy. Recently, trading in this market has become
increasingly automated as the use of electronic foreign exchange trading and algorithms
has increased. In this article, we analyse how and why algorithmic trading is used on
the market for Swedish krona. We also analyse how this trading affects the functioning
of the market. From interviews with market participants, we reach the conclusion that
algorithmic trading is a component of competition-driven technological development.
On the whole, it should therefore contribute towards a more efficient foreign exchange
market with lower transaction costs. At the same time, it may entail certain risks for both
individual market participants and the financial system as a whole. Finally, we observe
that the market itself has worked out a practice for the functioning of algorithmic trading
in the foreign exchange market.
1.AweLL-FuNcTioNiNGFoReiGNexcHANGeMARKeTiSiMPoRTANTFoRTHeecoNoMy
iNGeNeRAL
Theforeignexchangemarketplaysanimportantroleforthebasicfunctionsofthefinancial
system:mediatingpayments,convertingsavingsintofundingandmanagingrisks(Sveriges
Riksbank,2013).Forforeignexchangetradingtofunctionsmoothly,itisimportantthatthe
variousrisksinthetransactionchainaremanagedassecurelyandefficientlyaspossible,
fromthemomentoftradeuntilthetransactionissettled.inaddition,theremustbemany
buyersandsellers,sothatcurrenciescanbesoldandboughtrapidly,whichistosaythat
marketliquidityisgood.
Afurtherconditionforanefficientmarketisthatthosetradingonithaveconfidence
initswayofworking.However,itdoeshappenonthefinancialmarketsthatinformation
onbothmarketparticipantsandfinancialinstrumentsisdeficientandunevenlyallocated,
whichcanhaveanegativeeffectonconfidence.iftheparticipants’confidenceinthe
functionalityofthemarketsdeclines,theirwillingnesstotradewitheachothermayalso
decline,makingitmoredifficulttorapidlybuyandsellfinancialinstruments.Amongother
– 2 –
sveriges riksbank economic review 2013:1
consequences,thismayaffecttheabilityofbanksandcompaniestomanagetheirrisksand
obtainfundingonthecapitalmarket.
Algorithmictradingispresentlyanormalwayoftradingonthefinancialmarketsand
involvesautomatedtradinginwhichordersandexecutionsarehandledbycomputers.it
isasub-categoryofelectronictrading,whichistradingviaanelectronictradingplatform.
Algorithmictradingincludeswhatisusuallyreferredtoashighfrequencytrading.High
frequencytradingisthusasub-categoryofalgorithmictradingandinvolvesplacingor
executingordersataveryhighfrequency.Theaimistogenerateprofitthroughthe
speedoftradingandtheinformationaladvantagecreatedbythetechnologyoverother
participantsonthemarket.
Themediaandacademicstudieshaveprimarilyfocusedonalgorithmictradingonthe
stockmarket.Thesediscussionsweregivenparticularimpetusbythetemporaryfallin
stockpricesof6May2010.onthisday,theuSequityindexDowJonesfellby9percent
withinthespaceofaboutfiveminutesbeforereturningtolargelythesamelevelaspriorto
thefallafewminuteslater.AccordingtoinvestigationsbytheuSauthorities,thisshort-
termcrash(alsocalledthe’flashcrash’)wascausedbyahedgefundthathadsoldavery
largeblockofindexfutures1throughacomputerprogrammedwithalgorithms,which
alsocreatedpressuretosellonothermarketsandforotherinstruments(u.S.commodity
FuturesTradingcommissionandtheu.S.Securitiesandexchangecommission,2010).
According,amongothers,toKirilenkoetal.(2011)highfrequencytradersalsocontributed
tothefallinstockprices.
inthesubsequentdiscussionsofalgorithmictrading–andhighfrequencytradingin
particular–thefocushasbeenonthepossibleeffectsonandrisksforthestockmarket’s
functioning,amongotherareas.Haldane(2011)consideredthatthenewcomputerised
tradingmayentail’fattertails’onthefinancialmarkets,whichistosaygreaterprobabilities
ofmajorfluctuationsinmarketprices.ioSco(2011)describedalgorithmictradingand
highfrequencytradingonthestockmarketandpointedouttheriskswiththeseformsof
trading.
However,relativelylittleattentionhasbeenfocusedontheforeignexchangemarket,
wherealgorithmictradingalsooccurs.Theforeignexchangemarketismuchlargerthan
thestockmarket(seechart1).Duringthetimeperiodinchart1,theRiksbank’smonetary
policycounterpartiesandcounterpartiesinforeignexchangetransactionshadadaily
turnoverofaboutSeK72billionontheforeignexchangemarketforSwedishkrona.2This
canbecomparedwithaboutSeK14billion,whichwasthedailyturnoverfortheStockholm
Stockexchange’s30largeststocks.
1 Afutureisabindingagreementbetweentwopartiestobuyorsellanunderlyingassetatacertainpriceatagivenpointintime.
2 SverigesRiksbank’sSeLMAstatistics,www.riksbank.se.
– 3 –
sveriges riksbank economic review 2013:1
0
50
100
150
200
250
300
Oct-05 Oct-06 Oct-07 Oct-08 Oct-09 Oct-10 Oct-11 Oct-12
Foreign exchange market, SEK Stock market, OMXS30
Chart 1. Daily turnover on the foreign exchange and stock markets, spotSEK billion
Sources: Reuters, EcoWin and the Riksbank.
Theforeignexchangemarketanditsfunctioningaresignificantforthesafeandefficient
functioningofthepaymentssystem,whichisoneoftheRiksbank’sobjectives.
consequently,itisinterestingtostudyhowalgorithmictradingaffectstheforeign
exchangemarket’sfunctioning.inthisarticle,weanalysehowandwhyalgorithmictrading
isusedontheforeignexchangemarketforSwedishkronaandhowitaffectsthe
functioningofthemarket.unlessstatedotherwise,inthisarticle,‘foreignexchange
market’refersexclusivelytothespotmarketforcurrencies.
FromtheRiksbank’spointofview,itisparticularlyinterestingtoknowwhether
algorithmictradingtakesplaceintradingwithSwedishkrona.consequently,wehave
interviewedsevenmarketparticipantswhoareactive,invariousways,ontheforeign
exchangemarketandintradingwithSwedishkrona.Theinterviewedparticipantsinclude
Swedishandforeignbanks,pensionfundmanagers,hedgefundsandnon-financial
companies.Thebanksincludedareresponsibleforabout50percentofturnoveronthe
foreignexchangemarketforSwedishkrona.3
weopenthearticlebydescribingafewcharacteristicsoftheforeignexchangemarket.
inaddition,wepresentthedifferentfunctionsofalgorithms,whichwedivideintofour
categories.wethengoontoanalysebothconceivablereasonsfortheuseofalgorithmic
tradingandtheriskswehaveidentifiedinouranalysis.Finally,weobservethatthemarket
itselfhasworkedoutapracticeforthefunctioningofthealgorithmictradingintheforeign
exchangemarket.
3 SverigesRiksbank’sSeLMAstatistics,www.riksbank.se.
– 4 –
sveriges riksbank economic review 2013:1
2.THeSTRucTuReoFTHeFoReiGNexcHANGeMARKeT
Theforeignexchangemarketisworldwideandisopenaroundtheclock.itscharacteristics
includetradingwithlargeamounts,alargenumberofparticipantsandtherapid
disseminationofpriceinformation.Theglobalturnoverinthismarketeverydayinvolves
amountscorrespondingtotensofthousandsofbillionsofSwedishkrona.4Theforeign
exchangetradingcanbedividedintotwosegments,tradingbetweenbanks(interbank)
andtradingbetweenbanksandtheircustomers.
Market makers have a central function on the foreign exchange market
currenciesaretradedoverthecounter(oTc),meaningthattradingdoesnottakeplaceon
aregulatedmarket.Tradingontheforeignexchangemarketusuallytakesplacethrough
whatareknownasmarketmakers.Amarketmakerisabankthatundertakestoquotebid
andaskpricesandtoguaranteethatitispossibletobuyandsellaminimumvolumein
oneorseveralcurrencypairs.inthisarticle,’marketmaker’referstobankswithformaland
informalcommitmentstoquotepricesinvariouscurrencypairs.
Foreignexchangetradingentailsrisksforthemarketmaker,forwhichthemarketmaker
compensatesitselfbyquotingthebidandaskpricessothatthedifferenceintheseistothe
marketmaker’sadvantage.Themarketmakerisvulnerabletoliquidityrisk,asitisforcedto
matchordersandinterestonthemarket.Liquidityriskistheriskofbeingaffectedbyaloss
asaresultoftherenotbeingenoughliquidityonthemarket.Poorliquiditymeansthatitis
difficulttofindbuyerswhenonewishestosellandvice-versa.
Themarketmakermayalsobevulnerabletomarketrisk,asthevalueofthepositiona
transactiongivesrisetomayvaryovertimeifthemarketmakerdoesnotenteranopposite
position.Marketriskistheriskthatfluctuationsinmarketpriceswillchangethemarket
valueofassetsandliabilitiesnegatively.
inaddition,themarketmakerisvulnerabletocreditriskwithregardtothecounterparty
untilthetransactionhasbeensettled.creditriskistheriskthatacounterpartyina
transactionwilllosetheabilitytorepayitsdebts.Thisriskcanbemitigatedbyestablishing
limitstorestrictexposuretoaparticularcounterparty.inforeignexchangetrading–andall
tradinginfinancialinstruments–creditrisksarisebut,inthiscontext,theseareknownas
counterpartyrisksandsettlementrisks.
The counterparty risk in foreign exchange (spot) trading is limited
Thecounterpartyriskinaforeignexchangetransaction(sometimesknownasreplacement
costrisk)istheriskthatthecounterpartywillfailonitsdebtsandthatthecurrencywill
consequentlyhavetobeboughtorsoldanewforanotherprice.counterpartyriskislimited
forspottransactions,asthetimebetweendateoftradeandsettlementissoshort,only2-3
days.undernormalmarketconditions,pricefluctuationsandcounterpartyriskareminimal,
4 SeealsoKingetal.(2011)foracomprehensivesurveyoftradingontheforeignexchangemarket.
– 5 –
sveriges riksbank economic review 2013:1
butduringcrisestheriskcanbegreaterasexchangeratescanfluctuatemore.while
counterpartyriskisoftendeemedtobesmall,settlementriskishigher.
oncertainfinancialmarkets,counterpartyriskisrestrictedbyacentralcounterparty
(ccP).5AccPactsasanintermediarybetweenthebuyerandsellerinthemanagement
ofasecuritiestransaction.Theoriginalcontractbetweenthebuyerandtheselleristhen
replacedbytwocontractswiththeccP.Thismeansthattheoriginalcounterpartiesinthe
transactionarenolongerexposedtoriskinrelationtoeachotherbutinsteadinrelationto
theccP.ontheforeignexchangemarketatpresent,thereisnoccPthatclearsforeign
exchangetransactionsinSwedishkrona.consequently,neitheristhereanypossibilityof
restrictingcounterpartyriskbytradingviasuchacounterparty.
The settlement risk in foreign exchange trading can be eliminated through CLS
Settlementriskcanbeeliminatedbysettlingforeignexchangetransactionsthroughthe
systemcontinuousLinkedSettlement(cLS).AllpaymentsincLSaresettledaccording
totheprincipleofpaymentversuspayment.Thisisachievedbythemembersmaking
paymentstoandreceivingpaymentsfromcLSaccounts–oneforeachcurrency–with
bothstagesoftheforeignexchangetransactionbeingsettledthroughtheseaccounts
simultaneously.onJanuary2013,cLSsettlestradesin17currencies,includingtheSwedish
krona.Statisticsfromthefirstsixmonthsof2012showthatabout90percentofall
foreignexchangetransactionsmadebythefourmajorSwedishbanksweresettledviacLS.
However,theproportionoftransactionssettledviacLSvariesbetweenthebanks.
There are different types of trading platforms on the foreign exchange market
Foreignexchangetradingtakesplaceeitherbytelephoneandchatorelectronicallyvia
tradingplatforms.Allofthetradingplatformsmediatecontactbetweenthebuyerand
theseller,andprovideinformationoncurrentpricesforvariouscurrencypairs.Accessto
informationondetailsineachcurrencyorder–forexampletheorderdepth–ismuchmore
limitedontheforeignexchangemarketthanonmarketplacesonthestockmarket.even
ifeachpricesetbyamarketmakeronlyappliestoaspecifiedordervolumeatacertain
pointintime,abuyerorasellercannotknowexactlythesizeoftheordervolumebehind
theprice.Thismeans,ononehand,thatitismoredifficulttocarryoutspeculativeforeign
exchangetradingonthebasisoftheavailableinformationonordervolumesoneachside
ofthemarketpricethanitisinstockexchangetrading.6ontheotherhand,itmeans
lowertransparency(comparedwiththestockmarket).Ahighleveloftransparencyis
fundamentallyimportantforthemaintenanceoflong-termconfidenceinfinancialmarkets.
5 ccPsareregulatedbytheeuregulationeMiR(theeuropeanMarketinfrastructureRegulation)onoTcderivatives,centralcounterpartiesandtraderepositories,whichcameintoforceon16August2012.
6 itshouldbenotedthathiddenliquidityalsooccursinstockmarkettrading,inwhatareknownas’darkpools’.
– 6 –
sveriges riksbank economic review 2013:1
Tradingplatformscanbedividedintothreedifferenttypes:
• Inter-dealer electronic broking platforms.Theseplatformsweredevelopedinthe
1990sandareregarded,accordingtotheBankforinternationalSettlements(BiS,
2010),asthedominantsourceofinterbankliquidityontheforeignexchangemarket.
Theymediateinformationonvariousmarketmakers’indicativeprices.eBSand
Reuters,basedinLondon,arethetwodominantplatformswithinthiscategory.
• Multi-bank platforms. Theseplatformsarealsoknownasmulti-bankecNs
(electroniccommunicationnetworks).Theywerecreatedinthefirstdecadeofthis
centuryandresemblethepreviouscategoryinthattheymediateseveralmarket
makers’prices.onedifferenceisthattheyhavefreeraccessregulationsformarket
makers,whichmakesiteasierformarketmakerstojointheseplatforms.Another
differenceisthattheyarelargelyusedoutsidetheinterbankmarket,whichisto
saybymarketparticipantsthatarenotbanks.TheuSplatformsFxAll,currenex,
HotspotFx,StateStreetandFxconnectareexamplesofthistypeoftrading
platform.Therearealsoplatformsthatprovidestandardisedalgorithmictrading
functionsasaservice.currenexisonesuchplatform.
• Single-bank platforms.Thistypeofplatformisrunbyanindividualbank.The
platformmediatesonlytheindividualbank’sownpricesforvariouscurrencypairs,
unlikethetradingplatformsdiscussedabove,whichmediateseveralmarketmakers’
prices.inSweden,SeBhasaplatformofthistype,SeBTradingStation.other
examplesofbankswithsuchplatformsareJPMorgan,DeutscheBankandcitibank.
Aparticipantontheforeignexchangemarketcaneitherbedirectlyconnectedtoaplatform
orindirectlyconnectedviawhatisknownasaprimebroker.Abankthatisaprimebroker
allowscustomerstotradeontheforeignexchangemarketinthatbank’sname.Thismeans
thatthebankalsotakescreditriskexposurestothecustomerstradinginthismanner.
3.DiFFeReNTPuRPoSeSFoRALGoRiTHMSoNTHeFoReiGNexcHANGeMARKeT
Severalearlierstudieshaveobservedthatalgorithmictradingexistsontheforeign
exchangemarket.TheBiS(2011)claimsthathighfrequencytradingismostwidespreadin
tradinginthelargestandmostliquidcurrencypairssuchaseuR/uSD,GBP/uSDand
uSD/JPy,butthatitcouldspreadtotradinginlessliquidcurrencypairs.Kingetal.(2011)
alsodrawtheconclusionthathighfrequencytradingisresponsibleforalargeshareof
turnoverinthelargestandmostliquidcurrencypairs.
Themarketparticipantsweinterviewedinourstudyestimatethatalgorithmictrading
constitutes0-40percentoftradingincurrencypairsinwhichtheSwedishkronaisone
ofthecurrencies,and0-75percentoftradinginothercurrencypairs.Theseproportions
areassessedtobethehighestamongmarketmakers(20and39percentrespectively),
whiletheyareassessedtobesignificantlylowerbyotherparticipants(4and2percent
respectively).
– 7 –
sveriges riksbank economic review 2013:1
ontheforeignexchangemarketforSwedishkrona,ourinterviewshaveshownthat
thealgorithmsaremainlyappliedinspottrading,andtoalesserextentinforward7trade.
Aboveall,thisissaidtobebecausespottradingisthemoststandardisedformofforeign
exchangetrading,whileforwardtrademayinvolvespecificcharacteristicsandconditionsin
thetradedcontracts.Automatingforwardtradeisthusmoredifficultthanautomatingspot
trading.
Fromourinterviews,wedrawtheconclusionthatalgorithmsontheforeignexchange
marketforSwedishkronafulfilfourdifferentmainfunctions:executingorders;quoting
prices;managingrisks;andspeculatingandexploitingpricedifferenceswhenarbitrage
opportunitiesarise.
Order execution
Algorithmsareusedtoautomaticallyexecuteordersonthebasisofoneormore
predeterminedcriteria.Afewcommontypesofalgorithmsarethosethatspreadout
executionoverapredeterminedtimeperiod,executeanorderinbatchesofasmaller
amount,orexecuteanorderforacertainprice.Someallowarandomgeneratorto
determinetheamountconcernedoneachoccasionandwhenthetransactionistobe
executed.
Banksandtradingplatformsaredevelopingalgorithmsasapartoftheirrangeof
competitiveservicestoattractcustomersandultimatelygeneraterevenue.Forthe
customers,thealgorithmshaveentailedstandardisedandautomatedmethodsfor
minimisingmarketmanipulation,whentheycarryoutforeignexchangetransactionsand
enterordersanonymously.
Algorithmsalsomakeiteasierfortheusertoreportexactlyhowanorderhasbeen
executed,asallinformationisautomaticallystored.However,theassetmanagersandnon-
financialcompanies’treasurydepartmentsinterviewedsaythattheirusageofalgorithms
onlyamountstoatmost10percentofthetotalforeignexchangetransactionsinSwedish
krona.
Price quotation
certainalgorithmshavebeendevelopedtoautomaticallyquotepricesindifferentcurrency
pairs,sometimeswithveryhighfrequencies.withtheassistanceofsuchalgorithms,banks
cancontinuallyupdatetheirpricesfordifferentcurrencypairsonthetradingplatforms.
Thealgorithmsaresettoquotemarketpricesandactonthebasisoftheconditionsand
variablesforwhichtheyareprogrammed.consequently,undertherightconditions,they
makeitpossibleforthebankstorapidlyquotemarketprices,atthesametimeasmanual
involvementisminimisedandthenumberofworkinghoursperquotedpriceisreduced.
7 Aforwardisabindingagreementbetweentwopartiestobutorsellanunderlyingassetatacertainpriceatagivenpointintime.
– 8 –
sveriges riksbank economic review 2013:1
Risk management
Somebanksusealgorithmstomanagethemarketrisksarisinginforeignexchange
transactions.oneexamplewouldbeifatransactiongeneratedbyacustomerledtoabank
buyinguSD/SeK,followingwhichanalgorithmsoldtheequivalentamountinuSD/SeKto
eliminatethemarketrisk.Suchalgorithmscanthusbesaidtoautomaticallymake
transactionstoachievethedesiredriskexposure.Furthermore,thesealgorithmsfreeup
resourcesforthebanks,asriskmanagementnolongerneedstobeconductedmanually.
Speculation and exploitation of arbitrage opportunities
Therearealsoalgorithmsthatcanbeusedtoautomaticallytakepositionstoyieldreturn.
Thedecisioncriteriathatthealgorithmsareprogrammedforaredefinedbythealgorithm’s
owners.Thesecriteriamay,forexample,bebasedoncorrelationsbetweenassettype
anddifferentcurrencypairsorotherhistoricalrelationships.Forexample,aspeculative
algorithmmaybeprogrammedtoselldollarswhenthepriceofgoldrises,asthehistorical
relationshiphasdemonstratedanegativecorrelationbetweentheseprices.Thepositions
generatedbyspeculativealgorithmsarethusassociatedwithmarketrisk.
Therearealsoalgorithmsthatdonottakeanymarketriskbuttakepositionswhen
arbitrageopportunitiesarisethroughinconsistenciesonpricing.Suchalgorithmscan,for
example,beprogrammedtorapidlyidentifyandactondifferencesinpricebetweena
derivativeanditsunderlyingassets,orbetweendifferentmarketplaces.Thisoftenmeans
thatinconsistenciesinpricingarerapidlyeliminated.
certainalgorithmsarehighfrequencytraders,whichistosaythattheytradeatavery
highfrequencyandbenefitfromtemporarypossibilitiestomakerisk-freeoralmostrisk-free
profitsthattheycanexploitduetotheirrapidity.Byexploitingavailableinformationearlier
andmorerapidlythanothermarketparticipants,theycanactfasteronnewinformation
andthusgenerateyield.However,noneofthemarketparticipantswehaveinterviewed
workswithsuchalgorithms.
4.ALGoRiTHMicFoReiGNexcHANGeTRADiNGLeADSToiNcReASeDeFFicieNcy
Thereareseveralconceivablereasonsfortheuseofalgorithmictradingonthefinancial
markets.oneoftheseisstructuralchangessuchaschangedregulationsformarketplaces,
whichHaldane(2011),amongothers,haveindicatedforthestockmarket.Forexample,the
conditionsforalgorithmictradingareinfluencedbywhethertradingisconductedononeor
severalmarketplaces,andbyaccesstodifferenttypesofmarketinformation.Thesefactors
can,inturn,dependontheregulationssurroundingtrading.
Anotherreasonforalgorithmictradingistechnologicaldevelopmentonthefinancial
markets.Algorithmictradingmeansthatthebankscanstreamlinetheirforeignexchange
tradingandprovidecompetitiveservicestotheircustomers.AccordingtotheBiS(2011),
highfrequencytradingontheforeignexchangemarket(forexample)isanincreasingly
commonformoftradingandpartofadevelopmentoftechnologythatisincreasingthe
– 9 –
sveriges riksbank economic review 2013:1
useofelectronictrading.Algorithmictradingisalsoreplacingoftenoldertechnologyby
fulfillingthesamefunctionsmoreefficiently.Forexample,Menkveld(2013)saysthat
highfrequencytradersonthestockmarkettoagreatextentresembletraditionalhuman
marketmakers.However,itislikelythattradingbytelephonewillcontinuetobeneededto
maketransactionswithlargeamounts.Theinterviewedmarketparticipantsdeemthatthe
personalrelationshipsbuiltupbyforeignexchangetradersbytelephonewillcontinuetobe
valuable,astheseareconsideredtoleadtofavourablebusinessconditionsandthusbetter
pricesinmajortransactions.
Technologicaldevelopmentisthusastrongcontributoryfactorinthegrowthofvarious
typesofalgorithmictrading.Technologicaldevelopmentoftenleadstoareductionin
thenumberofhoursneededtosupplyaservice,whichshouldmeanthatthemarketis
functioningmoreefficiently.Amongothercharacteristics,anefficientmarketissignifiedby
lowtransactioncostsandthepresenceofassetswithhighliquiditysothattheflowoftrade
canfunctionsmoothly(Fama,1970).
5.ALGoRiTHMicFoReiGNexcHANGeTRADiNGMAyeNTAiLRiSKS
However,thereareanumberofconceivableriskswithalgorithmictrading:operational
risks,riskswithregardtotheliquidityofthemarket,risksofincreasedvolatilityand
temporarycrashessuchasthe’flashcrash’in2010,riskstotheaccesstoinformationand
theconfidenceinthemarket,andrisksofbarriersofentryformarketmakers.Theserisks
existbothfortheindividualmarketparticipantandforthemarketasawhole.Theycan
alsovarydependingonhowliquidthecurrenciesare.Thisisbecausetheliquidityinthe
tradeofacurrencyinfluencesthechoiceofappropriatetradingstrategyandthedegreeto
whichtradingcanbeautomated.itisthereforeimportanttomakeadistinctionbetween
tradinginlessliquidcurrenciesandtradinginthemostliquidcurrencies.
AccordingtotheBiS(2010),theSwedishkronawasthecurrencywiththeninthhighest
turnoverinApril2010,with1.1percentoftheforeignexchangemarket’stotaldaily
turnoverofuSD4000billion.undernormalconditions,theliquidityoftheSwedishkrona
ishigh,meaningthatmarketparticipantscanmakeforeignexchangetransactionswithout
influencingexchangeratesagainstothercurrenciestoanygreatextent.However,during
certain,oftenturbulentperiods,theremayariseashortageofliquidityonthemarketfor
theSwedishkrona.Thismaymakeitdifficulttomaketransactionsatareasonableprice,as
thereissuddenlyverylittleinteresttotradeamongmarketparticipants.oneexampleof
thisisfollowingtheoutbreakofthefinancialcrisisin2008,whenthemarketparticipants
experiencedinsufficientliquidityonthemarket(SverigesRiksbank,2009).
Operational risks
Thealgorithmictradingmaycontributetoadecliningmanualhandlingoftransactions
andinteractionbetweenindividualstradingonthefinancialmarkets.Thismayhaveboth
benefitsanddrawbacks.oneadvantageisthattradingislessfrequentlyaffectedby
– 10 –
sveriges riksbank economic review 2013:1
temporary,irrationalandoccasionallyimpulsivebehaviour.insteadofactingonthebasisof
emotions,thealgorithmsalwaysmakeassessmentsonthesamebasis,basedontheirpre-
programmedcriteria.
ontheotherhand,onedisadvantageofthedecreaseofmanualtradingmaybethat
algorithmsareunabletoquicklyassessscenariosforwhichtheyarenotprogrammed.
Forexample,ifamarketshouldsuddenlyloseliquidityorbesubjectedtoamajorfallin
prices,thereisariskthatthealgorithmwillnotchangeitsbehaviourbutwillcontinueto
tradeorquoteprices.Suchbehaviourmayleadtolossesforthemarketparticipantusing
thealgorithm.oneexampledatesfromAugust2012,whentheuSfinancecompany
KnightcapitallostuSD440millionafterthecompany’sincorrectlyprogrammedalgorithm
conductedloss-makingtransactionsin148differentstocks(FinancialTimes2,2012).
Market liquidity risks
Tomanageoperationalrisks,banksandothermarketparticipantsoftenusedifferent
typesofcircuitbreakersandothermechanismstodisablethealgorithmsinturbulent
situations.Thishelpsthemarketparticipantstoavoidlosses,butalsoleadstothesudden
disappearancefromthemarketplaceoftheliquidityotherwiseprovidedbythealgorithms’
trading.Kingetal.(2011)pointoutthatthereisuneasethathighfrequencytradingmarket
makersontheforeignexchangemarketmaychoosetostoptradingwhenturbulencearises
inmarketpricing.Thereareseveralstudiesofthisphenomenononthestockmarket.For
example,Johansson(2012)pointsouttheriskthatalgorithmictradersonthestockmarket
maystoptradingunderstressedmarketconditions.Suchbehaviourcanbothcreateand
worsenuneaseintrading,whichcanhaveanegativeeffectonmarketliquidity.
Severalstudiesshowthatalgorithmictradersusuallymakeapositivecontributionto
liquidityonthemarket.chaboudetal.(2011)indicatethatalgorithmictradingentails
increasedliquidityontheforeignexchangemarket,intermsofgreaterorderdepth.As
regardsthestockmarket,Hendershottetal.(2010)showthatalgorithmictradingon
thestockmarkethasgenerallycontributedpositivelytothemarket’sliquidity,butthat
anotherrelationshipcannotberuledoutunderstressedmarketconditions.Thebehaviour
ofalgorithmictraderscanalsobeexpectedtodiffer,dependingonwhichstrategythey
choosetoapply.HagströmerandNordén(2012)dividehighfrequencytradersintomarket
makersandopportunistictraders.Theyshowthatthemarketmakersarethedominant
highfrequencytradersontheSwedishstockmarketandthattheycontributepositivelyto
liquidityonthemarket.Gomberetal.(2011)alsopointoutthatthetraders’strategiesform
areasonablestartingpointfortheassessmentofhighfrequencytrading.
Risks of increased volatility and temporary crashes
Algorithmictradingthattakesplaceforspeculativepurposesortoexploitarbitrage
opportunitiescanaffectthepricingofcurrenciesandotherassetclasses.Bicchettiand
Maystre(2012)showthatinconsistenciesinpricingbetweendifferentassetclassesand
– 11 –
sveriges riksbank economic review 2013:1
currencies,andbetweenderivativesandtheirunderlyingassets,areexpectedtobecome
fewer.Suchadevelopmentcanbeseenaspositive,asitwillleadtoamorearbitrage-free
market.chaboudetal.(2011)alsoshowthatalgorithmictradinghasledtofewerarbitrage
opportunities,asthemarketreactsmorequicklytonewinformation.Thisdevelopment
couldalsobeimaginedtoleadtoincreasedco-variancebetweendifferentfinancial
instrumentsandbetweendifferentassetmarkets,asnewinformationismorerapidly
reflectedinprices.
Authoritiesandresearchershaveaskedwhetheralgorithmictradingmayleadto
increasedvolatilityinassetprices,whichcouldimpairthefunctioningofthefinancial
markets.Severalstudieshavebeenmadeofthisrelationshiponthestockmarket.Boehmer
etal.(2012)indicateapositiverelationbetweenalgorithmictradingandvolatilityonthe
stockmarket,thatisthatalgorithmictradingcontributestoincreasedvolatility.Atthesame
time,therearestudiesthatindicatetheopposite,forexampleHagströmerandNordén
(2012).Broogaardetal.(2012)cannotdrawanyconclusionsindicatinganyclearrelation
betweenhighfrequencytradingandvolatility.Asregardstheforeignexchangemarket,
chaboudetal.(2011)shownoclearsignsthatalgorithmictradinginfluencesvolatility.in
general,thereisthusalackofsupportforaclearrelationbetweenalgorithmictradingand
volatility.
However,theGovernmentofficeforScience(2012)demonstratesthatalgorithmson
thestockmarketcanactinakindoftechnologicalherdbehaviour,socalledfeedback
loops,whichcancreaterapidpricefluctuations.ontheseoccasions,algorithmsactby
sellingorbuyingsimultaneously,whichaffectsthepriceoftheassetinquestioneither
upwardsordownwards.Asregardstheforeignexchangemarket,chaboudetal.(2011)
showthatthealgorithmictraders’strategiesshowmoreofacorrelationthandothose
ofthehumantraders.Asdifferentfinancialinstrumentsandmarketsco-varymore,a
temporaryandexaggeratedmarketfluctuationononeassetmarketcanspreadtothe
foreignexchangemarket,whichcancreateuncertaintyonseveralfinancialmarkets.Sucha
scenariowouldbeparticularlyseriousforacurrencyliketheSwedishkrona,whichdoesnot
belongtothemostliquidcurrencies,asitmayleadtoasuddenreductionofliquidityonthe
market.
However,accordingtotheBiS(2011),theriskoftemporarycrashesissmalleron
theforeignexchangemarketthanitisonthestockmarket.Thisisbecausetheforeign
exchangemarketdiffersfromthestockmarketinseveralrespects.onefundamental
difference,accordingtothereport,isthatacurrencyispricedinrelationtoanother
currency,whileastockpriceisanabsoluteprice.whenthepriceofacurrencyfalls,one
partyloses,whiletheotherpartygains.Thismeansaredistributionofwealthratherthan
areduction,whichistheeffectwhenstockpricesfall.Afurtherdifferenceisthatthereis
ratherstabledemandforforeignexchangerelatedtocross-bordertradeandfinancialflows.
Thereisnoequivalentflowonthestockmarket.
– 12 –
sveriges riksbank economic review 2013:1
Risk related to the access to information and the confidence in the market
Anunevendistributionofinformationmayleadslowerinvestorstomakelessadvantageous
investmentdecisionsthanhighfrequencytraders,astheydonotreceiveinformationas
rapidly.Biaisetal.(2011),Jarrowetal.(2011)andMcinishandupson(2012)discussthis
andnotethathighfrequencytradershavevariousmeansofprofitingfromandactingon
informationmorerapidlythanothermarketparticipants.Finansinspektionen(2012)and
Johansson(2012)discusstheriskofimproperbehaviourandreachtheconclusionthathigh
frequencytradinggivesrisetonewbehaviouramongmarketparticipantsthatmayentail
ariskofmarketabuse.inthattradingtakesplaceatahigherspeed,thusthepossibilityof
improperlyinfluencingthemarketpricetoatrader’sownadvantage.
However,therearealsostudiesthatshowthathighfrequencytradersquotethemost
favourablepricesonthemarketandthuscontributetowardsefficientpriceformation.
ThesestudiesincludeHasbrouckandSaar(2011).Asregardstheforeignexchangemarket,
chaboudetal.(2011)simultaneouslyshowthatthepeoplewhoaretradingstillhavegood
accesstoinformationandthattheirtradinginfluencespriceformationmorethanthatofthe
algorithmictraders.
Risks of barriers of entry for market makers
Developingalgorithmsrequiresmoreorlesscomprehensiveinvestmentsintechnology.
Thosebanksinvestinginelectronicsystemsandalgorithmsdosopartlytostreamlinetheir
operationstogaineconomiesofscaleoverotherbanks.AccordingtoKingetal.(2011),this
isanexampleofareasonforthebankstochoosetodeveloptheirowntradingplatformsfor
customertradingtoagreaterextent.Thismakesitpossibleforthebankstointernallymatch
cashflowsindifferentcurrencies.Atthesametime,asthismayhaveadvantagesforthe
individualbanks,Kingetal.(2011)observethatanincreasedfragmentationofthemarket
may,overthelongterm,occasionallycreateliquidityshortagesontheinterbankmarket.
Thereisalsoariskthattheneedforcomprehensiveinvestmentswillcreatebarriersof
entryontheforeignexchangemarket.Thisisbecausesmallbankscanlackboth
competenceandresourcestoinvestinthetechnologyneededtocompetefortrade.King
etal.(2011)statethatsuchbarriersofentryhaveledfewerbankstochoosenottobeactive
inthelargest,mostliquidcurrencypairs.Atthesametime,Kingetal.(2011)saythatthe
smallerbanksarestilltradinginthesmallerlocalcurrencies,andthattheyarethereby
becomingexpertsandearningmoneyontradewiththesecurrencies.
6.ScoPeFoRSeLF-ReGuLATioNiNTHeFuTuRe
ingeneral,theforeignexchangemarketdiffersfromotherfinancialmarkets,asforeign
exchangetradingandtheclearingandsettlementofforeignexchangetransactionsare
notregulatedbyauthorities.Atpresent,thereisnoSwedishlegislationtoregulatetrading
ontheforeignexchangemarket.However,itisinthemarket’sowninteresttocounteract
tradingthatmaydamagethefunctioningofthemarket.itisthereforeself-regulatingin
– 13 –
sveriges riksbank economic review 2013:1
thatthemarketitselfhasformulatedapracticeforhowtradefunctions.Neitheristhereany
globallegislationforthetradingplatformsexistingforforeignexchangetrading,whichisa
differencetohowthestockmarketfunctions(BiS,2011).
Sincethefinancialcrisis,workhasstartedontheformulationofseveralnewregulatory
frameworks.However,undertheircurrentformulations,noneoftheseregulatory
frameworksspecificallyaffecttheforeignexchangemarket.intheeu,MiFiDii/MiFiR8
willentailexpandedregulationsfortheequityandfixed-incomemarkets(bothspotand
derivativemarkets),amongotherareasregardingincreasedtransparencyandrequirements
coveringwhichinstrumentsmaybetradedonaregulatedmarketplace.Thereareatpresent
noequivalentrulesfortheforeignexchangemarket.MiFiDiiwillprobablyalsoincluderules
placingrequirementsonsecuritiescompaniesusingalgorithmictrading(articles17and51
oftheeuropeancommission,2011).inasimilarway,theso-calledDodd-FrankActinthe
unitedStateswillprobablyincludenewrulesonalgorithmictrading.itisnotcompletelyclear
whetherandhowsecuritiescompaniesusingalgorithmictradingincurrencywillbeaffected
bytheseregulations.Thelegislationisprimarilyintendedtoaddressanynegativeeffectsof
highfrequencytradingonthestockmarket.
Recently,certaintradingplatformshavetriedtocounteractelementsofalgorithmic
trading,asithasbeenseenasdisruptivetothefunctioningoftheforeignexchangemarket.
oneexampleofsuchaplatformiseBS.Themeasuresadoptedincludeanincreasedtick
size9,aswellasareductionofthenumberofbidsatradermayenterwithoutactually
trading.Tradingbehaviourisdeemedtohavebecomehealthierduetothesemeasures,
withouttradingvolumeshavingbeenaffectednegatively(FinancialTimes1,2012).There
willcontinuetobescopeforthiskindofself-regulationtocounteractelementsofdisruptive
tradingontheforeignexchangemarket.
8 MiFiDstandsfortheMarketsinFinancialinstrumentsDirective.MiFiRstandsforMarketsinFinancialinstrumentsRegulation.Theseregulationswillprobablybeimplementedin2015.
9 Ticksizeisthesmallestpossiblepricechange.
– 14 –
sveriges riksbank economic review 2013:1
References
BiS(2010).TriennialcentralBankSurvey,Reportonglobalforeignexchangemarketactivityin2010.MonetaryandeconomicDepartment.1December2011.
BiS(2011).High-frequencytradingintheforeignexchangemarket.Report.Marketscommittee.27September2011.
Biais,B.,Foucault,T.,&Moinas,S.(2011).equilibriumHighFrequencyTrading.internationalconferenceoftheFrenchFinanceAssociation(AFFi).May2011.
Bicchetti,D.,&Maystre,N.(2012).Thesynchronizedandlong-lastingstructuralchangeoncommoditymarkets:evidencefromhighfrequencydata.MunichPersonalRePecArchive.PaperNo.37486.unitedNationsconferenceonTradeandDevelopment–uNcTAD.20March2012.
Boehmer,e.,Fong,K.,&wu,J.(2012).internationalevidenceonalgorithmictrading.AFA2013SanDiegoMeetingsPaper.14March2012.
Brogaard,J.A.,Hendershott,T.,&Riordan,R.(2012).HighFrequencyTradingandPriceDiscovery.workingPaper.30July2012.
chaboud,A.,chiqouine,B.,Hjalmarsson,e.,&Vega,c.(2011).RiseoftheMachines:AlgorithmicTradingintheForeignexchangeMarket.FederalReserveBoardinternationalFinanceDiscussionPaperNo.980.13June2011.
europeancommission(2011).ProposalforaDirectiveoftheeuropeanParliamentandofthecouncilonmarketsinfinancialinstrumentsrepealingDirective2004/39/ecoftheeuropeanParliamentandofthecouncil(Recast).20october2011.
Fama,e.F.(1970).efficientcapitalMarkets:AReviewofTheoryandempiricalwork.TheJournalofFinance,vol.25,no.2,p.383-471.
FinancialTimes1(2012).HFTcurbsledto“healthier”trading.Article.13November2012.
FinancialTimes2(2012).HFTtradersunderfireafteralgoglitch.Article.2August2012.
Finansinspektionen(2012).investigationintohighfrequencyandalgorithmictrading.Report.21February2012.
Gomber,P.,Arndt,B.,&Lutat,T.u.(2011).High-FrequencyTrading.workingPaper.Goetheuniversität.
GovernmentofficeforScience,London(2012).Foresight:TheFutureofcomputerTradinginFinancialMarkets.FinalProjectReport.
Hagströmer,B.,&Nordén,L.(2012).Thediversityofhighfrequencytraders.workingPaper.StockholmuniversitySchoolofBusiness.
Haldane,A.G.(2011).Theracetozero.SpeechbyAndrewG.Haldane.Bankofengland.8July2011.
Hasbrouck,J.,&Saar,G.(2011).Low-LatencyTrading.workingPaper.May2011.
Hendershott,T.,Jones,c.M.,&MenkveldA.J.(2010).DoesAlgorithmicTradingimproveLiquidity?TheJournalofFinance,vol.66,no.1,s.1-33.wFA2008Paper.30August2010.
ioSco(2011).RegulatoryissuesRaisedbytheimpactofTechnologicalchangesonMarketintegrityandefficiency.TechnicalcommitteeoftheinternationalorganizationofSecuritiescommissions.consultationreport.July2011.
Jarrow,R.,&Protter,P.(2011).ADysfunctionalRoleofHighFrequencyTradinginelectronicMarkets.JohnsonSchoolResearchPaperSeriesno.08-2011.29June2011.
Johansson,N.(2012).Högfrekventochalgoritmiskhandel–enöversikt(“Highfrequencyandalgorithmictrading–anoverview”).ReporttoFinansinspektionen.21February2012.
– 15 –
sveriges riksbank economic review 2013:1
King,M.R.,osler,c.,&Dagfinn,R.(2011).Foreignexchangemarketstructure,playersandevolution.NorgesBankResearchDepartment.workingPaper.No.2011/10.14August2011.
Kirilenko,A.,Kyle,A.S.,Samadi,M.,&Tuzun,T.(2011).TheFlashcrash:TheimpactofHighFrequencyTradingonanelectronicMarket.workingPaper.18January2011.
Mcinish,T.,&upson,J.(2012).StrategicLiquiditySupplyinaMarketwithFastandSlowTraders.workingPaper.1March2012.
Menkveld,A.J.(2013).HighFrequencyTradingandtheNew-MarketMakers.eFA2011Paper.AFA2012Paper.12February2013.
ProfitandLossMagazine(2011).Voiceofthemarkets,byJamesSinclair.Article.23September2011.
SverigesRiksbank(2013).TheRiksbankandfinancialstability.Report.4February2013.
SverigesRiksbank(2009).Swedishmarketparticipants’viewsofrisksandthefunctioningoftheSwedishfixed-incomeandforeignexchangemarkets.Report.20November2009.
u.S.commodityFuturesTradingcommissionandtheu.S.Securitiesandexchangecommission(2010).FindingsregardingthemarketeventsofMay6,2010.Report.30September2010.