A Systemic Risk Assessment Sheri MARKOSE Professor Of OTC Derivatives .

A systemic risk assessment of OTC derivatives reforms and skin‑in‑the‑game for CCPsSheri Markose, Simone Giansante and Ali Rais Shaghaghi(Tucker, 2011; Heath et al., 2015) and can reduceinterconnections in the system. In OTC markets,balanced positions come at the price of complexbilateral offsetting trades which add to the densityof links between the G‑SIB dealers.Hence, with the view to gaining the administrativeefficiency of CCP settlement and to reduce thecomplexity of financial links and their lackof transparency, the main thrust of the G20financial reforms mooted at the Pittsburgh Summitin 2009 has been to make it mandatory for allstandardised OTC derivatives (OTC‑D) contractsto be cleared through CCPs along with an extensivecollateralisation programme for both CCP andbilaterally cleared OTC derivatives.As the reliance on CCP clearing increases with theG20 reforms, Cœuré (2014, 2015) has famouslycalled CCPs “super systemic” players. CCPs havebegun to dominate an already crowded centrallyclustered network structure of the global derivativesmarkets with 16 or so G‑SIBs which currentlyaccount for over 85% of derivatives positions3in the OTC domain and as clearing membersof CCPs. The question here is can CCPs copewith an increased burden of clearing derivatives asthis migrates from the OTC domain? Have CCPsbecome too interconnected to fail (TITF)? TITFis a euphemism relating to the moral hazardproblem that the economic repercussions fromfailure of CCPs could be so wide ranging thatthey could become prime candidates to receive taxpayer bailouts (see Wendt, 2015; Blackrock, 2014;Markose et al., 2012).The purpose of this note is to examine frameworksfor assessing the systemic risk from CCPs inderivatives markets. Specifically in view of the TITFstatus of CCPs, in Section 2, we will discuss theproblem of determining the adequacy of CCPcapital in the context of what is now calledskin‑in‑the‑game (SIG) funds that are put at riskin the first tranche of losses to mitigate incentivesfor the CCP to free ride on the resources of clearingmembers or on those of the tax payer.One of the highlights we provide is an assessmentof the extant hybrid system of OTC‑D and CCPsusing the network approach in Heath, Kellyet al. (2016) which is based on the BIS MAGD4report (2013) data on the 2012 derivatives positionsfor the 40 G‑SIBs and using a reasonable OTC‑Dand CCP clearing spilt with five CCPs clearingeach of the main derivatives products.5 Retainingthe VaR method in Heath, Kelly et al. (2016),widely used for the calculation of initial marginand default fund contributions, the systemicrisk analysis is updated to cover the 2015 endderivatives product level data for the 40 MAGDG‑SIBs. Comparisons that can be made at thesetwo points of time provide interesting ballparkfigures for the extent to which progress has beenmade in the direction of mitigating systemicrisk in global derivatives markets. Further, theempirically calibrated hybrid network modelfor CCP and OTC‑D positions of 40 G‑SIBs givesa good basis to include the skin‑in‑the‑game capitalfunding of CCPs in addition to clearing memberinitial margin and default fund contributions toassess improvements in the stability properties ofthe network system. Following Alter et al. (2015),Markose (2012) and Heath, Kelly et al. (2016), werecommend the application of network centralitymeasures for CCPs to estimate the skin‑in‑the‑gamesurcharges that have the best potential to mitigatecontagion losses from clearing member defaultsthat can be transmitted by CCPs. Sections 4and 5 give some empirical evidence for the size ofthe SIG funds and their effectiveness in dealingwith TITF for CCPs clearing each of the fivemain derivatives products using the Heath, Kellyet al. (2016) CCP‑OTC clearing split involvingthe 40 MAGD G‑SIBs (see footnote 5).Finally, we conclude by reiterating the call toarms (see Haldane, 2009; Markose, 2013) for agranular data driven approach of digital maps forthe contractual obligations of G‑SIFIs, especiallyin global derivatives markets, at regular intervals oftime.6 Only this can vitiate the unacceptable levels ofmodel risk that prove a stumbling block to managingsystemic risk in financial markets. This case was alsoBanque de France Financial Stability Review No. 21 - April 2017 - The impact of financial reforms3 For the 2012 data,Markose (2012) showedthat this accounted for 97%of global derivatives interms of gross notional.Brunnermeier et al. (2013)study the CDS market on EUreference entities and note thatthe network of bilateral CDSexposures among counterpartiesresembles a “core‑periphery”structure with the CDS marketcentred around 13 or 14 G‑SIFIs.Likewise, Duffie et al. (2015)who have bilateral CDSexposure data for all participantsin the single name global CDSmarket, confirm a similarstructure of high concentrationof links around 13 G‑SIBs whodominate the CDS market.4 This stands for the Bankfor International SettlementsMacroeconomic AssessmentGroup on Derivatives (MAGD)Report.5In Heath, Kelly et al. (2016),the hybrid OTC‑CCP split inderivatives clearing networkmodel is called Scenario 1:CCP1 clears 75% of allinterest rate derivatives;CCP2 clears 15% Forexderivatives; CCP3 clears 15 %equity derivatives;CCP4 clears 50%credit derivativesand CCP5 clears 20%commodity derivatives.6 Our view is that systemicrisk does not happen overnightbut builds up and hence digitalmaps of extant who‑to‑whomobligations at reasonablyregular intervals can alertauthorities. The current practiceof calibration and simulationexercises undertaken to providereasonable replicas of the realworld interconnections, due to alack of data, can be avoided.113

A systemic risk assessment of OTC derivatives reforms and skin‑in‑the‑game for CCPsSheri Markose, Simone Giansante and Ali Rais Shaghaghimade by Brunnermeier et al. (2013) on why modelsbased on limited segments of G‑SIFI activities can bemisleading and “hence from an ESRB perspective,a holistic view of the exposures map is required.”2 Skin‑in‑the-game: CCPs as “supersystemic” or “super vulnerable”in a hybrid system of clearingThe regulatory reform process has set out extensiveinstitutional mechanisms that ensure that CCPshave: (i) sufficient resources in the form of stableand conservative initial margins that avoidprocyclicality by being precalibrated to meet stressedmarket conditions, (ii) higher capital charges andmargin requirements for non‑standardisable OTCinstruments, drawn up by BCBS and IOSCO (2013),and (iii) other risk management systems to deal withfailure of clearing members (see CPSS‑IOSCO,2012, EMIR, 2012).With regard to (iii) the current practice is forthe CCP to rely on the default fund contributionsfrom clearing members where the fund is calibratedto withstand failure of the two clearing memberswith the largest liabilities under extreme butplausible conditions. This goes by the name ofCover 2 (CPSS‑IOSCO, 2012). There are rulespertaining to how CCPs can mutualise lossesof defaulting members to surviving membersafter exhausting the former’s initial margin andcontribution to the default fund. This scheduleof loss settlement is called the default waterfallstructure. Rule books of CCPs include close outvaluation process and novation procedures foroutstanding positions of defaulting members tosurviving members. CCPs also have so calledassessment powers over surviving members tospecify the replenishment of the funds used inthe mutualised loses of the defaulting members.As CCPs are not public utilities (Lubben, 2014) butprivate firms competing for custom, there could be arace to the bottom in terms of less costly marginingrequirements and default fund contributions for114clearing members,7 and also undercapitalisation. Inorder to mitigate free riding by CCPs and moralhazard due to their TITF status, authorities suchas those implementing the EMIR have includedskin‑in‑the‑game (SIG) requirements for CCPs.CCP SIG is given precedence in the waterfallstructure ahead of the loss mutualisation basedon the prefunded default fund contributions ofsurviving CCP members. The implementation offormal capital requirements for CCPs will bringthem in line with banks which are subject toregulatory capital requirements.8It is customary in such regulations that some formulaicand absolute minimum standards are stipulated.European Union CCPs are required to hold aminimum capital buffer of EUR 7.5 million or,a larger sum sufficient to provide adequate coveragainst a number of risks which include credit,counterparty, business and operational risks.The latter can involve the cost of orderly windingdown. The SIG is viewed as a surcharge on top of theminimum CCP capital requirement. The EMIR SIGrule (Reg. 153/2013 Article 35, §2) specifiesa 25% surcharge on top of the minimum CCPcapital requirement.The discussions, to date, on whether the EMIR SIGrule of a 25% top up on minimum CCP capitalis adequate for the job at hand have mostly takena qualitative perspective or used rule of thumb.The size of the SIG, it has been argued, should belarge enough as the first loss tranche in order toprevent the CCP from free riding on the prefundedmargin and default fund contributions of itsclearing members. In the case of non‑existent orlow CCP SIG along with low initial margin anddefault fund contributions from clearing membersto attract custom, both the CCP and its clearingmembers have or potentially can have highlyleveraged uncollateralised positions that signal moralhazard problems that may require taxpayer bailout.Also, the CCP SIG should not be so large that thethreat of mutualised losses becomes remote andcan lose its power to discipline clearing membersto control the size of their open interest.97 Zhu (2011) in his survey ofa sample of CCPs does not findevidence of an obvious droppingof standards in regard tothis. However, initial margincalculations differ on detailssuch as length of the closeout period for which initialmargin is calculated. Hence,UK CCPs prescribe seven daysas opposed to the five day normand the former needs moreinitial margin than the latter.8See, BCBS 227 (DFCCP).9 Cœuré (2015) notesthat “the purpose of CCPsis to mutualise counterpartyrisk, not to remove it fromclearing members altogetherand bear it themselves.CCPs are risk poolers,not insurance providers”.The International Swaps andDerivatives Association (ISDA),has weighed in on thesuitable size of CCP SIG fund.In response to the EuropeanBanking Authority (EBA)consultation paper, ISDAstated “that having a 50%skin‑in‑the‑game requirementmay not strike the rightbalance between protectingnon‑defaulting membersand ensuring that they haveincentives to bid competitivelyin an auction of a defaultingclearing member’s portfolio ata time when resources needto be replenished.”Banque de France Financial Stability Review No. 21 - April 2017 - The impact of financial reforms

A systemic risk assessment of OTC derivatives reforms and skin‑in‑the‑game for CCPsSheri Markose, Simone Giansante and Ali Rais ShaghaghiAs in principle, the CCP operates a balancedbook and can become a source of financialcontagion only if the residual losses (in excessof the prefunded initial margin) of its defaultingclearing members are passed on in substantialamounts to non‑defaulting clearing members, wewill argue that the role of SIG should be viewedas a Pigou surcharge for the TITF status of CCPs,namely the negative externality that they pose toothers from their systemic vulnerability to theexposures of their clearing members that can arisefrom inadequate CCP capital.3 Frameworks for assessingthe systemic risk from CCPsin derivatives marketsClearly, there has to be empirical analysis to provideevidence for the efficacy or not for the CCP SIGin conjunction with the other CCP resources suchas the prefunded initial margin and default fund.There have been a number of studies that haveattempted to provide calibration and simulationstress test exercises to quantify and assess the riskmanagement capabilities of CCPs mostly in thecontext of the prefunded initial margin and thedefault fund rather than for CCP capital and SIGbuffers. Typically, formulaic calculations are madefor initial margin and default fund contributionsand it is conceivable that CCP SIG can be made ina similar vein and then the stress tests are conductedto see, under different scenarios, how CCPs fareunder extreme but plausible conditions. The latterinclude simultaneous defaults of several largeclearing members (CMs). The main difference inthe methodology of these studies lies in whetherthese stress tests are conducted with a modellimited to a single CCP and its clearing membersor one that can include G‑SIB positions in bothbilateral OTC clearing and with multiple CCPs.Table 1 summarises the key steps in such exercises.In Step 1, after having calibrated open interestpositions of clearing members at the CCPin question or within a hybrid OTC‑CCP splitclearing model,10 the first order of business isto determine the initial margin requirements.Step 2 involves Cover 2 default fund estimates.For both these steps, the best practice (see Linand Surti, 2015) is the conventional VaR typemetrics that are calibrated to satisfy stress period10 Exceptionally,Duffie et al. (2015) have bilateralexposure data for some 30%of the global market of singlename CDS. This data obtainedfrom the DTCC gives a snapshotof this fragment of the financialnetwork for 30 December 2011.T1  Steps in systemic risk assessment in stress test models for CCPand OTC-D clearingNote: At each step, the light blue box highlights the wider liquidity stress, while the darker blue boxes indicatesolvency risks.Step 1 Based on bilateral open interest positions ofderivatives market participants, determine initial margincollateral of OTC-D banks and as clearing membersbased on VaR type metrics precalibrated typically at 99%confidence level (Cij is the collateral for initial margin from ito j, see Box 1).Reduces solvency risk:initial margin reducesuncollateralised derivativesexposures for both OTCG-SIBs and CCPs.Heath, Kelly et al. (2016):initial margins add to liquidityencumbrance ratio of G-SIBsin both OTC and CCP positionsvis-à-vis their available highquality liquid assets.Current levels of CCP clearingat 35%‑45% on averageacross all derivatives productswith CCP fragmentationimplies high levels of collateraldemands, Duffie and Zhu(2011). Heath, Kelly et al.(2016) show this can triggerliquidity contagion.Step 2 Determine Cover 2 Default Fund (DF) contributionsof clearing members (CMs) for CCP.Reduces Solvency Risk:DF provides CCPs buffersagainst exposures to clearingmembers.Heath, Kelly et al. (2016) addsCM default fund contributionsto Liquidity EncumbranceRatio of CMs; New EMIR 2014rule permits CM to reducecapital of clearing members.Step 3 Stress test with extreme mark-to-market variationmargin conditions (this is done pairwise between participantsi and j as in Vij , see Box 1: Vij – Cij Uncollaterised residualliabilties from i to j is a source of counterparty risk.Interject CCP skin‑in‑game ahead of waterfall feature formutualisation of defaulted losses on non-defaulting clearingmember default funds.CCP skin‑in‑game canbe a panacea for TooInterconnected To Fail moralhazard problem; SIG canforestall variation marginhaircuts; latter are realisedlosses on derivatives assetsand threaten insolvency.Banque de France Financial Stability Review No. 21 - April 2017 - The impact of financial reformsCover 2 default fund exhaustedfully in Heath, Kelly et al.(2016) with 3.89 volatilitystress at Step 3.Step 3 stress of 2.67and 3.89 volatility can causeadded liquidity stress as G‑SIBswhich are clearing members ofmultiple CCPs have high liquidityencumbrance (Singh, 2010aand 2010b). Replenishmentof default funds and VMHC canbring about solvency stress.115

A systemic risk assessment of OTC derivatives reforms and skin‑in‑the‑game for CCPsSheri Markose, Simone Giansante and Ali Rais Shaghaghiconditions rather than use point in time estimateswhich suffer from the “paradox of volatility” (Borioand Drehmann, 2011; Markose, 2013; Markoseet al., 2017). The latter, in addition to beingprocyclical, will severely underestimate the riskbuffers needed in the run up to a financial crisis.Step 3 in Table 1 involves stress tests whereinmore extreme market conditions, than for whichprefunded buffers have been calibrated, to drivevariation margins and hence the size of residualuncollateralised positions. The systemic riskconsequences for CCPs and the liquidity andcapital shortfalls are typically assessed by the classicFurfine (2003) style failure of clearing members.Different scenarios involving CCP infrastructurerules and OTC‑D and CCP clearing splits havebeen investigated.Lin and Surti (2015) and Armakola andLaurent (2015) conduct detailed analyses of USand/or European CCPs and their specific clearingmembers.11 Armakola and Laurent (2015)analyse CCP resilience by conducting stresstests based on the capacity of clearing members,as determined by their ratings and defaultprobabilities, to make good on their derivativesobligations. They conclude that in a Cover 2situation with a failure of two major clearingmembers, many CCPs in their sample may faceserious liquidity problems.In Table 1, the pale blue boxes and darker blue boxes,respectively, highlight the wider implications forliquidity and solvency systemic risks as a function ofthe size of the margin and default fund requirementson G‑SIBs for derivatives clearing. The main findingshere show that the key factor in the demand forcollateral is the extent to which counterpartyexposures can be compressed by netting. Duffie andZhu (2011)12 show that multilateral netting benefitsfrom CCPs with few clearing members is limited.Hence, there has to be substantial migration frombilateral OTC to CCP and that too to a single CCPfor all product clearing to achieve close to 40%counterparty exposure reduction when comparedto the case of 100% bilateral clearing which benefits116from multi‑product netting efficiency germaneto OTC markets. Along the lines of Duffie andZhu (2011), for instance Heller and Vause (2011)show that margin requirements can be reduced byup to 15% if both credit default swaps and interestrate swaps are netted by one centralised CCP.Current levels of CCP clearing of derivatives, withgrowing fragmentation of CCPs, are estimated toaverage between 35%‑45%.13 Interestingly, our socalled Heath, Kelly et al. (2016) Scenario 1 OTC‑Dand CCP clearing with the latter being along singleproduct lines resembles Duffie and Zhu (2011)Table 3 Column 8 case which signals 20% reductionin counterparty exposures (see footnote 12).However, with collateralisation of both OTCand CCP exposures, Heath, Kelly et al. (2016)make a careful audit of the high quality liquidassets of the 40 MAGD G‑SIBs and find thatsome of them can suffer liquidity encumbranceof over 87% as a result of their collateralcommitments given in Steps 1 and 2 of Table 1.As pre‑funding of collateral grows, clearly residualuncollateralised counterparty risk from extremevariation margin volatility can be mitigated, butonly at the cost of triggering a liquidity contagionas G‑SIBs become more and more encumbered asmembers of multiple CCPs (see Singh, 2010a, b).At 3.89 volatility 14 stress tests at Step 3 of Table 1,Heath, Kelly et al. (2016) instigate a variationmargin hair cut (VMHC) to the winning sidenon‑defaulting clearing members of some CCPsas they become unbalanced from defaults ofclearing members. They also assume that CCPscan in principle exhaust all the non‑defaultingclearing member share of the default fund if thelosses of defaulting members exceed their initialmargin and default fund contributions. Clearly,this can be considered to be highly permissive offree riding on the part of CCPs and can resultin both solvency and liquidity contagion effects.To date, perhaps with the exception of Lin andSurti (2015), no paper has analysed the role of CCPcapital and SIG funds within a model in whichinitial margin and the default funds have been11 Lin and Surti (2015)study Swapclear for interestrates swaps and ICE for creditdefault swaps, while Armakolaand Laurent (2015) covereight European CCPs andfive US CCPs. Their analysiscan be compared to how CMEconducts its stress 12See Duffie and Zhu (2011)Table 3 column 9. Cont andKokholm (2014) show thatexposure reductions from CCPnetting are greater than whatDuffie and Zhu (2011) haveestimated for high volatilityunderlying assets.13 The 45% figure isgiven in EC Safer FinancialInfrastructure #saferCCPs.14 Based on daily data, this isonly about a one in 8 year event.Banque de France Financial Stability Review No. 21 - April 2017 - The impact of financial reforms

A systemic risk assessment of OTC derivatives reforms and skin‑in‑the‑game for CCPsSheri Markose, Simone Giansante and Ali Rais Shaghaghiquantified. Even more remarkably, despite the callto arms to model the interconnectedness of theextant financial exposures in complex derivativesmarkets (see Brunnermeier et al., 2013) virtually nonetwork analytics has been brought to bear on thestudy of systemic risk or of adequacy of buffers in suchsystems. Cont (2015) succinctly notes how essentialit is to model the links for G‑SIBs who are commonto multiple CCPs and also the OTC connectionsbetween G‑SIBS to make realistic systemic riskassessments of these markets and infrastructure rules.In the context of interconnected systems, we willfollow Alter et al. (2015) and Markose (2012)who find that network centrality, i.e. eigenvectorcentrality, based capital allocation and bailoutsurcharges are best placed to “stabilise” thesystem. Alter et al. (2014) show that other capitalallocation rules are less effective at preventingFurfine (2003) type contagion failures whenthe system is stress tested. Markose (2012) andMarkose et al. (2017) give a more extensiverationale for the use of a recursively derivedfixed point solution for the network centrality offinancial participants in propagating contagionfailures in the system. The principle of a Pigouor externalities tax that is proportionate to thenetwork eigenvector centrality of the financialinstitution to mitigate its TITF was first mootedin Markose (2012).4 Heath, Kelly et al. (2016 ) hybrid OTCand CCP global derivatives networkTable 2 gives the changes that have occurredin the balance sheet data for the 40 G‑SIBsfrom 2012 to 2015‑end with respect to theirtotal derivatives positions, both OTC and CCPcleared. Firstly, note the compression in grossnotional from about USD 755 trillion in 2012to USD 628.24 trillion in 2015, which is abouta 17 % fall. More impressive is the fair value ofderivatives payables that fell by just over a thirdfrom USD 14.34 trillion to USD 9.75 trillion.Derivatives receivables at fair value have fallenT2  Balance Sheet Data (for 40 G‑SIBs)(USD trillions)20152012All banksDerivatives liabilitiesnegative fair valueDerivatives assetspositive fair valueGross notionaloutstandingTier 1 capital20152012Top 16 core 8 534.7312.391.573633.491.34628.2492.630Source: 2012 financial reports data reported in Table 1 of Heath, Kellyet al. (2016); 2015‑end data obtained from financial reportsfor each of 40 G‑SIBs (from BIS MAGD).even more by 37% from USD 14.48 trillionto USD 9.05 trillion.The share of the top 16

A systemic risk assessment of OTC derivatives reforms and skin‑in‑the‑game for CCPs Banque de France Financial Stability Review o. 21 ‑ pril 2017 ‑ he impact of financial reforms (Tucker, 2011; Heath et al., 2015) and can reduce