CR 15 – Counterparty Risk and Beyond

When two parties sign a financial contract, each one is relying on the other to deliver what the contract promises. Counterparty credit risk, usually shortened to counterparty risk, is the possibility that the other side of the contract fails to honour its obligations, most obviously because it defaults. The term is attached above all to one particular combination of features: a derivative that is traded over-the-counter (OTC) rather than on an exchange, that is cleared bilaterally between the two parties rather than through a central counterparty, and that has no collateral standing behind it. A plain uncollateralised interest rate swap between a bank and a corporate client is the textbook case.

Counterparty risk does not disappear outside that combination, but it becomes smaller because some form of margin or collateral is always present. Four such cases are worth keeping in view:

• Exchange-traded derivatives. These are always centrally cleared, and margin is normally exchanged every day.
• Securities financing transactions such as repos. These are margined daily and are sometimes centrally cleared as well.
• Centrally cleared OTC derivatives. The central counterparty collects margin on a daily basis.
• Collateralised OTC derivatives. Here the two parties agree their own bespoke collateral terms, which govern how margin moves between them.

The risk in these cases is reduced by some mix of collateral, central clearing and short maturities; exchange-traded contracts and repos, for example, tend to run for weeks or months rather than years. A small risk per unit of notional can still add up to a large risk overall, however. Centrally cleared OTC derivatives are the clearest example: the counterparty risk on each trade is thought to be modest, yet the notional that a bank faces with a single central counterparty can be enormous.

A short point on vocabulary helps before going further. Margin is the word traditionally used with exchange-traded derivatives and collateral the word used with OTC derivatives. Strictly, margin is the requirement to settle or to post security, and collateral is the cash or securities actually handed over to meet that requirement. In practice the two words are used almost interchangeably, and this lesson follows that habit.

Traditional credit risk is lending risk. One party owes money to another (through a loan, a bond, a mortgage or a credit card balance) and may be unable to repay some or all of it because it becomes insolvent. Two things make lending risk relatively easy to describe.

First, the amount at risk is reasonably well known at every point in the life of the loan. Interest rates and other market variables move it only a little. A bondholder has roughly the par amount at risk for as long as the bond is outstanding. A repayment mortgage amortises, so the balance falls as payments are made, but the balance in three years can be forecast quite accurately today. A credit card or revolving loan has a maximum facility, and a lender will usually assume it is fully drawn when measuring credit risk, on the reasoning that a borrower who is unable to pay has probably used up most of the line already.

Second, only one side bears the risk. The bondholder can lose money if the issuer defaults, but the issuer suffers no loss if the investor who bought the bond goes bust.

Counterparty risk shares the basic cause of loss with every other kind of credit risk, namely an obligor that cannot or will not pay. What sets it apart are two different features:

• The future value of the contract is uncertain, often highly so. If a counterparty defaults on a derivative at some future date, the amount at stake is the net value at that date of every remaining cash flow under the contract. Seen from today, that value could turn out positive or negative and may swing widely.
• The risk usually runs both ways. Because the contract value can be positive for either party, each counterparty in a derivative is exposed to the other. The two exposures need not be equal. An option bought for an upfront premium creates counterparty risk only for the option buyer, since the seller has already been paid. Differences in credit quality, or collateral terms that require one side to post more than the other, also make the risk asymmetric.

A derivatives book contains far more settlements than trades. A swap exchanges cash flows on many dates, and each cash flow can also trigger a collateral call. For this reason the credit risk on a derivative has historically been split by timing into two parts:

• Pre-settlement risk is the risk that the counterparty defaults at any time before the contract has finally settled. This is what people normally mean by counterparty risk, and the older label is rarely used today, partly because it makes the risk sound less important than it is.
• Settlement risk arises at the moments when payments are actually exchanged, especially at maturity, because the two parties do not always perform their obligations at the same instant.

The two risks have opposite profiles. Settlement risk can involve a very large exposure, up to 100% of the notional, but it lasts only for the settlement window, which is often hours and at most a few days. Pre-settlement exposure is much smaller, being only the market value of the contract, but it is present throughout the life of the trade, and default at some point before expiry is far more likely than default on the one settlement date. Commodity deals where cash is paid against physical delivery spread over a period are an exception, since the settlement window itself is long.

The balance between the two risks depends on the product. A spot FX trade has almost nothing but settlement risk, since there is no life before settlement. A long-dated OTC derivative is dominated by pre-settlement risk, because the settlement window is tiny compared with the years of remaining maturity, even though the settlement exposure itself may be much larger, as the FX example shows. Netting of payments reduces both types.

The Herstatt failure and settlement risk controls

The classic illustration is Bankhaus Herstatt, a small German bank whose licence was withdrawn on 26 June 1974. The closure came after the German interbank payment system had shut at 3:30 pm local time. Several counterparties had already paid Deutschmarks to Herstatt that day, expecting US dollars in return later in New York, where it was only 10:30 am. The dollar payments out of Herstatt’s account were suspended, and those counterparties were left with a loss.

Measuring settlement risk precisely would require tracking contractual payment dates, the time zones involved and how long a bank takes to reconcile accounts in different currencies, because any trade that fails to settle keeps adding to the exposure until it finally does. So most institutions run separate settlement limits rather than folding this risk into counterparty risk. Some mitigation is simple, such as refusing to deliver securities until cash has been received. In FX, most flows now settle through continuous linked settlement (CLS), a multicurrency cash settlement system, and most securities settle by delivery versus payment (DVP), where payment occurs at the moment of delivery. Cross-currency swaps are among the exceptions where settlement risk still has to be recognised.

Collateral does not arrive instantly. Between the last time margin was received and the moment the trades of a defaulted counterparty can be closed out, the portfolio keeps moving with no fresh collateral behind it. That gap is the margin period of risk (MPoR), and it is the risk horizon used for a collateralised or margined portfolio. The usual assumption is 10 business days for bilateral collateralised OTC derivatives and a shorter value of five days for centrally cleared OTC derivatives, which are margined daily.

Because the MPoR is normally much shorter than the remaining maturity of the portfolio, collateral cuts counterparty risk sharply. It does not remove it. Collateralised portfolios still carry material counterparty risk, and one source of it comes from the way settlement and collateral interact.

Why a cash flow can create a collateral spike

Every net settlement of a cash flow changes the value of the remaining portfolio. Suppose the institution pays a coupon on a swap. Once paid, the swap is worth more to the institution, and that extra value is uncollateralised until the next margin call is met. If the counterparty defaults within the MPoR, that increase is lost. The result is a short spike in exposure that lasts for the length of the MPoR. Netting the cash flow and the collateral payment against each other would remove the problem; that is standard practice in exchange-traded markets, but not in OTC markets, where collateral has historically not had to be paid in cash. By convention this risk is still classed as counterparty risk even though it resembles settlement risk.

Margin rules can create settlement risk of their own. Central counterparties usually demand margin in cash in each currency of the portfolio. That splits a multicurrency book into currency silos, so members must make and receive large cash payments in several currencies, which adds to both settlement risk and liquidity strain.

Mitigants range from simple contract clauses to costly infrastructure, and none of them is perfect. Some residual counterparty risk always remains, and measuring that residual can be harder and more subjective than measuring the original risk. The more important point is that a mitigant rarely makes risk vanish. It converts counterparty risk into other forms of financial risk, as the table shows.

A few details deserve emphasis. With collateral, the reuse of collateral received (rehypothecation) and the segregation of collateral posted are central design choices, because they decide who bears the risk if the holder of the collateral fails. Margining terms work much like collateral terms but are usually tied to cash settlement or posting. CCPs act as intermediaries between market participants and so centralise counterparty risk at one point, which brings operational efficiency but means that the failure of a CCP would be a serious systemic event.

Mitigation is a double-edged sword. It reduces existing risk and supports market stability, yet it also loosens constraints such as capital requirements and credit limits, which encourages more trading. Without netting and collateral the OTC derivatives market could never have grown to its present size. It is more accurate to think of mitigation as risk transfer, since every mitigant creates new risks and new costs.

Most counterparty risk comes from bilateral OTC derivatives. A useful way to see where it sits is to look at credit value adjustment (CVA), the price of counterparty risk, broken down by asset class. An industry survey by Deloitte and Solum in 2013 averaged banks’ CVA across asset classes. Interest rate products accounted for by far the largest share, roughly two thirds as read from the survey chart. Foreign exchange came second at a little under a fifth, helped by volatile exchange rates and long-dated cross-currency swaps. Credit derivatives, commodities, exotics and equity each contributed single-digit shares, with credit derivatives notable because of wrong-way risk. These shares include collateralised OTC trades, since credit derivatives, for instance, are always traded under collateral agreements.

Institutions that take significant counterparty risk

Large global banks face counterparty risk in every asset class because they deal in all of them. Smaller banks tend to have narrower exposure, mainly interest rate products and some FX. End users such as corporations may hold only interest rate and cross-currency swaps. Central counterparties concentrate risk from all their members, and the counterparty risk that banks face with CCPs is an area likely to receive more attention in future.

A breakdown by rating and sector

Few banks publish a split of their counterparty risk. Royal Bank of Scotland did so in its 2016 annual report, decomposing CVA by rating and by counterparty type. The table reproduces the figures in £m and adds each row’s share of the total.

Source: Royal Bank of Scotland Annual Report (2016). Change and share columns are derived.

Highly rated counterparties contribute little because their default probability is low. Banks and other financial institutions are also a small slice, reflecting decent credit quality and the likelihood that most of those trades are collateralised. The bulk sits with mid-grade names rated BBB- to A- and with corporates and governments, where trades are mostly uncollateralised. Non-investment-grade names carry less than their high default probability would suggest, probably because banks are reluctant to trade with them in size. The breadth of this table also shows how the coverage of counterparty risk has widened: banks once concentrated on weak and uncollateralised counterparties and ignored the rest as immaterial.

Counterparty risk varies widely from one trade and one counterparty to the next, and any measure must give proper credit for netting and collateral. It helps to judge a new transaction at three levels:

• Trade level: the trade on its own, with all its features and risk factors such as the exact cash flows. This gives its standalone risk.
• Counterparty level: the trade added to everything else facing the same counterparty (or netting set), after netting and collateral. This gives the incremental effect on the existing portfolio.
• Portfolio level: the effect on total counterparty risk across all counterparties, recognising that only a small fraction will default within any given period.

Credit limits, also called credit lines, are the traditional tool for keeping counterparty risk in check over time, and they are the first line of defence. Capping exposure to any one counterparty, sector or region forces diversification, and the size of each cap reflects the underlying credit quality. Spreading trades across many names is not always realistic, because client relationships and business specialisation concentrate activity, and exposures that grow too large should then be reduced by other means.

A credit limit is checked against potential future exposure (PFE), a bad-case exposure at a chosen statistical confidence level, projected through time. For a loan, PFE is roughly the loan amount, but for derivatives it needs quantitative modelling that reflects the trades, the portfolio the limit applies to (usually a counterparty, sometimes a sector or region), current market variables such as rates and volatilities, and contractual mitigants including netting, collateral and hedges. The limit itself is set subjectively from the institution’s risk appetite and can vary with the horizon.

Limits are often reduced at longer horizons, favouring short-term over long-term exposure, because a counterparty’s credit quality has more time to deteriorate. Default probabilities for investment-grade names, both empirical and market-implied, tend to rise with horizon, which supports this. A limit at a future date should also be read as conditional on the counterparty surviving until then, since an earlier default is already covered by the limit at that earlier time.

Incremental PFE and hard and soft limits

Limits are applied dynamically: any new trade, unwind or restructuring that would breach a limit at any future date is normally refused unless specially approved. What matters is the incremental PFE, the portfolio PFE after the trade minus the portfolio PFE before it.

Thanks to netting and diversification inside the portfolio, the incremental PFE is smaller than the trade’s standalone PFE, and a risk-reducing trade can make it negative. A limit can be breached either by a new trade, which is easy to block, or by market moves, which are harder to handle. That is why banks use soft and hard limits. A soft limit may be exceeded through market movements but not through new trades; a hard limit breach forces remedial action such as unwinding or restructuring trades or buying hedges. For example, a soft limit of $10m might block trades that push PFE above $10m, while market moves are tolerated up to a hard limit of $15m. Near a limit, only risk-reducing trades may go through, which is awkward with end users whose OTC activity is mostly one-directional.

Limits give a consolidated view of exposure per counterparty and are the first step in portfolio management of counterparty risk, but they are binary. A fully used limit can block a more profitable trade, and the penalties some banks apply to trades that come close to a limit tend to be ad hoc.

Because credit limits judge a trade by its incremental risk and treat profitability as secondary, they can send the wrong signals. A barely profitable trade may be accepted because the line is mostly unused, while a highly profitable trade may be turned down because the line is nearly full.

Credit value adjustment (CVA) changes the question. CVA is the price of counterparty risk, so instead of asking whether there is room under a limit, the bank asks whether the trade is still profitable once the cost of counterparty risk has been charged. CVA builds the counterparty’s credit quality into that price and so defines the minimum revenue a trade must earn. Seen this way, a credit limit is a crude CVA that takes only two values: zero when the trade is accepted and infinity when it is rejected.

Like PFE, CVA is calculated for the whole set of trades with a counterparty, strictly per netting set since one counterparty may have more than one netting agreement. A new trade should be charged its incremental CVA, the change in CVA once netting with existing trades is captured. One consequence is that CVA adds up across counterparties and is blind to concentration, whether that is one huge exposure to a single name or several exposures to names that are highly correlated because they share a region or sector.

The two tools also pull in opposite directions on the number of counterparties. CVA rewards trading with few counterparties, since that maximises netting, whereas credit limits reward trading with many, since that keeps each exposure small and diversified. So banks use them together. In a typical workflow the credit risk department decides whether to approve a trade against limits, and if it does, the xVA desk prices in the CVA before the trade is done.

A financial product can be priced in two broad ways. The actuarial price is the expected value of future cash flows plus an adjustment for the risk premium; insurers price this way. The risk-neutral, or market-implied, price is the cost of hedging the product, which is the standard approach for derivatives.

CVA sits between the two traditions. It belongs to derivatives, where risk-neutral pricing is the norm and some hedging is possible, but credit risk in banks has long been treated actuarially because loans are illiquid and difficult to hedge. Historically many banks followed the actuarial approach and held CVA as a reserve equal to the expected loss from counterparty risk, much like a loan loss reserve. Today CVA is usually defined risk-neutrally, as the market cost of counterparty risk tied to a hedging strategy, and the larger, more sophisticated banks moved first. Four forces drove the shift:

• Market practice. Large US banks adopted risk-neutral CVA early. Other banks saw this through the prices quoted for novations, in which one party contractually takes over another’s position. A bank stepping into another bank’s shoes on a client portfolio would price CVA differently, forcing the original bank to ask whether its own CVA was market standard.
• Accounting. FAS 157 and IFRS 13 rest on an exit price concept, which auditors have increasingly read as requiring risk-neutral default probabilities. US and Canadian banks reporting under FAS 157 adopted risk-neutral CVA while many European banks under IAS 39 still used actuarial CVA, and IFRS 13, in force outside the US from 2013, has narrowed that gap. The exit price also generally requires a bank to recognise its own credit spread through debt value adjustment (DVA), which many see as problematic.
• Basel III. The capital rules define CVA in terms of credit spreads, which points to the risk-neutral approach, but they do not allow DVA, which clashes with the accounting treatment.
• Regulators. Local supervisors have said that counterparty credit risk models based on expected loss or history that ignore risk premia are not acceptable, and that market-implied risk premia can be observed in CDS and bond markets.

As a result, historical default probabilities are now rarely used in CVA, although other historical parameters still are. Even in 2012 an Ernst and Young survey reported that two banks used a blend of market and historical data and four relied mainly on historical data consistent with their Basel II reporting, and that these six banks were preparing for a more market-driven CVA, booking DVA on derivative liabilities and revising their hedging policies. This leaves an open practical question: how to set a risk-neutral default probability for a counterparty that has no traded credit spread.

The growth of credit derivatives made it possible to hedge counterparty risk. A single-name CDS works like insurance on a set notional of credit risk, paying out if a defined credit event occurs. The most direct use is to buy CDS protection on a counterparty so that its credit limit can be raised, as Figure 3 shows: the protection notional sits on top of the original limit. This is often called a jump-to-default hedge, though wrong-way risk between the protection seller and the counterparty can weaken it.

More tailored products aim at counterparty risk directly. A contingent CDS (CCDS) and a risk participation agreement (RPA) are essentially CDSs whose protection notional follows the exposure on a specified derivative or portfolio, so the counterparty risk of a particular trade can be passed synthetically to a third party. They are used to share the risk of new trades, but their market has never become liquid, and even single-name CDSs are fairly illiquid and cover a limited set of reference entities. In practice, CVA is hedged dynamically against credit spreads, often using CDS indices because they are more liquid than single names, and against market variables such as interest rates and FX rates.

The role of the CVA desk

As CVA grew in accounting and regulatory importance and hedging became more workable, banks set up CVA desks, in the largest banks as early as the late 1990s. The desk prices the counterparty risk of trades originated by trading or sales desks and may take ownership of that risk through an internal transfer price. A typical setup has a trader dealing uncollateralised with a client, hedging the market risk with a collateralised trade, and paying the CVA desk to take over the counterparty risk. Arrangements differ between banks, particularly between large and small ones. Because the same desk increasingly prices collateral, funding and capital as well, the names xVA desk or central desk are now common.

After the global financial crisis (GFC), regulators focused heavily on derivatives and counterparty risk, and accounting changes made CVA a core part of valuation. Alongside CVA, several related costs drew attention, all of them rooted in the existence of counterparty risk:

• Funding. Institutions borrow to finance their activities, and the more risk they carry on their balance sheet, the more that funding costs.
• Collateral. Collateral posted to mitigate counterparty risk can create a cost or a benefit depending on its type, such as its currency or the kind of security.
• Initial margin. Central clearing and some other settings require initial margin, which is over-collateralisation and must itself be funded.
• Regulatory capital. Banks must hold capital against counterparty risk and CVA.

These ideas combine into a picture of the full economic cost of a derivative over its life. The picture depends on thresholds, the levels of portfolio value beyond which collateral must be posted. In reality thresholds are often zero or infinite, but a general threshold shows every case at once.

• Positive value. When the portfolio is in the money (ITM), its uncollateralised part creates counterparty risk and a funding cost. Where collateral is posted, the counterparty may be able to choose which collateral to give, within a contractual range.
• Negative value. When the portfolio is out of the money (OTM), the counterparty faces the bank’s own default risk, and the bank enjoys a potential funding benefit to the extent the position is uncollateralised. If it must post margin, the bank may be able to choose what to post.
• Either way. The capital held against the trade and any initial margin posted must be funded throughout.

Some of these effects are symmetric because derivative values are: one party’s positive value is the other’s negative value, so one side’s CVA cost is the other’s DVA benefit. Capital and initial margin break that symmetry, since both sides bear them and neither receives an offsetting benefit.

Valuation adjustments are known collectively as xVA. Each xVA term measures the cost, or benefit, of one component such as counterparty risk, collateral, funding or capital across the whole life of a trade or portfolio. By convention a cost appears as a positive value and a benefit as a negative value. The calculation takes a profile of the relevant quantity over time, such as expected exposure or the capital amount, and integrates it against the matching cost curve, such as a credit spread, a funding spread, a collateral rate or a cost of capital.

The process starts from a base valuation that leaves out these effects, a clean starting point that matches reality only in special cases. The adjustments then move from the base valuation to the actual valuation:

• CVA and DVA together give the bilateral value of counterparty risk. CVA reflects the counterparty’s default; debt value adjustment (DVA) reflects the institution’s own default as seen by the counterparty.
• FVA (funding value adjustment) captures the cost and benefit of funding the trade. It splits into a funding cost adjustment (FCA) and a funding benefit adjustment (FBA).
• ColVA (collateral value adjustment) values the optionality in the collateral agreement, such as the right to choose the currency or type of collateral, and any other non-standard collateral terms.
• KVA (capital value adjustment) is the cost of holding capital, typically regulatory capital, over the life of the trade.
• MVA (margin value adjustment) is the cost of posting initial margin over the life of the trade.

These labels are fairly standard but not universal, and some effects fit more than one bucket. Collateral must be funded, for instance, so a given cost could sit in FVA or ColVA. Some terms also overlap; the clearest case is DVA and FBA, since the bank’s own default risk is widely seen as a funding benefit. Care is needed to avoid counting such effects twice.

Market and cost components of each xVA term

Counterparty risk combines market risk, which sets the exposure, with credit risk, which sets the counterparty’s credit quality. Every xVA term generalises this: a market component, linked directly or indirectly to the base portfolio value, multiplied in effect by a cost component that prices bearing it.

Every xVA calculation begins with a valuation, usually mark-to-market (MTM). The MTM is not an amount one party owes the other today. It is the present value of all the payments an institution expects to receive minus all those it must make. That one number drives each market component in the table above:

• Credit exposure: the net value of all positions with a counterparty, when positive, is what could be lost today if it defaulted.
• Funding position: a positive value is an asset to be funded; a negative value is a liability.
• Initial margin: where posted, it is usually sized from the variability of the valuation.
• Collateral amount: margin calls are driven mainly by the current value.
• Capital: every capital method is based on the underlying valuation.

Because the payments behind the valuation may fall many years ahead and depend strongly on market variables, each component matters both today and in the future. So xVA needs answers to two questions: what is the value now, and what could it be at each future date, given that the counterparty can default at any time? The first must be answered anyway for ordinary valuation; the second is far harder except in simple cases. Mitigants that change over time, above all margin, add to the difficulty, because future margin requirements and the future value of collateral securities must be projected too.

There is also a circularity. The valuation feeds the xVA calculation, yet a correct valuation should already include xVA. A common practical fix is to start from a base value without adjustments and add the adjustments linearly on top. This is only an approximation, since the true problem is recursive and non-linear.

Replacement cost

Default clauses, including close-out netting and termination rights, refer to replacement cost, the cost of entering an equivalent trade with a new counterparty. Replacement cost is close to the base valuation but not the same: it includes bid-offer spreads, which can be large for illiquid positions (and even a standard contract may become illiquid at the time of a default), and a portfolio may be replaced trade by trade or hedged at a macro level. Derivatives documentation has moved towards referencing replacement cost, defined as objectively as possible, so that genuine rehedging costs can be included in the amount owed between the surviving and defaulting parties. Since replacement cost can itself contain adjustments such as CVA, the circularity reappears. The ability to net trades in default and to adjust for collateral also depends on legal agreements and how a court reads them.

Credit exposure

Credit exposure, or simply exposure, is the loss if the counterparty defaults. It is asymmetric: a positive portfolio value becomes a claim on the defaulted counterparty, but a negative value must still be paid, so the institution loses when it is owed money and cannot gain from being released from a liability, apart from some special, non-standard cases.

Exposure only matters if the counterparty defaults, so strictly it should be measured conditional on default. Market practice often measures it independently of default, which implicitly assumes no wrong-way risk, the link between exposure and the counterparty’s default. That assumption is reasonable for most products but should be kept in mind. Credit exposure is specific to default and so to CVA; funding-related measures need not be conditional on default.

Derivatives can run for decades, so the credit quality of a counterparty must be judged over the whole life of the trades. Two questions matter: how likely is the counterparty to default over a given horizon, where default covers any credit event that could affect the counterparty, and how likely is its credit quality to decline, through a rating downgrade or a wider credit spread, over that horizon?

Credit migration, the discrete shifts in credit quality such as rating changes, matters because it shapes the term structure of default probability, and because a downgrade can cause trouble well before any default. Suppose the one-year default probability is known. The probability of default during a later year, say between four and five years from now, behaves differently for three reasons:

• It tends to fall with horizon, because the counterparty may already have defaulted before the year in question. Default between 20 and 21 years in the future may be very unlikely not because the counterparty is strong, but because a weak name is unlikely to survive 20 years at all.
• A counterparty expected to deteriorate has rising annual default probabilities, until the survival effect eventually reverses this.
• A counterparty expected to improve has falling annual default probabilities, and the survival effect speeds the decline.

Historical rating changes show mean reversion in credit quality: above-average firms tend to deteriorate and below-average firms tend to improve. So a high-quality counterparty usually has a rising default probability over time, while a poor-quality counterparty is more likely to default soon and less likely to default much later.

Real-world vs risk-neutral default probability

Default probability can be defined in two ways. A real-world probability asks how likely the counterparty actually is to default, usually estimated from historical data. A risk-neutral, or market-implied, probability is backed out of credit spreads observed in markets such as CDSs. Because of accounting standards, regulatory rules and market practice, risk-neutral default probabilities have become the standard input for CVA in recent years.

Exposure is measured assuming nothing is recovered, so it is the loss in value or replacement cost before any recovery. Recovery is brought in afterwards. The recovery rate is the percentage of the outstanding claim recovered after a counterparty defaults, and loss given default (LGD) is its complement.

Claims in default vary widely, so LGD is highly uncertain. While credit exposure is traditionally measured on its own, LGD is needed to quantify CVA. In bankruptcy, holders of bilateral derivative contracts normally rank pari passu with senior bondholders, meaning they share the same seniority and should expect the same recovery percentage. OTC derivatives, bonds and CDSs usually all reference senior unsecured credit risk and may seem to share one LGD, but structural features sometimes give certain contracts a lower LGD, or a higher one.

Timing also matters. A defaulted bond can be sold straight away, so its LGD is realised immediately, and CDS contracts settle within days of the credit event through the CDS auction, which also fixes the LGD. A derivative with a defaulted counterparty cannot be freely traded or sold, so its eventual LGD may differ from that of the bonds. These issues were very important in the Lehman Brothers bankruptcy of 2008.

Credit spreads and LGD price the credit risk, but a full xVA analysis also needs the cost of funding, collateral and capital. Like credit spreads, these are hard to pin down and usually require a blend of judgement and quantitative analysis. Three examples show where each cost arises.

• Funding. An institution enters an uncollateralised trade that requires it to pay cash upfront in return for cash flows received later. It must fund that payment, perhaps by issuing a bond, and lenders will demand compensation. Under a collateral agreement, margin would largely offset the upfront payment, so the funding need would mostly disappear. Initial margin is different: being over-collateralisation, it offsets nothing and always creates a funding requirement. Trades with no upfront payment still have funding effects, but the upfront case is the simplest to picture.
• Collateral. Funding is normally defined as the cost of raising cash in a base currency. Collateral in another currency, or in securities, earns a different rate, which should be reflected in the valuation. Receiving bonds may be a funding benefit, but turning them into cash costs the repo rate. If the posting party may choose among currencies or securities, that choice is an option with value.
• Capital. A trade that raises capital requirements must be supported by shareholders’ equity, which expects a return through dividends. That cost of capital belongs in the economics of the trade.

Some inputs can be observed. An institution can look at where its own bonds trade in the secondary market to estimate the cost of new funding. Others are subjective, such as the maturity of funding a given trade really needs, or the dividends shareholders will demand in future.