Does Financial Stress Affect Commodity Futures Traders' Positions?

Finance and Economics Discussion Series Federal Reserve Board, Washington, D.C. ISSN 1936-2854 (Print) ISSN 2767-3898 (Online) Does Financial Stress Affect Commodity Futures Traders’ Positions? Shengwu Du, Travis D. Nesmith, Yanggen Heppe 2025-082 Please cite this paper as: Du, Shengwu, Travis D. Nesmith, and Yanggen Heppe (2025). “Does Financial Stress Affect Commodity Futures Traders’ Positions?,” Finance and Economics Discussion Series 2025-082r1. Washington: Board of Governors of the Federal Reserve System, https://doi.org/10.17016/FEDS.2025.082r1. NOTE: Staff working papers in the Finance and Economics Discussion Series (FEDS) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or the Board of Governors. References in publications to the Finance and Economics Discussion Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers.

Does Financial Stress Affect Commodity Futures Traders’ Positions? ShengwuDu1 TravisD.Nesmith1 YanggenHeppe2 1QuantitativeRiskAnalysis,FederalReserveBoard, Abstract Washington,D.C.,U.S.A FinancialstresscanimpacttradingbehaviorintheU.S.commodityfuturesmarkets.Toclarifythe 2DepartmentofEconomics,UCSanDiego,La impact,westudyabsolutechangesandrelativeexposuredynamicsintraders’positionsduringtwo Jolla,C.A.,U.S.A recentcrises:the2008GlobalFinancialCrisis(GFC)andtheCOVID-19pandemic.Thenatureof Correspondence thesetwocrisesareverydistinct,andwefindthattradersbehavedquitedifferently.Thecommodity ShengwuDu marketcollapseduringthe2008GFCfollowedtheclassicpatternofaspeculativebubble;speculators, Email:shengwu.du@frb.gov includingfinancialinstitutionsandmoneymanagers,rushedtoclosetheirlongpositionsincommodity TravisNesmith futureswhilecommodityproducersorhedgersactivelyfacilitatedthesetrades.Consequently,the Email:travis.d.nesmith@frb.gov riskincommodityfuturesmarketsflowedfromspeculatorsbacktoproducers.Insharpcontrast,no evidenceisfoundtosupportthistypeofriskflowduringtheCOVID-19crisis.Stressinthefinancial FundingInformation systemwasrelativelymildcomparedwiththe2008GFC,andthecommoditymarketexperienceda Theauthorsreceivednospecificfundingforthis strongrallyearlyinthecrisis.Bothspeculatorsandhedgerstradedinanorderlyfashion.Intermsof work. traders’relativeexposures,wefindthattheimpactfromfinancialstresswasimmaterial.Wealsofind JELClassification:G01,G13,Q02 thatspeculatorsgenerallyreactedtochangingfinancialconditionsmorestronglythanhedgers,during theperiod. KEYWORDS Commodities,Futures,FinancialStability,MarketVolatility,COVID-19,2008GFC 1 INTRODUCTION institutions,particularlythoseheavilyexposedtocommodityrelated assets. On the other hand, the financial system’s Westudyhowriskistransferredbetweendifferentparticipants health—includinginterestrates,marketliquidity,andinvestor withincommoditiesfuturesmarketsinresponsetoafinancial sentiment—could attenuate stress in the commodity futures systemshock.Understandingrisktransferinthesemarketsis market. important,becausecommodityfuturesplayacriticalroleinthe Hedgersandspeculatorsaretwomajortypeofparticipantsin globaleconomy.Commoditiesareessentialtotherealeconomy commodityfuturesmarkets.Typically,hedgerseitherproduce andfuturesprovideessentialhedginginstrumentsforproducers the commodity or use the actual commodity. Futures enable andconsumersalike.Thetradingofcommodityfuturesalso themtomanagetheirrisksbylocking-inprices,sothattheyare facilitatespricediscovery,canhelpmanagemarketvolatility, protectedagainstunfavorablepricefluctuations.Speculators, andoffersopportunitiesforspeculationandinvestment.Conse- on the other hand, are the risk-takers of the futures market. quently,thefuturesmarketrepresentsanimportantintersection They aim to profit from price movements, for example, by between the real and financial sides of the economy and un- buyingorsellingafuturescontractiftheyanticipaterisingor derstanding how risk is transferred through those markets is fallingpricesrespectively,orconstructingmorecomplicated valuable. exposures to take advantage of perceived opportunities. The Commodityfuturesmarketsarehighlyintegratedwiththe behaviorofboththesetypesofparticipantsandtheirinteraction globalfinancialsystem.Theinterconnectednessbetweencom- candeterminewhethercommodityfuturesserveasastabilizing modity futures markets and the financial system can cause ordestabilizingchannelformarketstress. significantrepercussionsduringtimesofmarketstress.Com- Thispaper,therefore,studiesrisktransferbetweenthesetwo moditypriceshockscancreatearippleeffectthroughfinancial groupsunderstress.Inparticular,welookataggregatechanges inthepositionsofcommodityfuturestradersduringtworecent 1

2 periodsofseverefinancialstress:the2008GlobalFinancialCri- whetheropeninterestasanindicatorcontainsanyinformation sis(GFC)andthe2020COVID-19crisis(includingthe2022 thatcanbeusedtoanticipatepricemovements;seeDediand RussianinvasionofUkraine).Weareparticularlyinterestedin Mandilaras (2022) for example. The focus of this paper is whetherandhowtradersreacteddifferentlytoglobalfinancial broaderaswelookattherelationshipbetweentraders’position systemshocksduringthosetwocrises,andwhatdifferences changesandthegeneralconditionofthefinancialsystem. in the reactions might imply about how commodity futures Ourresearchismainlymotivatedbyatheoryoftheconvecmarketscanamplifyormitigatefinancialstress.Thisobjective tive risk flow proposed by Cheng et al. (2015).† They show contrastswithpriorresearchthatfocusedonthepotentialim- empiricallythatduringthe2008GFC,riskflowedfromthefipactofcommoditymarketshocksonfinancial-sectorstability, nancialinstitutionstocommoditiesproducers.Beforethecrisis, specificallyonbanking-sectorstability(Alodayni2016,Kinda financialinstitutionsactivelyfacilitatedhedgingtradesforcometal.2018,EberhardtandPresbitero2021). modityproducersusingtheirbalancesheet.However,during Boththe2008GFCandthe2020COVID-19crisesseverely thecrisis,asthosefinancialinstitutionsencounteredsignificant stressedthefinancialsystemaswellasthebroadereconomy. stressontheirbalancesheet,theywereforcedtoreducetheir However, the causes and consequences of those two crises commoditymarketexposures.Withoutthosefinancialinstituarequitedifferent.The2008GFC,whichwasstartedbythe tionstakingthehedgingpositions,commodityhedgerseither mortgagemeltdown,representedaninflectionpointwithinthe hadtoclosetheirpositionsorfindothertradingcounterparties financialsystem caused by excessleverageand poor-quality toreplacethefinancialinstitutions.Thiscommoditychannel mortgageloans;the2020COVID-19crisis,ontheotherhand, dynamiccomplementsthecredittransferdynamicbetweenthe representedasubstantialglobaleconomicshockfromtheviral networkofglobalfinancialinstitutionsdetailedbyYangand outbreakofCOVID-19thatwassomewhatexogenoustothefi- Zhou (2013) during GFC. Interestingly, Cheng et al. (2015) nancialsystemitself.AtthebeginningoftheCOVID-19crisis, argues that during the 2008 GFC, some commodity producglobalfinancialmarketsexperiencedunprecedentedupheaval ersactuallytooktheroleoflargefinancialinstitutions,which involatility.Theseverityoftheinitialstressonfinancialstabil- wouldimplythecommoditymarketbecamedisconnectedfrom ityandtheglobaleconomyexceededeventhe2008GFC.With thefinancialnetworkatleasttosomeextent.Inparticular,coinsupportfromthegovernmentandthecentralbank,thefinan- cidentallywiththeincreaseincreditrisk,marketriskflowed cialmarketalsostabilizedandrecoveredmuchfastercompared fromfinancialinstitutionstocommodityproducers.Kangetal. withthe2008GFC.TheRussianinvasionofUkraineduring (2020) further developed this theory by showing that specu- 2022 caused a tremendous shock to some specific commod- lators need to pay a risk premium to producers for closing ity markets,butthe overallimpact on financial stability was theirpositionsbecauseofexternalfinancialconditions.Bonnier contained. (2021)alsofoundthatshort-termfluctuationsinopeninterest We use the Financial Stress Index (FSI) provided by the mightprimarilybedrivenbyspeculatorsdemandforliquidity. OfficeofFinancialResearch(OFR)tomeasurethestabilitycon- Thisstudymakesfouruniquecontributionstotheliterature. ditionsoftheglobalfinancialsystem.TheOFRFSIprovidesa First, our analysis extends the previous literature by considdailymarket-basedmeasurementofstresslevelinglobalfinan- ering a broader metric for financial stress by using the OFR cialmarkets.AccordingtotheOFR’swhitepaper(OFR2023), FSI. This index is not only a better indicator of broad finantheOFRFSIperformswellinidentifyingsystemicfinancial cialsystemconditions,butalsodecomposes"financialstress" stressasacoincidentalindicator.Inaddition,theOFRFSIleads into five specific categories: (1) credit, (2) equity valuation, theChicagoFedNationalActivityIndexinaGrangercausal- (3)funding,(4)safeassets,and(5)volatility.Thisapproach ityanalysis,suggestingthatthisindexmightbepredictiveof allowsustoassessthetypeoffinancialstressmarketparticidecreasesinglobaleconomicactivity. pantstowhicharereacting.Forinstance,iffundingconstraints Tomeasuretraders’positionchanges,weuseCommodity arethemaindriversforfinancialinstitutionsliquidatingtheir FuturesTradingCommission(CFTC)weeklyDisaggregated positions, we would expect there to be high correlation be- Commitments of Traders (DCOT) data. The CFTC collects tween the FSI funding sub-index and speculators’ positions. dataonthedailypositions(expressedthroughopeninterest)of Second,ouranalysisconsidershowtheCOVID-19financial largeparticipantsinindividualcommoditymarkets.TheCFTC shockaffectedcommodityfuturestraders’position.Thereare aggregates these data and groups individual trading entities manydifferencesaswellassimilaritiesbetweentheCOVIDintoseveralclassesoftraders,andreleasesthepositionsofthe 19shockandthe2008financialcrisis,allowingustotestthe aggregategroupsinitsweeklyDCOTreport.Someresearch convective risk flow hypothesis in different financial stress using these reports has focused mainly on finding whether certain groups of traders positions lead or lag returns, and howhedgingpressureaffectsprices.Thepurposeistodiscern †AnotherpaperthatisclosertothisoneisTokic(2012),whichstudiesthebehaviorof differenttradersduringthe2008oilbubble.

3 conditions.Third,weinvestigatebothabsoluteandrelativepo- Weidentifythefirstgroupasproducersorhedgersandaggresitionchangesforhedgersandspeculators.Previousresearch gatethesecondandthirdgroupsasspeculators.Weexclude onlystudiedthepositionchangesforindividualcommodities. theothercategoryfromouranalysis. Weextendtheanalysistorelativeexposurebetweenhedgers We use the OFR FSI to measure the level of stress in the andspeculatorsattheaggregatemarketlevel.Lastly,westudy financialmarkets.AccordingtoOFR(2023),theFSImeasures the positions of both active investors like money managers systemicfinancialstress–disruptionsinthenormalfunctioning andpassiveinvestorslikecommodityindextraders(CIT)in offinancialmarkets.TheOFRFSIincorporatesfivecategories commodityfuturesmarket.Wetestwhetherthosetwotypeof ofindicatorsandisconstructedfrom33financialmarketvariinvestors,bothcategorizedasspeculators,respondtofinancial ables,suchasyieldspreads,valuationmeasures,andinterest systemstressdifferentlyduringthetwocrises. rates.Eachvariableintheindexmeasuresafeatureoffinancial Theremainderofthepaperisorganizedasfollows.Inthe stress.Financialstresscanbecapturedbyhowthevariables followingsection,weprovideabriefintroductiontotheDCOT movetogetherthroughtime.Astatisticalalgorithmcaptures data and the FSI. Section 3 provides an overview of market theco-movementandproducesasetofweightsforthevaridynamicsandtraderpositionsduringthe2008GFCandthe ables.ThevalueoftheOFRFSIonagivendayistheweighted COVID-19crisis.Section4studiestherelationshipbetween averagelevelofeachvariableobservedinthemarketonthat financialstabilityandcommoditymarket.Section5presentsthe day,relativetoitshistory.TheOFRFSIispositivewhenstress regressionresultsfortheabsolutechangesintraders’positions. levelsareaboveaverage,andnegativewhenstresslevelsare Section6providesastudyontherelativeexposureofdifferent belowaverage.Themagnitudeoftheindexindicateshowfar traders,andsection7concludes. theindexdeviatesfromtheaverageatzero.TheindexiscalculatedaftereachU.S.tradingdayandisaconcurrentindicator. TheOFRFSIincorporatesfivecategoriesofindicators: Credit Trackscreditspreads,whichindicatethecostdifference forborrowingbetweenfirmswithvaryingcreditworthi- 2 DCOTDATAANDTHEFSI ness.Widerspreadssuggestinvestorsaremorehesitantto holddebt,raisingborrowingcosts; This study focuses on four groups of commodity futures ac- Equityvaluation Monitors stock valuations across multiple tively traded on the Chicago Mercantile Exchange (CME): market indexes, reflecting investor sentiment and risk (1) energy, (2) agriculture, (3) base metals, and (4) precious tolerance; metals. For traders’ positions, we use the DCOT report pub- Funding Assesses the ease with which financial institutions lishedbytheCFTC.Thereportisreleasedasasnapshotofthe cansecurefunding.Stressinfinancialmarketscanleadto aggregatedpositionsoffourcategoriesoftradersonTuesday fundingfreezesifmarketparticipantsperceiveheightened eachweek.Thecategoriesare:(a)Producer/Merchant/Proces- counterpartycreditorliquidityrisks; sor/User,(b)SwapDealers,(c)MoneyManagers,and(d)Other Safeassets Evaluatesthevaluationofassetsconsideredstable Reportables.TheCFTCdefinesthesecategoriesasfollows: storesofvalueorthosewithpredictablecashflows.During periods of stress, higher valuations of safe assets may Producer/Merchant/Processor/User Entity that trades pre- indicatea"flighttoquality"asinvestorsshiftfromriskier dominatelyforthepurposeofhedgingriskthatistiedto orlessliquidassetstosaferoptions;and theirnon-financialbusiness; Volatility Measuresbothimpliedandrealizedvolatilityacross SwapDealer Entitythatprimarilydealsinswapsforacom- equity, credit, currency, and commodity markets. Inmodityandusesthefuturesmarketstomanageorhedge creaseduncertaintyaboutassetvaluesorinvestorbehavior theriskassociatedwiththoseswaptransactions.Theswap duringstressedperiodsoftenmanifestsashighervolatility. dealers counterparties may be speculative traders, like hedge funds, or traditional commercial clients that are OurDCOTdatasampleextendsfromJune2006throughDemanagingriskarisingfromtheirdealingsinthephysical cember2022,coveringthreemajorcommoditymarketshocks: commodity; the2008GFC,the2020COVID-19crisis,andthe2022Russian MoneyManager Entity that trades speculatively, such as a invasionofUkraine.Table1presentsthestatisticsoftraders’ commoditytradingadvisor(CTA),aregisteredcommodity netpositionsovertheentiresampleperiodforselectedcompooloperator(CPO),oranunregisteredfundidentifiedby modityfuturescontracts.Generally,producershavenetshort theCFTC;and positionsastheyaresellingfuturescontractstohedgeagainst Other EntitiesthattheCFTCcannotreliablyplaceintoanyof thepricerisk.Swapdealersandmoneymanagersareusually theothercategories.

4 TABLE 1 SummaryStatisticsforNetPositionsofTraders’OpenInterest Producers SwapDealers MoneyManagers Commodities Mean StdDev Mean StdDev Mean StdDev CrudeOil –100,860 96,304 –184,264 235,129 185,870 105,634 Wheat –85,531 47,528 109,924 37,931 –15,692 46,838 NaturalGas –32,518 36,424 111,588 72,190 –26,872 108,789 Copper –40,177 26,234 37,474 26,234 10,944 32,684 Gold –105,193 55,524 –76,222 61,237 103,905 70,480 Soybean –161,845 103,365 100,421 24,782 69,596 18,780 Unitsarenumberoffuturescontracts 30 20 10 0 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 2020 2022 2024 eulaV Index FSI Credit Safe_assets Funding Volatility Equity FIGURE 1 HistoryoftheOFRFSIanditsSub-indexes ontheothersideofproducers’hedgepositions,andtherefore on the global financial system, but the impact was muted in tendtohavenetlongpositions. termsofmagnitude,althoughtheindexdidremainpositivefor Figure 1 shows the daily history of the OFR FSI and its severalmonths. sub-indexfrom2000through2024.TheFSIpeakedin2008 andremainedinpositiveterritoryforanextendedperiod.The 2008GFCwasabankingcrisis,andtheglobalfinancialsystem 3 COMMODITY MARKET AND wasunderpersistentseverestressforseveralyears.TheFSI TRADERPOSITIONOVERVIEW didnotreturntoitsneutralbaselineuntilFebruary2010.The COVID-19crisis,ontheotherhand,wastriggeredbyeconomic Commodityfuturesmarketshaddistinctpricemovementsbelockdownsaimedatpreventingthevirus’sspread.Itwasnota tweenthe2008GFCandtheCOVID-19crisis.Inthe2000s, bankingcrisis,andtheimpactontheglobalfinancialsystem commoditymarketsexperiencedapriceboom,drivenbythe wastemporaryandrelativelycontained,particularlycompared growth of the global economy and the inflow of investors’ to the GFC. The FSI jumped in 2020, but too less than 50 money.Commoditypricesrosesteadilythroughthe2000sunpercentofitspeaklevelduring2009.Thislevelwassimilar tilthebeginningofthe2008GFC,whenthemarketcollapsed. totheinitial(local)peakintheFSIobservedin2008thatwas Whatcausedthisboomandbustcycleincommoditymarkets subsequentlydwarfedbythelater(global)peak.Followingthe hasbeenextensivelystudied(DominguezandReinhart2008, initial jump in 2020, the FSI quickly fell back into negative Carter et al. 2011, Irwin and Sanders 2011, Hamilton 2009, territory relatively quickly, indicating that financial stability etc.). Researchers found strong interconnectedness between conditionsrevertedtonormalafterashort-livedmarketturmoil. financialsystem,speculativetrading,andcommoditymarket In2022,theRussianinvasionofUkrainecreatedsomestress movements.Somehavearguedthatexcessivespeculationwas

5 400 300 200 100 0 2006 2008 2010 2012 2014 2016 2018 2020 2022 eulaV Gold Wheat Copper Crude Oil Natural Gas FIGURE 2 DailyHistoryofNormalizedandInflation-AdjustedCommodityFuturesPrices 300 200 100 2006 2008 2010 2012 2014 2016 2018 2020 2022 eulaV Gold Wheat Copper Crude Oil Natural Gas FIGURE 3 HistoryofNormalizedOpenInterest amajorcauseofthedramaticpricemovementincommodity thelossfrom2009through2012.Sincethen,thecommodity prices.Incontrast,thecommoditymarketswererelativestable marketswereinamulti-yeardownwardmovement.InthebebeforetheCOVID-19crisisandgainedmomentumfortheup- ginningoftheCOVID-19crisis,thecommoditymarketswere wardmovementaftertheCOVID-19,partiallydrivenbythe weakened by the economic shutdowns and lack of demand. imbalanceofsupplyandthestrongdemandfromtheeconomy Then,theeconomicrecoveryandsupplychainshortagescaused recovering(MongeandLazcano2022,ZhangandWang2022). an imbalance between supply and demand for commodities, Figure2showsthepricemovementforfivecommodityfu- resultingincommoditypricessharplyincreasingsince2020. tures.ThepriceisnormalizedtothelevelofAug25th,2005, Traders’positions,asmeasuredbythetotalopeninterestof anditisalsoadjustedforinflationusingtheconsumerprice futurescontracts,haveshownsimilarmovementwiththeprice inflationindex.Exceptforgoldfutures,theinflationadjusted changes.Figure3presentsthenormalizedtotalopeninterest priceforcommodityfuturesdroppeddramaticallyafterreach- forselectedcommodities.Followingthepricemovement,total ing the historical high in early 2008. It recovered most of openinterestforcommodityfuturessteadilyincreasedinthe

6 600,000 400,000 200,000 0 2007 2009 2011 2013 2015 2017 2019 2021 2023 2025 stcartnoC Producer_Long Producer_Short Speculator_Long Speculator_Short Total_OI FIGURE 4 DailyHistoryofOpenInterestforWheatFutures 2000sandreachedapeaklevelinearly2008.Whenthemar- marketselloff.DuringtheCOVID-19crisis,theaggregateposiket collapsed, traders dramatically reduced their commodity tionsincreasedwhenthecommodityindexreboundedsharply futuresexposures.Totalopeninterestquicklydroppedbelow fromthebottomin2020andgraduallydecreasedafterreach- 2005 levels.When marketsstarted to recoverafter the GFC, ingthepeakin2022.Thisresultisconsistentwiththefindings openinterestincreasedaswell.Itremainedrelativelystablebe- of Bonnier (2021) and Kang et al. (2020), who showed that foretheCOVID-19crisis.AftertheCOVID-19,openinterest thecommoditypricemovementisthemaindriveroftraders’ increasedwiththefuturesprices. positionschangeforbothspeculatorsandhedgers. Figure 4 shows the long and short positions of producers (hedgers)andspeculators(swapdealersandmoneymanagers) for wheat futures. Hedgers’ and speculators’ positions have 4 FINANCIAL STABILITY AND COMdisplayed similar historical patterns as the commodity price MODITYMARKETS movements.SincethebeginningoftheGFC,producershave beenreducingtheirshorthedgingpositions,whilespeculators Tostudytherelationshipbetweenfinancialstabilityandtraders’ havebeenactivelyclosingtheirlongexposures.Thechange positions,wefirstneedtounderstandthefactorsdrivingthe intotalopeninterestwasmainlydrivenbythecross-trading tradingbehaviorsofhedgersandspeculators.Lehecka(2015) betweenproducersandspeculators.Whenthemarketrecovered provided a systematic empirical investigation of lead-lag refromthesell-off,speculatorsstartedtobuildtheirlongexposure lationshipsamongtradingpositionsandpricesincommodity andhedgersincreasedtheirshortpositions.Othercommodity futures markets. They employed Granger-causality tests apfuturesdisplayedasimilartradingpatternduringthe2008GFC plied to a variety of measurements of trading activities and period.Incontrast,duringtheCOVID-19crisis,hedgers’and futuresprices.Theirresultsindicatedlittleevidencetosupport speculators’ positions remained relatively stable comparing thattradingactivitiesleadcommoditypricesmovement.Inconwiththe2008GFCperiod. trast,therearestrongevidencesthatpricestendtoleadtraders Figure5showstheaggregatelongandshortpositionsfor hedging and speculation activity. These results appear to be speculators and hedgers cross all major commodity futures, generallypersistentovercommodities,measurementsofhedgalongwiththeS&PGSCI(formerlytheGoldmanSachsCom- ingandspeculation,andperiods.BoschandSmimou(2022) modity Index‡). All the data are normalized to August 25th, showed that pricing in commodity markets can be predomi- 2005.Itshowsasizabledroponspeculators’longpositionsand nantlyattributedtohedgersandinfluentialspeculators(money hedgers’shortpositionsduringthe2008GFC.TheS&PGSCI managers).Ekelandetal.(2019)studiedhedgingpressureand declinedsharplyasallthecommoditiesexperiencedaserious speculationincommoditymarkets.Theirstudyalsoattributes priceactiontotraders’positionchanges.BoosandGrob(2023) showedthattrendsignalslargelyexplainpositionchangesof ‡S&PGSCIdatawasaccessedviaBloomberg speculatorsincommoditymarkets.Priceactionisoneofthe

7 160 120 80 2007 2009 2011 2013 2015 2017 2019 2021 2023 eulaV Producer_Short Speculator_Long S&P GSCI FIGURE 5 NormalizedAggregateTraders’PositionsandtheS&PGSCI main factors causing traders to change their positions. This variablesarethelaggedweeklyreturnoftheS&PGSCI,the resultisconsistentwithwhatweshowedintheprevioussection. volatilityoftheS&PGSCIweeklyreturn,theweeklychange Theothertwomainfactorsdrivingtraderpositionchanges intheFSI,andthelaggedweeklychangeintheFSI. aremarketvolatilityandinitialmarginlevel.Marketvolatility Table2showstheregressionresults.Thecoefficientsforthe willimpactbothvariationmarginandinitialmargin,thecostof FSIweeklychangearenegativeandstatisticallysignificantfor openingandmaintainingafuturesposition.Hartzmark(1986) all four sample periods. This finding implies that increasing studiedtherelationshipbetweenmarginsandthedegreeofex- financialstresscanaddpressuretocommodityfuturesprices. cessivespeculatorparticipationinfuturesmarkets.Theyfound Aone-pointincreaseintheFSIresultsina1.3percentdropin that margin changes will result in significant changes in the theS&PGSCIduringthe2008GFCperiodand2.7percent compositionoftradersinthemarket.DaskalakiandSkiadopou- fortheCOVID-19period.Thecommodityfuturesmarketwas los(2016)foundthatmarginincreasesmakehedgersexitfrom more reactive to the change of financial stability conditions grainandmetalmarkets,andtheeffectofmarginchangesis duringtheCOVID-19crisisthantheGFC,thiscouldbedue morepronouncedincommodityfuturesmarketsthaninequity tothefactthatthechangeofFSIduringtheCOVID-19crisis andinterestratefuturesmarkets. isrelativesmall.ThelaggedchangeoftheFSIdoesnotshow anystatisticallysignificantimpactontheS&PGSCI,meaning thatthecommodityfuturesmarketmainlyrespondstothemost 4.1 Financial Stability and Commodity recentchangesinfinancialstabilityconditions. Prices Figure 6 shows the scatter plot of the weekly changes in the S&P GSCI and the FSI. The left-hand panel is for the Givenpricemovementisthekeydriverfortraders’position 2008GFC,andtheright-handpanelisfortheCOVID-19pechanges,werunaregressionanalysisbetweentheFSIandthe riod.Theupperpanelplotstheweeklychangeofbothindices, S&PGSCItostudytherelationshipbetweenfinancialstability andthebottompanelplotstheweeklychangeintheS&PGSCI andcommodityprices.TheS&PGSCIcurrentlycomprises24 with the level of FSI. During the 2008 GFC, the weekly recommodities from all commodity sectors - energy products, turn of S&P GSCI was mostly negative, and the FSI stayed industrialmetals,agriculturalproducts,livestockproducts,and inpositiveterritory.Commodityfuturespricesdeclinedwhen preciousmetals.Itservesasabenchmarkforinvestmentinthe thefinancialsystemwasunderstress.Incontrast,duringthe commoditymarkets.Weruntheregressionusingfourdifferent COVID-19period,theweeklyreturnofS&PGSCIwasmostly historical data samples: the 2008 GFC (from 2007 through positive and the FSI stayed in negative territory most of the 2009),theCOVID-19crisis(from2020through2022),thefull time.Commodityfuturespricessurgedwhenfinancialstability sample(from2007through2022),andthedatasamplebetween conditionsnormalizedaftertheCOVID-19lockdowns. twocrises(from2009through2020).Thedependentvariable is the weekly return of the S&P GSCI, and the independent

8 TABLE 2 GSCIWeeklyReturnRegressionResultswithFSI Sample GFC COVID-19 All Between (1) (2) (3) (4) –0.091 –0.086 –0.038 0.001 GSCILagWeeklyReturn (0.084) (0.083) (0.035) (0.046) –0.001 0.0003 –0.0003 0.001 LagFSI (0.003) (0.004) (0.001) (0.002) –0.519*** –0.136 –0.153** 0.036 GSCIVol (0.198) (0.165) (0.074) (0.130) –0.013*** –0.027*** –0.017*** –0.020*** FSI (0.002) (0.004) (0.001) (0.002) Observations 151 151 830 480 R2 0.185 0.267 0.179 0.206 AdjustedR2 0.163 0.247 0.175 0.199 Note:***p<0.01,**p<0.05,*p<0.1;standarderrorsin() 0.15 0.10 0.05 0.00 −0.05 −0.10 −6 −3 0 3 FSI Change nruteR ICSG 2008 GFC 0.1 0.0 −0.1 −2 0 2 4 FSI Change nruteR ICSG COVID−19 0.15 0.10 0.05 0.00 −0.05 −0.10 0 10 20 FSI nruteR ICSG 0.1 0.0 −0.1 0 5 FSI nruteR ICSG FIGURE 6 RelationshipbetweentheS&PGSCIandtheFSI 4.2 Financial Stability and Volatility/Mar- producers’shortpositions,whileFigure8showsthesamechart forspeculators’shortandproducers’longpositions.Theleftgin handpanelpresentsforthe2008GFC,andtheright-handpanel isfortheCOVID-19period.Duringthe2008GFC(leftpanels), Increasingmarketvolatilityraisesthevariationmarginfora bothspeculators’longpositionsandproducers’shortpositions futuresportfolio.Clearinghouseswillalsoincreasetheinitial exhibitastrongnegativecorrelationwithinitialmarginlevels, margin requirement to address high market volatility. Both especiallywheninitialmarginsincreasedsignificantly;when changesraisethecoststoopenandmaintainfuturespositions initial margin was up over 50%, traders’ positions have deforhedgersandspeculators,andthehightradingcostscould creasedsubstantially,indicatingasignificantimpactonmarket forcetraderstoliquidatetheirpositions,triggeringafire-sell participation.Theelevatedinitialmarginmateriallyincreased event,especiallywhenthefinancialsystemisunderstress. tradingcostsduringtheGFC,whichcouldexplainwhytraders Weusewheatfuturesasanexampletoillustratetheimpact weredramaticallyreducingtheirfuturespositions.Incontrast, of initial margin on traders’ positions. Figure 7 presents the theCOVID-19period(rightpanels)showsamuchweakerrescatterplotbetweeninitialmarginandspeculators’longand lationship,withonlyaslightnegativetrendvisible.Notably,

9 275000 250000 225000 200000 175000 150000 100 150 200 250 gnoL rotalucepS 2008 GFC 210000 180000 150000 100 120 140 160 gnoL rotalucepS COVID−19 175000 150000 125000 100000 75000 100 150 200 250 IM trohS recudorP 160000 140000 120000 100000 100 120 140 160 IM trohS recudorP FIGURE 7 NormalizedInitialMarginversusTraders’Positions-PanelA 80000 60000 40000 100 150 200 250 trohs rotalucepS 2008 GFC 110000 90000 70000 50000 100 120 140 160 trohS rotalucepS COVID−19 40000 30000 20000 100 150 200 250 IM gnoL recudorP 60000 50000 40000 30000 20000 100 120 140 160 IM gnoL recudorP FIGURE 8 NormalizedInitialMarginversusTraders’Positions-PanelB the range of normalized initial margin values is smaller for chartrevealsdistinctmarketbehaviorsduringthetwomajor the COVID-19 period compared to the GFC, meaning there crises:2008GFCandCOVID-19pandemic.DuringtheGFC, was also a very limited change on the initial margin during commoditymarketvolatilityspikeddramatically,accompanied theperiod.Forspeculators’shortpositionsandhedgers’long byaneartriplingofnormalizedinitialmarginsforbothwheat positions, the impact from the change of initial margin was andcornfutures.Thiselevationinmarginspersistedthroughrelativelymutedduringbothcrises. outthecrisisperiod,indicatingprolongedmarketuncertainty The level of initial margin generally follows the market andrisk.Incontrast,theCOVID-19crisisin2020exhibiteda volatility. Figure 9 shows the annualized volatility for the differentpattern.WhiletheS&PGSCIvolatilitysurgedbriefly, S&PGSCIandnormalizedinitialmarginlevelforwheatand itquicklysubsided.Interestingly,theinitialmarginsforwheat cornfuturesfrom2005to2024.Theinitialmarginlevelismea- andcornfutureshavenotshowedsignificantincreasesduring suredastheratioofinitialmargintothevalueforonefutures thisperiod,suggestingthatagriculturalfutureswererelatively contract, normalized to the level of August 25th, 2005. The insulatedfromthebroadercommoditymarketvolatility.The

10 300 200 100 0 2006 2008 2010 2012 2014 2016 2018 2020 2022 eulaV GSCI_Vol Wheat_IM Corn_IM FIGURE 9 S&PGSCIVolatilityandNormalizedInitialMarginforWheatandCorn 1.0 0.5 0.0 −0.5 −1.0 2008 2010 2012 2014 2016 2018 2020 2022 noitalerroC Correlation GSCI−FSI GSCI−IM IM−FSI FIGURE 10 MovingWindowCorrelationamongS&PGSCIVolatility,theFSI,andInitialMarginofWheatFutures S&PGSCIindexhasover20differentcommoditiesinthebas- Tofurtherillustratethedivergencebetweeninitialmargin ket;itsvolatilitylevelisameasureforthebroadcommodity andgeneralmarketvolatilityduringCOVID-19crisis,wecalmarketcondition,notforaspecificcommodityfutures.During culatetheone-yearmovingwindowcorrelationamongtheFSI, theGFC,Commoditymarketsweremoreinter-correlateddur- theS&PGSCI,andinitialmarginforwheatfutures.Figure10 ingthe2008GFC.Therewasamarketsell-offacrossallthe presentsthehistoryfrom2008to2022.Notably,thosethree commodityfutures.Incontrast,duringtheCOVID-19crisis, variableswerehighlycorrelatedduring2008GFC.Initialmarcommodity futures markets were less correlated and volatil- ginforaspecificcommodityfuturesisusuallydeterminedbyits ity level varied across different commodities. This explains ownvolatility,notbythebroadmarketvolatility.Giventhecomwhy the S&P GSCI volatility can diverge from the margin moditymarketswerehighlyconnectedduringthe2008GFC, requirementsofindividualagriculturalfutures. correlation between initial margin for a specific commodity futuresandthebroadmarketvolatilitywashigh.Thisresult

11 TABLE 3 VARResults–FSIandS&PGSCIVolatility Sample 2008GFC COVID-19 FullSample Dependentvariable⇒ GSCIVol FSI GSCIVol FSI GSCIVol FSI 1.020*** 8.671 1.094*** –3.457 1.048*** –0.081 GSCIVolLag1 (0.082) (24.229) (0.093) (13.644) (0.036) (7.009) 0.001** 0.081 0.002*** 0.248*** 0.001*** 0.076** FSILag1 (0.0003) (0.081) (0.001) (0.094) (0.0002) (0.036) –0.047 –19.423 –0.165* –5.772 –0.100*** –6.492 GSCIVolLag2 (0.083) (24.461) (0.093) (13.669) (0.036) (7.033) 0.001* 0.216*** 0.001 0.123 0.0005*** 0.154*** FSILag2 (0.0003) (0.082) (0.001) (0.089) (0.0002) (0.036) 0.001 0.417 0.002*** 0.327** 0.001*** 0.185*** const. (0.001) (0.257) (0.001) (0.130) (0.0003) (0.061) Observations 149 149 149 149 799 799 R2 0.933 0.092 0.921 0.145 0.915 0.047 AdjustedR2 0.931 0.067 0.919 0.121 0.915 0.042 GCF-Test 4.4637 8.8882 22.463 GCp-value 0.01233 0.0002 0.0000 Note:***p<0.01,**p<0.05,*p<0.1;standarderrorsin() is consistent with Abricha et al. (2024), who found that in- onS&PGSCIvolatility.Granger-causality(GC)testswhether terconnectedness increases at all quantiles during periods of thepredictionofonetimeseriescanbeimprovedbyincorpohighmarketstress.Elevatedinitialmargincrossallcommodity ratingtheknowledgeofasecondtimeseries.Ifitdoes,then futuresduringtheGFCcausedbypersistentfinancialsystem thelatterissaidtohaveacausalinfluenceonthefirst. stress materially increased transaction costs for traders and We run the VAR models and Granger-causality tests uscould force them to close their futures position. In contrast, ingthreedifferentsampleperiods:the2008GFC(from2007 duringtheCOVID-19crisis,correlationamongdifferentcom- through2009),theCOVID-19(from2020through2022),and moditieswasrelativelylowandinitialmarginforindividual the full sample period (from 2007 through 2022). Table 3 commodityfuturesdidnotcloselyfollowwiththebroadmarket presentstheVARmodelestimationsaswellastheGCtestrevolatility.Financialsystemstressandgeneralmarketvolatility sults.BoththeVARandGCtestacceptthehypothesisthatthe did not cause a spike in the initial margin for wheat futures. changesoftheFSIcanimprovethepredictionofS&PGSCI As a result, traders can maintain their positions without the volatility,implyingthatsystemicfinancialstresscanincrease economicpressurefromtheincreasingoftransactioncost. thevolatilityincommodityfuturesmarkets.Thisresultholds Tobetterunderstandthestatisticalrelationshipbetweenthe forallthreesampleperiods.Interestingly,wedon’tfindthat FSIandS&PGSCIvolatility,weappliedavectorautoregres- theS&PGSCIvolatilityshowsanymeaningfulimpactonthe sive(VAR)modelontheweeklychangeofthosetwovariables. FSIasallthecoefficientsarestatisticallyinsignificant. VARmodelsdescribethejointgenerationprocessovertime and can be used for investigating relationships between two timeseriesvariables.Ourmodelsarespecifiedas: 4.3 Financial Stability and Traders’ Positions ∆GSCIVolatility =α+∆GSCIVolatility t t–1 +∆FSI +∆GSCIVolatility +∆FSI +(cid:15) (1) AfterabriefstudyingontheconnectionbetweenFSIandthe t–1 t–2 t–2 1t factorsdrivingthechangesoftradingpositioninfuturesmarket, and weexaminethedirectrelationshipbetweenFSIandtraderpositions.Theliteratureonhowfinancialstresscanaffecthedgers ∆FSI =α+∆GSCIVolatility +∆FSI andspeculatorsiswell-developed.ChenandYang(2021)found t t–1 t–1 +∆GSCIVolatility +∆FSI +(cid:15) (2) thatduringaperiodwhenmarketvolatilitywasespeciallyelet–2 t–2 2t vated,dealersandleveragedfundmanagerswouldaltertheir WealsoperformstandardGranger-causalitytests(Granger tradingstrategies.Röthig(2011)providedsomeempiricalre- 1969)tostudythetemporalinfluenceoffinancial(in)stability sultsthatspeculatorsleadhedgersincurrencyfuturesmarkets

12 240000 200000 160000 120000 0 10 20 gnoL rotalucepS 2008 GFC 210000 180000 150000 0 5 gnoL rotalucepS COVID−19 150000 100000 50000 0 10 20 FSI trohS recudorP 160000 140000 120000 100000 0 5 FSI trohS recudorP FIGURE 11 TheFSIandTraders’Position andattracthedgerstoopen/closepositionsincurrencyfutures fromthemacroeconomicconditions.§Theregressionmodelis markets.Chengetal.(2015)arguedthatduringthe2008GFC, specifiedasthefollowing: somecommodityproducersplayedtheroleoflargefinancial institutionsbytakingtheotherendoftradesofdistressedbanks ∆Position t = α+∆FSI t +∆GSCI t (3) and money managers. Kang et al. (2020) tested for the exis- +∆GSCIVolatility +∆Margin t t tence of a liquidity premium paid by speculators to hedgers. +Controls +(cid:15) t t Allthesestudiessupporttheideathatfinancialstabilityconditionscanchangetraders’positionsinthecommodityfutures Table4showstheregressionresults,coveringthreedifferent market,especiallyforthespeculatorswhoaremorereactiveto sampleperiods:the2008GFC(from2007through2009),the thefinancialsystemstress. COVID-19crisis(from2020through2022),andthefullsample Figure11presentsthescatterplotbetweentraders’positions period(from2007through2022).Wefocusedonhedgers’short andtheFSIinwheatfuturesmarket.Theleft-handpanelisfor positionsandspeculators’longpositions.Ingeneral,thecoefthe2008GFCandtheright-handpanelisfortheCOVID-19 ficientsoftheFSIarenotstatisticallysignificantforhedgers’ crisis.Speculators’longpositionsandhedgers’shortpositions shortpositions,implyingthatfinancialstresshaslessimpact displayedastrongnegativecorrelationwiththeFSIduringthe onhedgers.Forspeculators’longpositions,thecoefficientsare 2008 GFC. The higher the FSI level, the lower the open in- statisticallysignificant.Increasesinfinancialstresscanreduce terests. During the 2008 GFC, financial stress put pressures speculators’ long positions. To address the multicollinearity oncommoditypricesandincreasedmarketvolatility.Asare- problem in the regression model, we run a step-wise regressult, speculators aggressively closed the long positions and sionanalysistoselectthemostsignificantvariablesdrivingthe producersalsoquicklyreducedtheirshortpositions.Duringthe changeoftraders’positions.Forhedgers’shortpositions,the COVID-19crisis,thecorrelationbetweentheFSIandtraders’ regressionselectspricereturnandinitialmarginlevel;Forspecpositions was not as significant as it was during the 2008 ulators’longpositions,themodeladdstheFSI.Thisfinding GFC;financialstresshadrelativelylimitedimpactontraders’ confirmsthatspeculatorsreacttothefinancialsystemstability positions. conditionsmoresubstantiallythanhedgers. Using wheat futures as an example, we conduct a regres- WealsoruntheregressionanalysisusingtheFSIsub-index. sionstudytoshowhowtraders’positionchangesrespondto Table5providestheregressioncoefficientsforeachsub-index. the change of financial stability conditions. We run separate Again,wefindthathedgersareingeneralnotveryresponsive regressionsforhedgersandspeculators.WeusetheBalticDry tothechangesinfinancialstabilitycondition.Thecoefficients Index(BDRY)astheindependentvariabletocontroltheimpact forallFISsub-indicesarestatisticallyinsignificantforhedgers’ §TheBDRYisreporteddailybytheBalticExchangeinLondontomeasurethecostof movingthemajorrawmaterialsglobally;datawasaccessedviaBloomberg.

13 TABLE 4 Traders’PositionChangeRegressionwiththeFSI Sample 2008GFC COVID-19 FullSample Positions⇒ Hedge Speculate Hedge Speculate Hedge Speculate 0.425*** 0.134** 0.153 0.074 0.306*** 0.135*** GSCIReturn (0.137) (0.063) (0.199) (0.132) (0.089) (0.045) 0.021 0.045* –0.016 –0.023 –0.023 –0.016 BDRY (0.051) (0.023) (0.064) (0.043) (0.029) (0.014) 0.667* 0.149 –0.084 –0.253 0.279 –0.075 GSCIVol (0.385) (0.176) (0.382) (0.253) (0.212) (0.106) –0.004 –0.003* –0.011 –0.023*** –0.004 –0.006*** FSI (0.004) (0.002) (0.010) (0.006) (0.004) (0.002) –0.0002** –0.0001* –0.0005 –0.0001 –0.0003*** –0.0001** Margin (0.0001) (0.00005) (0.0003) (0.0002) (0.0001) (0.00004) Observations 111 111 111 111 831 831 R2 0.199 0.225 0.056 0.186 0.041 0.053 AdjustedR2 0.161 0.188 0.011 0.147 0.035 0.047 Note:***p<0.01,**p<0.05,*p<0.1;standarderrorsin() TABLE 5 Traders’PositionChangeRegressionwithFSISubindex Sample 2008GFC COVID-19 FullSample Positions⇒ Hedge Speculate Hedge Speculate Hedge Speculate –0.023 –0.021*** –0.022 –0.111*** –0.020 –0.039*** Credit (0.017) (0.008) (0.042) (0.027) (0.017) (0.009) –0.003 –0.006 0.017 –0.150*** –0.007 –0.012** Funding (0.009) (0.004) (0.059) (0.039) (0.011) (0.006) –0.047 –0.018 0.120 0.137 –0.009 0.023 SafeAssets (0.049) (0.023) (0.140) (0.098) (0.040) (0.020) –0.014 –0.008* –0.025 –0.031*** –0.005 –0.009*** Volatility (0.009) (0.004) (0.017) (0.012) (0.007) (0.004) Note:***p<0.01,**p<0.05,*p<0.1;standarderrorsin() shortpositions.Speculators,ontheotherhand,aremuchmore 5 ABSOLUTEPOSITIONCHANGE reactive to the increase of financial stress, especially to the conditionsofcredit,funding,andmarketvolatility.Thisfinding Thissectionextendsourregressionanalysistomoreindividual impliesthatspeculatorsareparticularlysensitivetowidening commodities.FollowingthemethodofChengetal.(2015)and creditspreads,deterioratingfundingmarkets,andincreasing Kangetal.(2020),weemployastandardOLSregressiontoexmarketvolatility.Theseresultsalignwellwiththeliterature. aminetheimpactoffinancialstabilityontraders’exposuresto Chengetal.(2015)showedthatfinancialinstitutionsdecrease differentcommodityfutures.Theregressionmodelisspecified theirmarketexposureincommoditieswithrespecttochanges asthefollowing: inequitymarketvolatility.Speculatorssensitivitytocreditrisks alsoisconsistentwiththesensitivityoffinancialinstitutions ∆Position = α+∆FSI +∆Price (4) t t t–1 tocreditduringtheGFC(YangandZhou2013);ouranalysis +∆FSI +∆Position showsthatsuchsensitivityextendedtotheCOVID-19period. t–1 t–1 +Controls +(cid:15) t t We study four different position changes: the change in traders’netposition(thelongpositionoffsetbytheshortposition);thechangeinnetoftrade,whichisthechangeinthenet

14 TABLE 6 RegressionCoefficientoftheFSIforTraders’NetPositionChange–GFC FSICoefficients Producers SwapDealers MoneyManagers Commodities Pre-GFC GFC Pre-GFC GFC Pre-GFC GFC –1119.99 674.92 30.28 1645.59 –1850.29 –3415.41** CrudeOil (1612.51) (832.81) (1604.51) (1299.03) (1756.33) (1621.53) 133.38 1784.49** –92.90 –826.11** 420.77 –772.12 Wheat (834.25) (690.26) (512.21) (346.64) (887.40) (610.46) –443.29 766.58** –833.39 611.88 179.75 –2339.93 NaturalGas (656.18) (373.30) (692.33) (657.56) (1302.59) (1420.90) 159.44 586.41** –125.88 10.03 –43.42 –363.34 Copper (221.54) (248.69) (132.04) (192.88) (247.34) (286.33) –231.77 830.15 –2245.11** 340.93 1259.19 –1035.69 Gold (1379.31) (860.39) (1028.18)‘ (634.59) (1776.64) (904.10) –4535.65 4831.55 1500.50 –1000.64 2803.39 –4071 Soybean (4317.77) (3550.48) (1363.03) (902.68) (3321.56) (3084.28) Note:***p<0.01,**p<0.05,*p<0.1;standarderrorsin() positiondividedbytotalopeninterest;thechangeinhedgers’ Table6showstheregressioncoefficientoftheFSIforthe shortpositions;andthechangeinspeculators’longpositions. pre-GFCperiodandtheGFCperiod.Thedependentvariableis Tobeprecise,thepositionchangesarecalculatedasfollows: theweeklychangeinnetposition.Theregressionresultsshow thatproducersgrewtheirnetpositionswiththeincreaseofthe ∆NetPosition t =NetPosition t –NetPosition t–1 , FSIovertheGFCsampleperiod.Theyeitherclosedtheirshort ∆NetofTrade = (NetPosition t –NetPosition t–1 ) , positionsoropenedmorelongpositions.Thischangeadded t TotalOpenInterest moreexposurestocommoditypricemovementsasproducers t–1 ∆Hedger’sShortPosition = are less hedged. The coefficients for wheat, natural gas, and t (Hedger’sShortPosition –Hedger’sShortPosition ) copperarestatisticallysignificantintheGFCsample.Forthe t t–1 , Hedger’sShortPosition pre-GFCsample,thecoefficientsarenotstatisticallysignificant t–1 forallthecommodities.Forswapdealers,thecoefficientsof and theFSIaremostlynotstatisticallysignificant,exceptforGold inthepre-GFCperiodandWheatduringtheGFC.Thereisnot ∆Speculator’sLongPosition = aclearpatterntoshowtheimpactsfromfinancialstabilitycont (Spec.’sLongPosition –Spec.’sLongPosition ) ditions.Formoneymanagers,thecoefficientsarenegativefor t t–1 . Speculator’sLongPosition theGFCsampleperiod,althoughmostofthemarenotstatist–1 ticallysignificant.Moneymanagersasagrouplikelyreduced Thelaggedchangeinthecommoditypriceisusedtocontrol theirlongexposureswhenthefinancialsystemwasunderstress fortheimpactfrompricemovements.Wealsointroduceother duringtheGFC.Thereductionincrudeoilpositions—theconvariablesliketheBDRYandtheindustrialproductionindexto tract arguably most strongly connected to global economic controlformacroeconomiceffects. conditions—wasstatisticallysignificant.Overall,the money Weconducttheregressionanalysisonsixmajorcommodity managers response during the GFC was consistent with the futures contracts traded on the CME: wheat, soybeans, cop- convectiveriskflowtheory. per,gold,naturalgas,andcrudeoil.Thosecommodityfutures Table7showstheregressionresultsfortheweeklychanges contractsareheavilytraded.Foreachofthecommodities,we innetoftrade,whichisthechangeofnetpositiondividedby runtheregressionmodelusingfourdifferentsampleperiods: thetotalopeninterest.Thisvariablemeasurestraders’position thepre-GFCperiod,theGFCperiod,thepre-COVIDperiod, changerelativestothetotalopeninterest.Ifopeninterestwas andtheCOVIDperiod(includingthe2022Russianinvasionof constant, changes in net of trade for producers implies the Ukraine).Fortheseregressions,wedonotaggregateintospec- hedgingdemandfromcommodityproducersandchangesin ulators,butratherruneachregressionseparatelyforthethree netoftradeformoneymanagersmeasuresthetradingpressure differentmarkettypesidentifiedintheDCOTdata:producers, from speculation activity. The imbalance in the net of trade swapdealers,andmoneymanagers.Intotal,12regressionsare studiedforeachcommodityfuturecontract.

15 TABLE 7 RegressionCoefficientoftheFSIforTraders’NetofTrade–GFC FSICoefficients Producers SwapDealers MoneyManagers Commodities Pre-GFC GFC Pre-GFC GFC Pre-GFC GFC –0.001 0.001 –0.0001 0.001 –0.001 –0.003** CrudeOil (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) 0.001 0.01*** –0.0004 –0.002*** 0.0002 –0.003* Wheat (0.002) (0.002) (0.001) (0.001) (0.002) (0.002) –0.001 0.001** –0.001 0.001 0.0002 –0.003 NaturalGas (0.001) (0.0005) (0.001) (0.001) (0.002) (0.002) 0.002 0.01*** –0.002 0.001 –0.001 –0.004** Copper (0.003) (0.002) (0.002) (0.002) (0.003) (0.002) –0.001 0.002 –0.01** 0.001 0.003 –0.003 Gold (0.003) (0.002) (0.003) (0.001) (0.005) (0.002) 0.0001 0.01*** 0.001 –0.002** –0.001 –0.005** Soybean (0.003) (0.002) (0.001) (0.001) (0.002) (0.002) Note:***p<0.01,**p<0.05,*p<0.1;standarderrorsin() TABLE 8 FSICoefficientforProducers’ShortandSpeculators’LongPosition–GFC FSICoefficients Producers’ SwapDealers’ MoneyManagers’ ShortPosition LongPosition LongPosition Commodities Pre-GFC GFC Pre-GFC GFC Pre-GFC GFC 0.005 –0.002 0.004 0.01 –0.002 –0.02*** CrudeOil (0.01) (0.004) (0.01) (0.004) (0.01) (0.01) –0.005 –0.01*** 0.0002 –0.005*** 0.003 –0.004 Wheat (0.01) (0.005) (0.003) (0.002) (0.01) (0.01) 0.01 –0.002 –0.01* 0.001 –0.01 –0.02** NaturalGas (0.01) (0.005) (0.01) (0.004) (0.01) (0.01) –0.02* –0.01 –0.002 –0.0002 –0.02 –0.02** Copper (0.01) (0.01) (0.003) (0.003) (0.02) (0.01) –0.003 –0.003 –0.05*** –0.001 0.01 –0.01 Gold (0.01) (0.004) (0.01) (0.01) (0.01) (0.005) 0.002 –0.01** 0.004 –0.003 –0.001 –0.005** Soybean (0.01) (0.004) (0.003) (0.002) (0.002) (0.002) Note:***p<0.01,**p<0.05,*p<0.1;standarderrorsin() betweenproducersandmoneymanagerscouldbringvolatility Table8showstheregressionresultsfortheweeklychanges tothefuturesmarket. inproducers’shortpositionsandswapdealersandmoneyman- The regression results for the pre-GFC and GFC sample agers’longpositions.Again,thoseresultsaremostlyconsistent periods are mostly consistent with those from regression on with the regression on net of trade. This finding highlights thechangesinnetposition.WhentheFSIincreases,produc- the fact that during the 2008 GFC, speculators’ net position ersreducetheshorthedgingpositions,whilemoneymanagers changesweremainlyfromthereductioninlongpositions,while closetheirlongexposures.Swapdealersalsoreducetheirex- hedgers’netpositionchangesweredrivenbyclosingshortposiposure to agricultural contracts. As a result, the total open tions.Thetwogroupsreversedtheirtradingpatterns.Insteadof interestdecreases.Normalizingthechangesinnetpositionby supplyingliquiditytohedgersastheyusuallydoinnormalmarthedecreasingopeninterestappearstoyieldmorepowertoour ketconditions,speculatorsdemandedliquidityfromhedgersto estimates,asmorecoefficientsbecomestatisticallysignificant, reducetheirexposurestothecommodityfuturesmarket. especiallyforthemoneymanagers.

16 TABLE 9 RegressionCoefficientoftheFSIforTraders’NetPositionChange–COVID-19 FSICoefficients Producers SwapDealers MoneyManagers Pre- Pre- Pre- Commodities COVID COVID COVID COVID COVID COVID 1141.40 2839.10 –382.26 –4442.99* 4102.71 –5657.70* CrudeOil (2127.29) (1806.09) (3964.64) (2578.42) (6073.16) (3035.44) 1740.99 1615.53 –1101.83 2.73 –2198.96 –1889.56 Wheat (1594.58) (1144.67) (694.16) (600.24) (2006.02) (1470.11) –558.91 135.73 –264.12 –1206.25 2316.48 –2236.72 NaturalGas (1743.39) (1121.74) (2628.67) (1762.65) (4896.29) (3506.02) 5109.09*** 2584.53** –273.94 171.91 –4777.76** –3323.57** Copper (1273.36) (1072.91) (284.65) (303.97) (1822.41) (1285.22) 1657.58 233.50 –3423.74 1604.95 4209.54 –2052.23 Gold (1808.85) (926.90) (3355.86) (1961.17) (4695.30) (2791.77) 2087.00 2839.18 –396.78 –63.99 –528.42 –3162.72 Soybean (4075.40) (2626.93) (1027.43) (1083.79) (3887.75) (2582.55) Note:***p<0.01,**p<0.05,*p<0.1;standarderrorsin() TABLE 10 RegressionCoefficientoftheFSIforTraders’NetofTrade–COVID-19 FSICoefficients Producers SwapDealers MoneyManagers Pre- Pre- Pre- Commodities COVID COVID COVID COVID COVID COVID 0.001 0.002** 0.0001 –0.003* 0.002 –0.003 CrudeOil (0.001) (0.001) (0.002) (0.002) (0.003) (0.002) 0.003 0.01 –0.003 –0.001 –0.005 –0.01 Wheat (0.004) (0.004) (0.002) (0.002) (0.004) (0.01) –0.001 –0.0003 –0.00001 –0.001 0.002 –0.001 NaturalGas (0.001) (0.001) (0.002) (0.002) (0.004) (0.004) 0.02*** 0.01 –0.001 –0.0004 –0.02** –0.01 Copper (0.01) (0.01) (0.001) (0.002) (0.01) (0.01) 0.002 0.001 –0.01 0.004 0.01 –0.01 Gold (0.003) (0.002) (0.01) (0.004) (0.01) (0.01) 0.002 0.01 –0.001 –0.002 –0.001 –0.003 Soybean (0.01) (0.004) (0.001) (0.001) (0.01) (0.004) Note:***p<0.01,**p<0.05,*p<0.1;standarderrorsin() Table9,Table10,andTable11showthesameregression movementduringtheCOVID-19crisis,aswasseeninFigure2. resultsforthepre-COVIDandCOVIDsampleperiods.Ingen- Tradingactivitywasgrowing,andopeninterestincreased,as eral,thecoefficientsfortheFSIarenotstatisticallysignificant. seeninFigure3.ThestartoftheRussianinvasionofUkraine Theimpactsfromthechangesinfinancialstabilityconditions in2022createdapoliticalshocktotheglobalfinancialsystem. are negligible during both sample periods. The COVID-19 TheFSIwaspositiveforseveralmonths,indicatingastressin crisis created a large shock to the financial system, but the theglobalfinancialsystem.However,thestresslevelwasnot overall stress level was not comparable with the 2008 GFC. comparablewiththe2008GFC.Thewaractuallytriggereda Moreover,financialsystemconditionsquicklystabilizedwith strongmarketrallyinthecommodityfuturesmarkets,instead governmentinterventions.Thestressinthefinancialsystemdid ofcreatingasell-off.Overall,thefinancialstressmeasuredby nottransformtoalargeshocktothecommodityfuturesmar- theFSIduringtheCOVID-19crisisperiodwasnotassevere kets.Commoditymarketsactuallyhadaverystrongupward

17 TABLE 11 FSICoefficientforProducers’ShortandSpeculators’LongPosition–COVID-19 FSICoefficients Producers’ SwapDealers’ MoneyManagers’ ShortPosition LongPosition LongPosition Pre- Pre- Pre- Commodities COVID COVID COVID COVID COVID COVID 0.001 –0.01 0.004 –0.02* 0.01 –0.03*** CrudeOil (0.01) (0.01) (0.01) (0.01) (0.02) (0.01) –0.02 –0.03** –0.01 –0.02*** –0.01 –0.03* Wheat (0.01) (0.01) (0.01) (0.01) (0.01) (0.02) 0.004 –0.001 –0.01 –0.01 0.004 –0.02* NaturalGas (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) –0.05** –0.01 –0.01 –0.003 –0.06*** –0.01 Copper (0.02) (0.01) (0.01) (0.01) (0.02) (0.02) –0.01 0.004 –0.01 –0.01 0.005 –0.02 Gold (0.02) (0.01) (0.01) (0.01) (0.02) (0.02) –0.003 –0.01 –0.01 –0.02*** –0.004 –0.002 Soybean (0.01) (0.01) (0.01) (0.01) (0.02) (0.02) Note:***p<0.01,**p<0.05,*p<0.1;standarderrorsin() asthe2008GFC;andtheimpactoffinancialstressontraders’ in the samples (Robe and Roberts 2024).¶ Sun et al. (2023) positionswasrelativelylimited. alsofindthat‘non-reportable’tradersreinforcepositivemarket Ourregressionresultsprovidesomesupportstothehypoth- feedbackintheirtrading.Intheappendix,wedoanalyzesmall esisofconvectiveriskflowtheory:producersincreasedtheir traders’sensitivitytofinancialconditions;theresultssuggest net long positions on commodity futures and speculators— that,particularlyinthepre-COVIDandCOVIDperiods,small particularly money managers who are likely more active in tradersareevenmoresensitivethanlargetraderssothattheir speculating—reducedtheircommodityexposureswhenthefi- activitywouldreinforceconvectiveriskflows. nancial system was severely stressed during the 2008 GFC. Convectiveriskflowtheoryarguesthatduringthe2008GFC, However,weshouldapplycautionwheninterpolatingthere- commodityriskflowedfromspeculatorstohedgers.Ourstudy gressionresults.First,traders’positionchangescouldbemainly showsthat,atgrouplevel,bothspeculatorsandhedgershave driven by the market price movement (For example, by dy- beenunwindingtheircommodityfuturespositionsandreduced namics like in Jiang et al. 2024). As the commodity market the risk exposures. Further studies are needed to understand collapsedin2008,itwasnormalforspeculatorstopromptly themechanismofthisreversedriskflow.Moredetailedtrader closetheirriskexposurestoavoidfurtherlossesfromthenega- positiondatacanshedlightwhoisprovidingliquidityinthe tivemarketmovement.Hedgers,ontheotherhand,couldtake commodityfuturesmarketsduringthe2008GFCandwhatthe theopportunitytoprofitfromtheirshorthedgingpositionsand motivationisfordoingso. reducetheirhedgingpositions.Second,elevatedinitialmargin levelandmarketvolatilitysignificantlyincreasedtradingcost andcouldforcetraderstoreducetheexposures,aswediscussed 6 RELATIVEEXPOSUREDYNAMICS intheprevioussections. Ontheotherhand,ourresultscouldbebluntedbylimitations Previoussectionsexaminedsixindividualcommodityfutures inthedata,bothintermsofthehighlevelclassificationsandthe contracts.Thissectionextendsourempiricalanalysistostudy focusonlargetraders.Forexample,Sunetal.(2023)findthat theaggregatedpositionsofdifferenttypeoftraderscrossthe CommodityIndexTraders(CIT)actaspassivetraders,which sixcommodityfuturesmarket.First,wecalculatethesimple is consistent with Hamilton and Wu (2015). Unfortunately, averageofthepercentageoftraders’longandshortpositions thereisnoclearmeanstoremovetheimpactofCITactivity. overtotalopeninterest,whichshowstherelativeexposuresof CertainlythebulkofCITactivityisincludedintheswapdealers differenttraders.Second,weapplytheregressionanalysisto category,butthepreciseoverlapisnotclear,anditmaychange ¶TheCFTCprovidesseparatedataonCITactivityforagriculturalfutures.Recentdata showsthatCITactivityislargerthanthatofSwapDealers.Intheappendix,weanalyze thepositionchangeofCITandfindthatitdisplayedsimilartradingbehaviorstoswap dealers.Possibly,CITactivityisprimarilyresponsiblefortheswapdealerresults.

18 80 60 40 20 0 2007 2009 2011 2013 2015 2017 2019 2021 2023 tnecreP Producer_Long Producer_Short Speculator_Long Speculator_Short Speculation_Index FIGURE 12 SpeculationIndexandAveragePercentageofTraders’LongandShortPositions investigatewhethertherelativeexposureofhedgersandspecu- TABLE 12 RegressionResultsforAggregateProducers’Position latorschangesunderfinancialstress.Studyingtheaggregated Producers’PositionType positioncouldhighlightthesystematicimpactoffinancialstress LongPosition ShortPosition onthecommodityfuturesmarket.Lastly,westudythespeculationindexproposedbyWorking(1953)toexaminewhether Sample⇒ GFC COVID-19 GFC COVID-19 financialstabilityconditionschangethespeculativetradingbe- 0.0003 0.104* 0.014 0.130* FSI (0.039) (0.056) (0.056) (0.078) haviors.Thisindexalsoreflectstherelativeexposurebetween speculatorsandhedgers. –5.480*** –2.808** 8.304*** 3.273** GSCI (1.259) (1.083) (1.808) (1.508) Figure12showsthedailyhistoryofthespeculationindex andtherelativeexposuresofproducersandspeculators.For BDRY –0.604 –0.322 0.562 –0.217 (0.467) (0.345) (0.671) (0.480) illustrativepurpose,thespeculationindexisscaledupbyafac- –0.807 4.863** 6.733 0.088 torof30.Ingeneral,thespeculationindexhasbeensteadily GSCIVol. (2.960) (2.148) (4.253) (2.991) increasingsincethe2000s.Nomajorstructuralchangeswere Observations 151 151 151 151 observedduringthetwocrises.Thisresultimpliesthatfinan- R2 0.166 0.154 0.166 0.035 cialstresshasnotsignificantlychangedtherelativeexposure AdjustedR2 0.144 0.130 0.143 0.009 between hedgers and speculators. During the 2008 GFC pe- Note:***p<0.01,**p<0.05,*p<0.1;standarderrorsin() riod,speculatorsreducedtheirlongexposureswhilehedgers closedtheirshortpositions,therelativeexposurebetweenthose twogroupsdidnotchangesignificantly.AftertheCOVID-19 financial stress increases. For speculators, the coefficient is lockdowns,ascommodityfuturespricesgainedastrongmo- negativeduringthe2008GFC,meaningspeculatorsreduced mentumtomoveupward,speculativetradingactivitiesoutgrew exposureswhenfinancialstabilityconditionsdeteriorated.The thehedgingactivitiesandthespeculationindexincreased. speculationindexalsoshowsanegativerelationshipwiththe Followingtheframeworkoftheprevioussection,weruna FSI.ThecoefficientsarenegativeforbothGFCandCOVID-19 regressionanalysisontheweeklychangesinproducers’and sampleperiods.Overall,thosecoefficientsarerelativelysmall, speculators’relativeexposures,measuredbythepercentageof indicatingthattheimpactoffinancialstabilityontraders’rellongandshortpositionsovertotalinterest.Theregressionis ativeexposuresislimitedingeneral.Forexample,one-point estimatedusingtwodifferentdatasamples:the2008GFCand increasingintheFSIindexonlychangesproducers’shortposithe COVID-19 crisis. Tables 12 and 13 show the regression tionsby0.104percent.Insummary,financialstressdidslightly resultsforproducersandspeculatorsseparately. changetraders’relativeexposuresduringthetwocrises. Forproducers,theregressioncoefficientsoftheFSIareall ThecoefficientoftheS&PGSCIaremorestatisticallysignifpositive,andtheyaremorestatisticallysignificantduringthe icantcomparedwiththecoefficientfortheFSI,implyingthat COVID-19period.Producers’relativeexposuresexpandswhen traders’ relative exposures change more noticeably with the

19 TABLE 13 RegressionResultsforAggregateSpeculators’PositionandSpeculationIndex Speculators’PositionType LongPosition ShortPosition SpeculationIndex Sample⇒ GFC COVID-19 GFC COVID-19 GFC COVID-19 –0.0110 0.092* 0.029 0.076 –0.004*** –0.011** FSI (0.028) (0.050) (0.049) (0.055) (0.002) (0.005) 1.001 2.239** –3.273** 1.188 –0.214*** –0.265** GSCI (0.899) (0.966) (1.599) (1.073) (0.053) (0.103) –0.232 –0.370 –0.289 0.440 0.011 0.045 BDRY (0.334) (0.308) (0.594) (0.342) (0.020) (0.033) –1.923 4.162** –3.378 –0.643 –0.342*** 0.108 GSCIVol. (2.114) (1.917) (3.762) (2.130) (0.124) (0.205) Observations 151 151 151 151 151 151 R2 0.020 0.070 0.054 0.027 0.131 0.062 AdjustedR2 –0.007 0.045 0.028 0.0003 0.107 0.036 Note:***p<0.01,**p<0.05,*p<0.1;standarderrorsin() TABLE 14 RegressionResultswithFSIDummyVariableforAggregatePosition PositionType Producers’ Producers’ Speculators’ Speculators’ Speculation LongPosition ShortPosition LongPosition ShortPosition Index –4.966*** 3.740** 2.434*** 0.241 –0.050 GSCI (1.063) (1.490) (0.778) (1.087) (0.075) 0.032 0.147 –0.205 0.733** –0.001 BDRY (0.295) (0.414) (0.216) (0.302) (0.021) –1.890 4.457 –2.476 –1.990 0.034 GSCIVol. (3.487) (4.885) (2.551) (3.564) (0.247) –0.136 0.077 –0.062 0.001 0.002 FSIDummy (0.114) (0.160) (0.084) (0.117) (0.008) –0.145 2.244 –0.998 –1.417 –0.071 GSCI×Dummy (1.313) (1.840) (0.961) (1.342) (0.093) 4.468 –1.862 3.561 1.189 –0.105 GSCIVol×Dummy (4.053) (5.678) (2.965) (4.143) (0.287) –0.278 0.160 –0.119 –0.814* 0.012 BDRY×Dummy (0.449) (0.629) (0.328) (0.459) (0.032) Observations 831 831 831 831 831 R2 0.079 0.048 0.025 0.011 0.007 AdjustedR2 0.071 0.039 0.016 0.002 –0.002 Note:***p<0.01,**p<0.05,*p<0.1;standarderrorsin() commodityprices.Whencommoditypricesincrease,produc- Tofurtherconfirmourresults,werunadifferentsetofreersincreasetheirshorthedgingpositionsandreducestheirlong gressions.Wecreateadummyvariabletorepresentfinancial exposures. Speculators, on the other hand, tend to add more stressbasedonthevalueoftheFSI.IftheFSIisnon-positive, longexposures.Surprisingly,thecoefficientsoftheS&PGSCI the dummy variable is 0, meaning financial stability is norforthespeculationindexareconsistentlynegative.Risingcom- mal;iftheFSIispositive,thedummyvariableis1,meaning modity price actually decreases the speculation index. This financialstabilityisunderstress.Table14showstheregression resultcouldmeanthatproducers’relativeexposuresoutgrow resultsusingthisdummyvariable.Otherthanthecoefficientof speculators’ positions when commodity prices increase, not theS&PGSCI,allothercoefficientsarestatisticallyinsignifinecessarilyimplyingareductioninspeculativetrading. cant.Thisresultconfirmsthatfinancialstabilityconditionshave minimalimpactsontherelativeexposuresforproducersand

20 speculators,andpricemovementplaysakeyroletodetermine commodityfuturesaremainlydrivenbycross-tradingbetween traders’relativeexposures. those two groups: one group provides liquidity to the other. This finding is consistent with our prior analysis. During Undernormalmarketconditions,speculatorsprovideliquidnormalmarketconditions,hedgersopenshortfuturespositions itytohedgers,takinghedgersshortpositionandaccumulating tohedgetheirproductionrisk,whilespeculatorstakehedgers’ netlongexposures;whenfinancialmarketsareundersevere shortpositionsandarecompensatedwithariskpremium.This stress,likeduringthe2008GFC,thetradingflowscouldretradingrelationwasreversedduringthe2008GFC;speculators verse.Hedgersstarttofacilitatethetradesforspeculatorsand paid a risk premium to hedgers in order to close their long supplytheliquiditytothemarket.Furtherstudyisneededto positions.Thetotalopeninterestdecreasedsignificantly,but understandthemechanismofthisreverseriskflowduringthe therelativeexposurebetweenthosetwogroupswasrelatively 2008GFC.Itisimportanttostudywhattriggerstherisk-flow stable. reversal,whoaretheplayersinsidetheproducersgroupprovidingtheliquiditytospeculators,andwhatthemotivationsare forthoseproducers.Moredetailedpositionandtradedataare 7 CONCLUSION neededforthisresearch. The other main finding of this paper is that during both The commodity futures market is highly integrated with the crises, the impact of financial stress on traders’ positions is globalfinancialsystem.Thispaperexaminestherelationship relativelylimitedcomparedwiththeimpactofpricemovement. between traders’ positions and the global financial stability Speculators in general are more reactive to the changes of during the two most recent crises: the 2008 GFC and the financial stability conditions than producers, who are more COVID-19 crisis (including the 2022 Russian invasion of responsive to price movement and economy condition. This Ukraine). We study the dynamic of both absolute position result could explain why we do not observe the reverse risk changesandrelativeexposuresforhedgersandspeculators. flowduringtheCOVID-19crisis,whenfinancialstresswasnot Our regression results provide evidence in support of the assevereasthe2008GFC. convectiveriskflowstheoryproposedbyChengetal.(2015) Ourfindingshavesomestrongimplicationsforregulators duringthe2008GFC.Astheglobalfinancialsystemcameun- seekingtoaddressthefinancialstabilityissuesrelatedtocomderseverestressin2008,speculatorsaggressivelyreducedtheir moditymarketshocks.Afiresalebyfinancialinstitutionsata commodityfuturesexposure.Incontrast,commodityproducers, timewhentheglobalfinancialsystemisunderstressincreases inaggregate,activelytookthelongpositionfromspeculators theprobabilityofamarketcrash,likewhathappenedtoLME and closed their short positions. The 2008 GFC was mainly nickelsfuturesin2022.Commodityproducers,whoareless a banking crisis, reflecting the balance sheet constraints and regulatedingeneralthanfinancialinstitutions,maynotbeable diminishedrisktoleranceoftheinvestmentbanksandtrading tosurvivealargeunexpectedmarketmovementwhentheyare firms.Thefinancialstresslimitedtheircapacitytowarehouse overly exposed to commodity market risk by taking the pocommoditymarketriskandforcedthemtoclosethelongpo- sitions from financial institutions. A default event of a large sitions on commodity futures to limit their losses from the commodityproducercouldbringmorestresstothefinancial massivemarketsell-off. system, triggering a downward spiral in commodity futures Incontrast,duringtheCOVID-19crisis,thecommoditymar- markets.Addressingsuchriskswithinthecurrentmarketand ketshadastrongupwardmovement;The2022Russianinvasion regulatorystructurewheresubstantialtradingoccursbetween of Ukraine created more upward momentum for some com- moreandlessregulatedentitiesischallenging.Theasymmemodities.Largeinvestmentbanksgrewtheirtradingbusiness try of regulation implies that increased regulatory standards across all asset classes, including commodities. Speculators mayhaveunforeseenresultsduetotheasymmetriceffectson alsoincreasedtheircommodityfuturesexposures.Theconvec- institutionalsoundnessversusmarketfunctioning. tive risk flow was not observed during the COVID-19 crisis. Theseresultshighlightedthecomplexityofthedynamicsbetween financial stress and traders’ behaviors. How traders Acknowledgements respondtofinancialstressisacomplicatedmatter,involving thepricemovementoftheunderlyingcommodity,thetypeand WewishtothankourcolleaguesattheBoardandtheFederal severityofstressfinancialinstitutionsexperience,aswellas ReserveBanksofChicagoandDallasformanyfruitfuldiscusthefinancialhealthofthoseinstitutions. sions.ThepaperbenefitedfromcommentsreceivedatEnergy Wealsostudyspeculators’longandhedgers’shortpositions andtheEconomy:MeetingRisingEnergyDemand—the9th relativetototalopeninterest.Wefindthatfinancialstability jointenergyconferenceoftheDallasandKansasCityFederal conditionshadalimitedimpactontherelativeexposuresdur- ReserveBanksin2024.Wealsowishtothanktheorganizers ingbothcrises.Ingeneral,thechangesoftotalopeninterestfor and participants of the J.P. Morgan Center for Commodities

21 Research2025symposiumNewDirectionsinCommoditiesRe- Eberhardt, M. & Presbitero, A.F. (2021) Commodity prices and searchhostedbytheUniversityofColoradoDenver.Thepaper banking crises. Journal of International Economics, 131, 103474. doi:10.1016/j.jinteco.2021.103474. wasgreatlyimprovedfromthefeedbackreceivedtherefrom Ekeland,I.,Lautier,D.&Villeneuve,B.(2019)Hedgingpressureand ourdiscussantHaiboJiang.Nevertheless,theviewspresented speculationincommoditymarkets.EconomicTheory,68(1),83–123. aresolelyourownanddonotnecessarilyrepresentthoseofthe URLhttp://www.jstor.org/stable/45200371 Granger, C.W.J. (1969) Investigating causal relations by econometric FederalReserveBoardoritsstaff;anyremainingerrorsareour models and cross-spectral methods. Econometrica, 37(3), 424–438. soleresponsibility. doi:10.2307/1912791. Hamilton,J.D.(2009)Causesandconsequencesoftheoilshockof200708. BrookingsPapersonEconomicActivity,2009,215–261. URLhttp://www.jstor.org/stable/25652719 ConflictofInterest Hamilton,J.D.&Wu,J.C.(2015)Effectsofindex-fundinvestingoncommodityfuturesprices.InternationalEconomicReview,56(1),187–205. doi:10.1111/iere.12099. Theauthorsdeclarethattheyhavenoconflictofinterestfor Hartzmark,M.L.(1986)Theeffectsofchangingmarginlevelsonfutures thismanuscript. marketactivity,thecompositionoftradersinthemarket,andprice performance.TheJournalofBusiness,59(2),S147–S180. URLhttp://www.jstor.org/stable/2352786 Irwin,S.H.&Sanders,D.R.(2011)Indexfunds,financialization,and DataAvailability commodityfuturesmarkets.AppliedEconomicPerspectivesandPolicy, 33(1),1–31. Most of the data is publicly available. In particular, the URLhttp://www.jstor.org/stable/41237206 Jiang,H.,Kapadia,N.,Xing,Y.&Zhang,Y.(2024)Extrapolativeex- CFTC publishes the DCOT data at https://www.cftc.gov/ pectationsandcorporateriskmanagement.SocialScienceResearch MarketReports/CommitmentsofTraders/index.htm and the Network. WorkingPaper4709420. Treasury department publishes the FSI data at https:// Kang,W.,Rouwenhorst,K.G.&Tang,K.(2020)Ataleoftwopremiums: Theroleofhedgersandspeculatorsincommodityfuturesmarkets.The www.financialresearch.gov/financial-stress-index/.TheBaltic JournalofFinance,75,377–417.doi:10.1111/jofi.12845. DryIndexandtheS&PGSCIarebroadlyavailable,butwere Kinda,T.,Mlachila,M.&Ouedraogo,R.(2018)Docommodityprice accessedhereviaBloomberg. shocks weaken the financial sector? The World Economy, 41(11), 3001–3044.doi:10.1111/twec.12667. Lehecka, G. (2015) Do hedging and speculative pressures drive com- REFERENCES modity prices, or the other way round? Empirical Economics, 49. Abricha,A.,BenAmar,A.&Bellalah,M.(2024)Commodityfutures doi:10.1007/s00181-014-0886-7. marketsunderstressandstress-freeperiods:Furtherinsightsfroma Monge,M.&Lazcano,A.(2022)CommoditypricesafterCOVID-19: quantileconnectednessapproach.TheQuarterlyReviewofEconomics Persistenceandtimetrends.Risks,10(6).doi:10.3390/risks10060128. andFinance,93,229–246.doi:10.1016/j.qref.2023.12.005. OFR(2023)Officeoffinancialresearchfinancialstressindex.Updated Alodayni,S.(2016)Oilprices,creditrisksinbankingsystems,andmacro- daily. financiallinkagesacrossGCCoilexporters.InternationalJournalof URLhttps://www.financialresearch.gov/financial-stress-index/ FinancialStudies,4(4).doi:10.3390/ijfs4040023. Robe, M.A. & Roberts, J.S. (2024) Four commitments of traders Bonnier,J.B.(2021)Speculationandinformationalefficiencyincommod- reports puzzles, revisited: Answers from grains and oilseeds ityfuturesmarkets.JournalofInternationalMoneyandFinance,117, futures markets. Journal of Commodity Markets, 34, 100389. 102457.doi:10.1016/j.jimonfin.2021.102457. doi:10.1016/j.jcomm.2024.100389. Boos, D. & Grob, L. (2023) Tracking speculative trading. Journal of Röthig,A.(2011)Onspeculatorsandhedgersincurrencyfuturesmarkets: FinancialMarkets,64,100774.doi:10.1016/j.finmar.2022.100774. Who leads whom? International Journal of Finance & Economics, Bosch,D.&Smimou,K.(2022)Tradersmotivationandhedgingpressure 16(1),63–69.doi:10.1002/ijfe.410. incommodityfuturesmarkets.ResearchinInternationalBusinessand Sun, H., Bos, J. & Rodrigues, P. (2023) Destabilizing or passive? Finance,59,101529.doi:10.1016/j.ribaf.2021.101529. the impact of commodity index traders on equilibrium prices. Carter,C.A.,Rausser,G.C.&Smith,A.(2011)Commodityboomsand International Review of Economics & Finance, 83, 271–285. busts.AnnualReviewofResourceEconomics,3,87–118. doi:10.1016/j.iref.2022.08.014. URLhttp://www.jstor.org/stable/43202691 Tokic,D.(2012)Speculationandthe2008oilbubble:TheDCOTreport Chen,Y.L.&Yang,J.J.(2021)TraderpositionsinVIXfutures.Journalof analysis.EnergyPolicy,45,541–550.doi:10.1016/j.enpol.2012.02.069. EmpiricalFinance,61,1–17.doi:10.1016/j.jempfin.2020.12.003. Working,H.(1953)Futurestradingandhedging.TheAmericanEconomic Cheng,I.H.,Kirilenko,A.&Xiong,W.(2015)Convectiveriskflows Review,43(3),314–343. in commodity futures markets. Review of Finance, 19, 17331781. URLhttp://www.jstor.org/stable/1811346 doi:10.1093/rof/rfu043. Yang,J.&Zhou,Y.(2013)Creditriskspilloversamongfinancialinstitu- Daskalaki,C.&Skiadopoulos,G.(2016)Theeffectsofmarginchangeson tionsaroundtheglobalcreditcrisis:Firm-levelevidence.Management commodityfuturesmarkets.JournalofFinancialStability,22,129–152. Science,59(10),2343–2359.doi:10.1287/mnsc.2013.1706. doi:10.1016/j.jfs.2016.01.002. Zhang, Y. & Wang, R. (2022) COVID-19 impact on commod- Dedi,V.&Mandilaras,A.(2022)Traderpositionsandthepriceofoilin ity futures volatilities. Finance Research Letters, 47, 102624. thefuturesmarket.InternationalReviewofEconomics&Finance,82, doi:10.1016/j.frl.2021.102624. 448–460.doi:10.1016/j.iref.2022.06.018. Dominguez,K.M.&Reinhart,C.M.(2008)Financialcrash,commodity prices,andglobalimbalances.commentsanddiscussion.Brookings PapersonEconomicActivity,2008,56–68. URLhttp://www.jstor.org/stable/27720395

22 APPENDIX FSIandchangesinCITlongpositionsacrossallfoursample period,suggestingthatasfinancialstressincreases,CITlong FINANCIAL SYSTEM STRESS EFFECTS ON THE positionstendtodecrease.Thisrelationshipisstatisticallysig- POSITIONSOFCITANDSMALLTRADERS nificantinbothGFCandCOVID-19crises.Thoseresultsshow Commodityindextraders(CIT)playacriticalroleinfutures passive CIT investors displaying similar trading patterns as markets, particularly in the agricultural and energy sectors. speculators.However,asitisjustforagriculturalcontracts,the Theseinstitutionalinvestors,oftenlargefinancialinstitutions patternmostcloselyresemblesthefindingsforswapdealers, or pension funds, seek exposure to commodity price move- wherethedeclinewasstatisticallysignificantforwheatandsoy ments by tracking broad-based commodity indices. Unlike (seeTable7). traditionalspeculatorsorhedgers,indextraderstypicallymain- Figure2presentsthesmalltraders’positionsinthosefour tain long positions across a basket of commodities, rolling agricultural futures markets over time. The left panel shows theircontractsforwardastheynearexpiration.Theirapproach theirlongpositions,whiletherightpaneldisplaysshortposiis generally passive, aiming to replicate the performance of tions.LiketheCITandmanagemoneypositions,smalltraders’ commodityindicesratherthanactivelytradingbasedonmar- positionsalsoshowadramaticdecreaseduringthe2008Global ket views. CFTC publishes a supplemental report including FinancialCrisis,indicatingthatsmalltradersmightbealsosen- 13 select agricultural commodity contracts for combined fu- sitivetobroaderfinancialmarketstressjustliketheinstitutional turesandoptionspositions.Supplementalreportsbreakdown investors.WerunasimilarpanelregressionbetweenFSIand thereportableopeninterestpositionsintothreetraderclassifi- smaller traders’ position. Table 2 show the regression result cations:non-commercial,commercial,andindextraders.For for their long position. The panel regression results reveal a smalltraderswhodonotmeetthereportablethreshold,their consistentnegativerelationshipacrossallsampleperiods,with positionswillbeaggregatedtogetherasnon-reportableposition. increasingstatisticalsignificanceaftertheGFC.GivetheFSI Weutilizethereporttostudyhowthefinancialsystemstress has been remained relative stable after the GFC, the regrescanimpactthepositionofpassiveinvestorslikeCITaswellas sionresultscouldimplythatsmalltradersaremoresensitiveto smalltradersduringthe2008GFCandCOVID-19crisis.We thechangeoffinancialsystemconditionthanlargeinstitution focusonfouragriculturalfutures:soybean,corn,wheat-SRW, traders. andwheat-HRW#. Figure1presentstheCITpositionsinthosefouragricultural futuresmarketsovertime.TheleftpanelshowsCITlongpositions,whiletherightpaneldisplaysshortpositions.Anotable featureisthesharpdecreaseinlongpositionsacrossallcommoditiesduringthe2008GlobalFinancialCrisis,particularly evidentincornandsoybeans.Thiscontrastsstarklywiththe COVID-19period,wherepositionsremainedrelativelystable orevenincreased,suggestingdifferentmarketdynamicsand trader responses between these two major economic events. CIT’sshortpositions,whilegenerallymuchlowerthanlong positions,showmorevolatilityandhaveincreasedsignificantly since2020,particularlyforcornandsoybeans. ThesharpdecreaseofCIT’sagriculturallongpositionsduringtheGFCisconsistentwiththechangeofactivespeculators’ positions that we identify in section 5. To test whether increasingfinancialsystemstresscansignificantlyimpactCIT positions,werunapanelregressionbetweenthechangeofCIT positionsandthechangeofFSIforfourdifferentsampleperiods:pre-GFC,GFC,pre-COVID,andCOVID.Table1presents thepanelregressionresultforthelongpositions.Theresults showaconsistentnegativerelationshipbetweenchangesinthe #HardRedWinter(HRW)WheatfuturestradeontheKansasCityBoardofTrade(now CMEGroup),reflectingtheprimarygrowingregioninthecentralandwesternplainsof theUnitedStates.SoftRedWinter(SRW)wheatfuturestradeontheChicagoBoardof Trade(nowCMEGroup),representingproductionintheEasternCornBeltandGreat Lakesregions.

23 600,000 400,000 200,000 0 2010 2015 2020 2025 tnInepO CIT Long Positions 300,000 200,000 100,000 0 2010 2015 2020 2025 CORN SOYBEANS WHEAT−HRW WHEAT−SRW tnInepO CIT Short Positions CORN SOYBEANS WHEAT−HRW WHEAT−SRW FIGURE 1 CITLongandShortPositions TABLE 1 CITLongPositionsPanelRegression ∆ofCITLongPositions Pre-GFC GFC Pre-COVID COVID –210.963 –600.139*** –440.177 –865.058* ∆ofFSI (448.997) (190.398) (323.984) (449.007) Observations 380 448 2,092 588 R2 0.001 0.022 0.001 0.006 AdjustedR2 –0.002 0.020 0.0004 0.005 FStatistic 0.221 9.935*** 1.846 3.712* Note:***p<0.01,**p<0.05,*p<0.1;standarderrorsin() 400,000 300,000 200,000 100,000 0 2010 2015 2020 2025 tnInepO Small Traders' Long Positions 400,000 300,000 200,000 100,000 0 2010 2015 2020 2025 CORN SOYBEANS WHEAT−HRW WHEAT−SRW tnInepO Small Traders' Short Positions CORN SOYBEANS WHEAT−HRW WHEAT−SRW FIGURE 2 SmallTraders’LongandShortPositions

24 TABLE 2 SmallTraders’LongPositionsPanelRegression ∆ofSmallTrader’sLongPositions Pre-GFC GFC Pre-COVID COVID ∆ofFSI –505.909 –312.743 –567.205*** –1259.571*** (749.406) (243.292) (191.789) (263.098) Observations 380 448 2,092 588 R2 0.001 0.004 0.004 0.038 AdjustedR2 –0.001 0.001 0.004 0.036 FStatistic 0.456 1.652 8.747*** 22.920*** Note:***p<0.01,**p<0.05,*p<0.1;standarderrorsin()