How are the responsibilities of derivative traders changing in line with methodological advances? Business Resources Management Paul Shah 2 Sep 2013

Inventory and portfolio trading Analogy with CVA, and where it breaks The role of the portfolio trading desk A case study : a trading model of rate swaps Using technology Conclusions

Inventory and portfolio trading Traditional business management structures to adapt to meet resource constraint challenges Trading floors oriented to inventory trading within defined asset class management lines Hierarchy problem : translation of portfolio resource constraints to asset, business, inventory resource constraints Standalone, marginal, incremental approach to splitting portfolio risks? Standalone robust, but neglects to incentivise macro risk offsets Marginal and incremental useful for management actions but project assumptions on current state to future Reorganise risks and management structure into Inventory (Micro) and Portfolio (Macro) Inventory traders Make markets and manage inventory market risks Are given standard contract assumptions (eg SCSA+) Execute hedges on behalf of Portfolio traders Receive allocations of Portfolio costs net of those hedges Portfolio traders Assume risks that cannot be discretised to trade level May also assume certain trade level risks that funge across asset classes (eg non-standard collateral risks) Receive portfolio funding costs allocations from Treasury Execute portfolio hedges, buffering winner/loser effect at the Inventory level Provide portfolio cost allocations and incentive models to Inventory traders 2

Analogy with CVA, and where it breaks CVA desks as the original Macro desks Portfolio valuation methods analagous to VAR calculation set up to net risks across asset classes CVA traders price trades to net increase/decrease in portfolio cost, hedge embedded greeks Low frequency assumption : netting set does not evolve in realtime Persistence assumption : costs presumed to exist until final maturity of transaction Computational assumption : portfolio tractable to point-of-trade computation Examples of other portfolio costs CCP Initial Margin CSA Thresholds 2notch downgrade Collateral outflow in stressed market Intraday liquidity reserve CCP Initial Margin fails the low frequency and persistence assumption CSA Thresholds fails the computational assumption etc Portfolio risk manager needs to design new techniques for pricing and risk Incentive models (for example, risk accepted within tolerance, pricing fee/benefit for liquidity risk outside tolerance) Level at which control risk and allocate incentive models (business, trading desk, trader) 3

Resource Management Desk Model Trading Floor Desk A Unfunded inventory Funded inventory reserve (stress haircuts, rollover) Loans Derivative VM Initial Margin on Exchange/CCP CVA Derivative Working Capital (Thresholds, CIT, Net IA deposits) 2-Notch Downgrade Provision Collateral Outflow in Market Stress Nostro Pre-Funding Provision Other buffers Debt Buy-backs Debt maturing during scenario window Desk A Book A,B,C Inventory Risk managed and transfer priced at trade level Derivative funding discount risk could be managed either at trade level or risk stripped to a central book Business unit net cost allocations Transfer Pricing and Usage Reporting Resource Management Desk Portfolio Hedging to the Street Financing bake into funding rate 4

The role of the portfolio trading desk Portfolio desk acts as a buffer between Corporate Treasury and trading desks Choice of models to implement funding cost transfer “Bake into rate” : take cost of liquidity reserves and adjust business funding rate : Pros : simple to implement. Cons : charges the wrong trades, distorts trading incentives “Actual cost” : realised cost of unsecured funding and liquidity reserves split by allocation methodology across desks. Pros : transparency to actual usage drivers. Cons : complicated methodology to implement, point of trade assessments tricky “Incentives model” : model designed to discourage behaviour risk-additive beyond some tolerance. Pros : human pilot, implementation can be easier. Cons : could seem arbitrary (eg “ageing cost”). Inventory risk Add haircut unsecured funding cost, repo rollover risk, haircut widening risk to produce blended all-in asset funding cost by inventory category Secured funding desk to manage residual risk (realised spread to blended) and concentration targets Derivative discount risk funges across product, so more efficient to implement a risk-stripping methodology Traders price to CSA discount model PV split into SCSA and “Collateral switch” piece Collateral switch risk stripped to central FVA book, who receive actual collateral balance and provide synthetic balance to desk FVA book manages CSA decisions and portfolio trading decisions (bulk clearing) etc 5

The role of the portfolio trading desk, cont… Resource management desk acts as a buffer between Corporate Treasury and trading desks Portfolio risk Collateral outflow in market stress Can be implemented by a new blended CVA charge/rebate, managed by CVA desk Incentivise risk reduction trades to offset current portfolio outflows Derivative working capital : CCP Margin, thresholds, 2notch Technologically feasible to price CCP margin at point of trade Portfolio vol question Franchise supportive? Incentive model : “Risk accepted” provided CCP Margin stays within certain bounds Implement risk management strategies (portfolio trades, ccp switches) and incremental charges for risk-additive trades administered by FVA desk Allocation of CCP margin cost net of risk mitigation provided to business divisions CSA negotiation, assigns, tearups pnl buffered by FVA desk Winner/loser effect per trade netted out Allocation of net residual cost benefit either buffered centrally or allocated contra-revenue Tradeoff on accruals pnl vs realised pnl 6

Case study : an FVA model for OTC Derivs Portfolio desk manages embedded risks in collateral switch and high threshold pv implements charge/rebate model for CCP IM, LCR costs to incentivise staying within desired risk tolerance facilitates compression/clearing/csa negotiations buffering trading desk pv Counterparty1 Counterparty2 Colt Mgmt Desk Treasury Desk Collateral received CCP Margin funding cost CSA threshold funding cost Non-seg OTC IM benefit LCR costs (eg 2notch, collateral outflow) Desk 1 Desk 2 FVA Desk SCSA collateral Collateral switch pv High threshold pv CCP IM fee model LCR fee model CCP1 CCP2 VM IM Default fund Collateral switch pv = trade pv difference to between actual and standard collateral pv High threshold pv = reserve cost on funding asymettric CSAs above some threshold CCP IM fee model = rebate for risk reduction, charge for increase beyond tolerance, duration assumption FVA Desk sets budget based on expected under/over recovery difference in actual and priced unsec exposure difference in priced rate and Treasury rate CCP and LCR fee model vs actual costs 7

Autofee process for VM runs on EOD curves, CSA curves marked by FVA Using technology Trading Desk pays fee of $1 to FVA desk for IM Trading Desk pays fee of $10-$7 to FVA desk for VM $10 -$3 $7 Trading Desk FVA Desk Client Val Risk Manage: Actual – Standard CSA Price to clients: Actual CSA Risk Manage: Standard CSA Net position: Actual CSA Autofee process for VM runs on EOD curves, CSA curves marked by FVA Autofee process for IM runs according to scripted rule set marked by FVA eg pay Xbp for rec USD delta cleared at CME future : bilateral IM 8

Conclusions Inventory trading desks well equipped to understand costs per inventory unit Portfolio costs discretise well by counterparty but not by asset class Resource management desks “translate” actual funding costs to incentive models Portfolio funding costs not practical to decompose into inventory in many cases Substitute incentive models : eg liquidity charge for trading beyond risk tolerance Resource management desks can also aggregate fungible inventory costs and execute portfolio hedges Collateral switch risk Unsecured exposure risk Frees trader resource and maximises efficiency Risk stripping and incentive models can be built into pricing systems Other models could be explored Sales driven resource management etc 9