GS: The Role of Enterprise-Wide Risk Management Systems in Derivatives Pricing
SEE LAST PAGE OF THIS REPORT Howard Mason
FOR IMPORTANT DISCLOSURES 203.901.1635
September 2, 2014
GS: The Role of Enterprise-Wide Risk Management Systems in Derivatives Pricing
“In the old days, there was complexity on products with leverage and pay-off. Now, the derivatives are simple – the complexity is in the accounting. The world has been turned upside down.” Head of markets at a major dealer.
- The financial crisis has transformed the derivatives business which is now relatively less about value-pricing a differentiated product (with a complex payoff, presented as a “customized risk solution”, but also tending to obfuscate fair-value) and more about cost-advantage in commoditized product.
- In addition to dependence on the credit spreads of reporting dealer and counterparty (given, post-crisis, the decisive industry move away from default-free pricing), the relevant costs are increasingly driven by portfolio effects and particularly the relevant “netting set” as defined by regulators and clearing houses. Indeed, dealers now explain uncompetitive pricing to clients in terms of being “wrong way round” at the clearing house (meaning that a given trade adds to, rather than nets against, credit and/or funding costs).
- Investors understand the impact on trade profitability of capital charges, but valuation adjustments can have a greater impact. For example, pioneering analytics provider, Quantifi, estimates the all-in cost of these adjustments at 3 basis points for an at-the-money (ATM) 10-year received-fixed swap versus less than ½ basis point for the cost of capital charges (assuming these are debt-, rather than equity-, funded); in comparison, the bid-ask spread is less than ¼ basis point.
- As shown in the Exhibit below, the relative impact of valuation adjustments is also high for in-the-money (ITM) and out-of-the-money (OTM) swaps as well as for a stressed-, rather than only base-case market-, volatility scenario.
Exhibit: Valuation Adjustments vs. Capital Charges (as par rate adjustments in basis points) for 10Y US$ Receiver Swaps
Source: Quantifi. Annual dollar costs converted to basis points using dv01 of ~8.8. CVA assumes 40% recovery rates, counterparty credit spread of 2%, and own credit spread of 0.5%; this own credit spread also used for FCA and to monetize capital charges; FBA set at zero.
- An increasingly important element of these valuation adjustments, currently accounting for 20-25% of the total, is due to funding; specifically, a funding valuation adjustment or FVA reflects the present-value of expected funding costs for a derivatives-related obligation to post collateral. In 2014Q1, JPM commented that “for the first time, we were able to clearly observe the existing of funding costs in market clearing levels”, and the impact is likely to increase with inter-dealer collateral requirements (as the industry shifts to central-clearing and extends the collateral requirements of central-clearing to bilateral trades) and as rates normalize.
- For example, Deloitte estimates that the cost of posting T-bills or agency securities as collateral will double with an increase in the effective fed funds rate from 0.1% today to 2%.
- Mitigating this increase in the funding costs of collateral will be important to competitive advantage, and requires enterprise-wide technology to link front-office pricing to optimal choices in the middle- and back-offices for both clearing venue (and trading venue when clearing is not “open” because of vertical integration with trading) and cheapest-to-deliver collateral (across repo, OTC derivatives, and listed derivatives businesses, for example). We see GS as the leader in this enterprise-wide risk-management technology, and associated pricing-support tools for traders.
- In May, for example, President & COO Gary Cohn commented: “We made a decision to have a single unified risk management system for all of Goldman Sachs. We have one risk management system that everything in the world goes in that calculates every potential cost, and shows every trader every possible outcome, every possible way to hedge that can go on. And they can literally manipulate anything they want within the system. It’s got an enormous amount of data”.
The core activity of a swaps trader on the interest-rates desk of a dealer is to arbitrage the internal pricing models used to value or “mark” positions against execution prices available from clients. The spread between the execution price and the model price constitutes the trader’s profit, and increasing the spread through derivatives with leverage and complex pay-offs (for a “customized risk solution” which also tends to muddy pricing) was a key driver of pre-crisis profitability and compensation. Today, derivatives are generally simpler but the accounting more complex with Jamie Dimon referring to one element (the debit valuation adjustment or DVA) as “one of the most ridiculous concepts that’s ever been invented in accounting”.
The accounting controversies are important given the reliance of equity investors on financial statements and size of the numbers: in 2014Q1, JPM took a $1.5bn charge related to the accounting for funding expected derivatives-related obligations to post collateral, and C has since indicated it will take a similar charge of ~$550mm later this year. The cornerstone issue is that for instruments such as bonds that settle (where, by definition, settlement occurs when all obligations between counterparties are extinguished) within a matter of days, the model price is simply the market price; however, for instruments such as swaps which may not settle for years, the model price is more theoretical and relies on an idea from mathematical finance. Specifically, arbitrage-free pricing values an instrument as the present value of expected cash flows (between execution and settlement) where discounting is at a risk-neutral rate asserted by the theory to exist and to be unique.
The financial crisis exposed the flaws in using the same “risk-neutral” rate for discounting derivatives-related cash flows as for computing dealer-funding costs, and has led to a profound rethink in derivatives pricing and workflows with implications not only for dealers but also for market infrastructure players such as exchanges and clearing houses. A key reason is that derivatives can no longer be priced in the front-office assuming Libor funding rates for both counterparties, and then passed “over the wall” to the middle- and back-office for clearing and margining; rather, the expected over-the-life cost of funding margin, at the borrowing-cost of the reporting dealer (typically some credit spread over Libor), is now integral to pricing. In addition to a reporting dealer’s “own” credit spread, derivatives pricing is thereby also linked to expected over-the-life collateral requirements which, in turn, depend on a trajectory for the relevant “netting set” of other positions against which the collateral requirement can be offset.
Funding costs are not the only driver of adjustments to derivatives valuation beyond the market factors referenced in the contract, and it is worth reviewing other examples as context.
Adjusting Derivatives Prices for Counterparty Credit Risk
It is hard to imagine now but, in the early 1990s, the notion of marking credit to market was controversial even at derivatives pioneers such as Bankers Trust. Loan officers believed most loans were “money good”, albeit with a buffer from provisions, and that internal credit analysis was more important to loan pricing than reference to the price of publicly-traded bonds. They understood the dangers of over-concentrating in a particular region or industry, and presented grids (to the asset-liability management committee or ALCO) showing credit exposure, measured at book value, by industry-region bucket; there was additional input from risk-scores (“1” for good credits and “5” for less good credits, for example) based on internal credit analysis. The various buckets were subject to credit limits, and active management of the portfolio was restricted to opportunistic action in secondary loan markets that traded by appointment if at all. During the mid-1990s, this loan framework was extended to manage the credit risk arising from interest-rate swaps; for each derivative, exposure was expressed as a “loan-equivalent” amount based on the maximum possible exposure (i.e. replacement-value or in-the-money present value) over the contract life to an agreed, albeit arbitrary, level of statistical confidence. As these loan-equivalent amounts increased with swap activity, banks found themselves pushing up against industry-region credit limits and looked for ways to shed exposure. Credit default swaps (CDS), first launched by JPM in 1997, allowed banks to do so by “buying protection” against counterparty default and launched a new derivatives product category whose outstanding notional peaked at ~$60tn by end-2007 (and now stands at ~$20tn).
Initially, banks managed swaps-related credit risk at the ALCO level by buying default protection in the CDS market to relieve credit-limit constraints on the portfolio. However, as the associated costs increased and particularly after the financial crisis, it was natural for ALCO management to look to pass them down to traders through incorporating into derivatives pricing models a credit valuation adjustment (“CVA”). Conceptually, the CVA for a derivative represents the price of default-risk: the valuation difference between a derivative with a default-free counterparty and the same derivative with a default-prone counterparty. In practice, there is no standard method for computing CVA and approaches differ widely with some end-users relying on current exposure (and pricing it based on current spreads), others computing expected future exposure based on forward rates (and pricing it by approximation through CDS of different maturities), and a few running simulations of possible future outcomes. The largest dealers, with access to quantitative experts and extensive computing resources, employ “copula” simulation of the joint distribution of swap exposure and default probability to capture, among other things, the “wrong-way risk” which arises when the two are positively correlated.
Since the default of major counterparties during the financial crisis, CVA has become a routine part of derivatives pricing and, in January 2013, IFRS 13 required that derivative fair-values incorporate counterparty credit risk and, where possible, reference market-observable credit spreads. As a result, all swaps traders are now credit traders: a rates-trader can perfectly hedge all market risks and yet still see P&L swings from day to day because internal pricing models incorporate a CVA which changes with counterparty credit spreads. Indeed, particularly post-Volker, the majority of trades offset from the standpoint of market risks so that credit can be a main driver of risk. Furthermore, the Basel Committee has commented that “roughly two-thirds of counterparty credit losses [in the financial crisis] were due to CVA losses and only about one-third were due to actual defaults,” and responded by introducing in Basel 3 a capital charge, referred to as the “CVA charge”, to improve bank resilience against mark-to-market losses associated with the deterioration in counterparty creditworthiness.
Adjusting Derivatives Prices for Own Credit Risk
Counterparty credit risk is a two-way street in that if both counterparties incorporate the credit spread of the other in derivatives pricing, but one (the “untenable” dealer) does not incorporate their own credit spreads, there can be no agreement on pricing even with market-observable data for credit spreads. Specifically, the untenable dealer wants to be compensated (through CVA) for the counterparty’s option to default, but does not want to provide that counterparty with compensation for their own option to default. This is obviously unworkable, and the untenable dealer will see market share decline as volumes flow to competitors with a more symmetrical approach to pricing default risk. As a result, derivatives pricing increasingly reflects the default risk of both counterparties and, in particular, the fact that more default-prone dealers will receive inferior pricing (e.g. faces a higher liability through the pay- or liability-leg of a swap) to compensate counterparties for elevated default risk.
The corollary is that if the credit of a default-prone dealer improves, the replacement cost of outstanding derivative contracts declines (because there is less need to compensate counterparties for default risk). In short, the default-prone dealer paid more in the past than would be necessary in the present and this leads to a mark-to-market loss (just as, on a bond liability, there is a mark-to-market loss when the credit spread of a reporting borrower improves). In the case of the derivative, this mark-to-market loss is reportable under fair-value accounting as a “debit valuation account” or DVA line item. This accounting may be counterintuitive, and it may make prudential sense to exclude DVA amounts from Basel 3 capital as is current practice, but it is not ridiculous; in fact, it follows from the less controversial standards (SFAS 157 in the US and IFRS 13 internationally) setting the fair value of a derivative as “the price that would be received to sell an asset or paid to transfer a liability in an orderly transaction”. Given that, through CVA, the market prices default-risk as measured by counterparty credit spreads, the reporting dealer must price in its own credit spread if it is to transact; changes in this credit spread therefore lead to mark-to-market, and hence, reportable “DVA” gains and losses. The net of the CVA (which reduces the derivatives price for the reporting dealer relative to a default-free assumption for the counterparty) and the DVA (which increases the derivatives price for the reporting dealer relative to a default-free assumption for the reporter) is sometimes referred to as the bilateral CVA (“BCVA”).
Adjusting Derivatives Prices for Funding Costs
The cost of insuring against counterparty default or, equivalently, bearing the uninsured default risk is not the only liability (beyond the pay-leg) that arises from a derivatives contract. Dealers are increasingly subject to collateral requirements with the shift to central-clearing and rising “credit support” amounts on bilaterally-cleared trades. Since most trades ultimately offset from a market risk standpoint, margin-funding costs are not meaningful if there are matching collateral requirements on the original trade and its hedge, but this is sometimes not the case. For example, trades with corporates and other collateral-exempt end-users are typically hedged in the inter-dealer market which is collateralized particularly if centrally-cleared; in this case, the original trade with the end-user does not give rise to collateral flows while the hedging trade in the inter-dealer market does. More specifically, when the end-user trade is in-the-money, the present value of the hedge is negative giving rise to a need to post collateral with the hedge counterparty; this collateral is not funded by the original counterparty (because the original trade is uncollateralized) so that the reporting dealer has to fund it with borrowings. Conversely, when the end-user trade is out-of-the-money, the present value of the hedge is positive giving rise to collateral flows from the hedge counterparty; this collateral does not need to be posted to the original counterparty (again because the original trade is uncollateralized) so that the reporting dealer has the opportunity to lend it out.
In a pre-crisis world, where the funding and lending of collateral flows was assumed to occur at the same Libor rate used for discounting, there was no impact on derivatives pricing. Now, however, the assumption is that a dealer will fund a collateral deficit at Libor plus a borrowing-spread giving rise to cash flows which, when discounted at an assumed (and so presumably different) risk-neutral rate, generate a present-value “funding cost adjustment” or FCA to the derivatives price. Conversely, the assumption that a dealer will lend a collateral surplus at Libor plus a lending-spread gives rise to cash flows which, when discounted at a risk-neutral rate, generate a present-value “funding benefit adjustment” or FBA to the derivatives price. Netting the FCA and FBA generates a funding valuation adjustment (FVA) which varies with the lending and borrowing spreads for the reporting dealer. The amounts can be meaningful; in 2014Q1, for example, JPM reported a $1.5bn cost as it implemented FVA and C has indicated it expects a charge of ~$550mm when it implements the methodology later this year.
Impact on Trade Profitability
The cumulative impact of these various valuation adjustments, along with capital charges for risk-weighted assets and CVA, can be meaningful both in absolute terms and relative to capital charges. For example, pioneering analytics provider, Quantifi, estimates the all-in cost of valuation adjustments at 3 basis points for an at-the-money (ATM) 10-year received-fixed swap versus less than ½ of a basis point for the cost of capital charges (assuming these are debt- rather than equity-funded); by comparison, the bid-ask spread is less than ¼ of a basis point. The estimate for the impact of valuation adjustment on trade profitability is also high, relative to the cost of capital charges, for in-the-money (ITM) and out-of-the-money (OTM) swaps; as well as when the assumption for volatility in the underlying interest-rate is increased for a stressed, as opposed to base-case market, scenario (see Exhibit 1).
Exhibit 1: Valuation Adjustments/ Capital Charges (as par rate adjustments in basis points) for 10Y US$ Receiver Swaps
Source: Quantifi. Annual dollar costs converted to basis points using dv01 of ~8.8. CVA assumes 40% recovery rates, counterparty credit spread of 2%, and own credit spread of 0.5%; this own credit spread also used for FCA and to monetize capital charges. FBA is set at zero and there is an assumed 8% capital charge against RWA (net of CVA to avoid a double-count) which are modeled to decline linearly through time.
- There is a case for using the shadow cost of capital which would be higher than the own credit spread at 0.5% and as high as the target ROE (i.e. at least 10%) for firms with tight capital constraints.
- Of course, there is no FVA in the pre-crisis environment when derivatives cash flows are discounted as the same rate assumed for dealer funding costs. Now, however, dealer funding-rates are not assumed to be risk-neutral so that an FVA arises when there is a collateral mismatch and the dealer needs to borrow to fund the difference as can occur when the original trade is with a collateral exempt end-user but the hedge, as is typical, is with another dealer and increasingly centrally-cleared. In this instance, when the original trade is in-the-money, the hedge has negative present value creating a collateral requirement in the inter-dealer market particularly in a centrally-cleared environment; this collateral is not funded by the original (uncollateralized) trade, so that the reporting dealer needs to borrow in the open market.
- In the theory of asset pricing, the condition of no-arbitrage for complete markets is equivalent to the existence of a unique, risk-neutral rate.
- Until the financial crisis, standard desk practice was to compute risk-neutral discount factors by bootstrapping from deposits at the short end of the curve and swaps at the long end; since the financial crisis, the industry is clearer that a risk-neutral discount rate is a theoretical construct and that OIS rates may be a better approximation in practice than Libor. In addition, the industry no longer assumes counterparties are default-free and can fund at Libor flat.
- In a typical credit default swap (CDS) the buyer pays a fee, often similar to the coupon on a bond, in return for receiving a payoff from the seller in the event of default on a referenced issuer.
- Before 2007, CVA was often ignored or too small to have meaningful impact.
- The canonical example is a cross-currency swap where an emerging-market counterparty pays US$ to receive local-currency; in the event of sovereign default, currency devaluation causes dealer exposure to spike at just the time the counterparty credit spread is likely widening.
- In the US, SFAS 157 set this standard in 2007
- Bank for International Settlements, Strengthening the Resilience of the Banking Sector, (2009).
- In book-keeping terms, the value of the pay-leg of the swap increases leading to a credit on the liability side of the mark-to-market balance sheet and hence a mark-to-market loss.
- Recall that arbitrage-free pricing requires cash flows be discounted at a risk-neutral rate which does not change with the “own” credit spread of the dealer.
- There is a case for using the shadow cost of capital which would be higher than the own credit spread at 0.5% and as high as the target ROE (i.e. at least 10%) for firms with a capital deficit.