CR 14 – Derivatives

A derivative is a contract under which two parties agree to make payments to each other, or to buy or sell some underlying security, at one or more future dates. How much the contract is worth at any moment tracks one or more underlying assets or indices, and sometimes on choices the parties themselves make during its life. The horizon can be very short, as with a futures contract that expires in a few weeks, or very long, as with an interest rate swap that runs for decades. Many contracts are deliberately set up so that neither party pays anything at the start, which means the initial value is zero for both sides.

Derivatives have a long history; forward contracts were already in use in medieval Europe. What changed in the last thirty or so years is the sheer scale and complexity of the market. The main attraction remains hedging efficiency. A corporate treasurer facing three typical risks can reach for a different instrument in each case:

In economic terms a derivative often just replicates a position in the cash market synthetically. An airline worried about rising oil prices could store physical oil, but buying cash-settled oil futures gives it the same long exposure without any transport or storage costs. In the other direction, an investor who wants to cut exposure to an asset without selling it can enter a total return swap. The user base runs from sovereigns, central banks and local authorities through to hedge funds, pension funds, asset managers, insurers and ordinary corporates. Because each user has particular hedging and accounting needs, many contracts are bespoke; a company hedging a floating-rate loan wants a swap whose maturity, payment dates and reference rate match the loan exactly.

Where the credit risk comes from

Banks sell derivatives to end users and then hedge the resulting positions with one another, and a fairly small group of large dealers dominates most markets. These dealers are central nodes of the financial system; around 35 banks are currently designated as globally systemically important banks (G-SIBs), a label that broadly means “too big to fail” and that brings tougher rules such as higher minimum capital.

Throughout the life of a contract each party holds a claim on the other, and the size and direction of that claim move with markets. If one party becomes insolvent, the other may not receive what it is owed. This is counterparty credit risk, usually shortened to counterparty risk: the possibility that a counterparty fails to meet its contractual obligations when they fall due. It is managed over the life of the trade through clearing, which happens either bilaterally, where the two sides each look after the risk the other poses, or centrally through a central counterparty (CCP). The collapse of Lehman Brothers showed how damaging the default of a major derivatives player can be, and it pushed counterparty risk to the top of the agenda for banks and regulators.

The simplest derivatives, chiefly futures and listed options, trade on exchanges. An exchange gathers buyers and sellers of standardised contracts in one venue, which concentrates liquidity and makes it cheap to open a position and later close it by selling the identical contract. Prices are visible to everyone, and the exchange also supports price discovery, meaning the process by which the interaction of buyers and sellers sets the price of an asset.

Over-the-counter (OTC) derivatives are negotiated privately between two parties, either a dealer and an end user or two dealers, with no exchange in between. Because no third party is involved they have traditionally been private contracts with little or no secondary market, and historically their trades were not reported anywhere. Their great strength is flexibility. A company that needs to hedge production on particular dates can have those dates written into an OTC contract. On an exchange it would be limited to the listed maturities, and the gap between its own dates and the contract dates would leave it with basis risk. So, for a hedger with precise needs, a customised OTC contract can be the better risk management tool. An industry survey found that over 94% of the largest companies globally rely on derivatives to manage financial risk, and for exactly this reason much of that use is OTC.

The costs of customisation

Flexibility has a price. A client who later wants to exit an OTC trade, and with it the counterparty risk, generally has to go back to the original dealer, who knows it is the only natural buyer and may quote poor terms. Transferring the trade to someone else, called novation or assignment, normally also needs the original counterparty’s consent. This lack of fungibility is a real drawback of the OTC market. Customisation itself is not the problem when it genuinely serves a client’s hedging needs; the criticism of OTC derivatives tends to centre on a small set of exotic products built for regulatory arbitrage or that arguably mislead the buyer.

Any financial transaction passes through three stages. Execution is the moment the buyer and seller commit to the trade. Settlement is the point at which every legal obligation has been met, either because all payments and deliveries have been made or because the contract has been closed out in some other way, for example by offsetting it against another position. Clearing sits in between: it is the work of calculating, and frequently netting, the payment obligations between the parties, together with the management of the trade (margining, interim cash flows) until settlement. One of its central concerns is the risk that a party fails to perform before the contract is complete.

The length of the clearing period is the crucial difference between markets. A spot equity trade clears in a few days and a futures contract within a few months, while a long-dated OTC swap may not settle for years or even decades. A simple three-box picture can hide this, but it is the reason clearing of OTC derivatives has become so important.

Bilateral and central clearing

In bilateral clearing the two parties carry the clearing responsibilities and risks themselves, including any collateral arrangements. In central clearing a third party, the CCP, takes on the job of managing counterparty risk and its components such as collateral. Every derivatives exchange now has an associated CCP, and exchanges have offered central clearing since the mid-1980s. OTC derivatives were historically cleared bilaterally, but a growing share is now centrally cleared.

Central clearing is in a sense the natural next step after exchange trading, but it needs a degree of standardisation, so a meaningful slice of the OTC market will always stay bilateral. Tools such as multilateral trade compression bring some of the benefits of central clearing into the bilateral world by tearing up offsetting trades.

Back in 1986, total OTC notional stood at about $500 billion, a little below the exchange-traded total. Even then the OTC side was arguably the riskier, because its contracts are longer-dated; a 10-year swap carries far more risk than a three-month future. Over the following twenty years OTC notional grew exponentially, driven by customised hedging and investment demand and by entirely new products. The credit default swap market, for instance, expanded by a factor of 10 from end-2003 to end-2008. By 2008 the global derivatives market had passed $700 trillion of notional, around nine-tenths of it OTC.

Growth later flattened, partly because of the global financial crisis (GFC), which pushed banks away from capital-hungry products and clients away from structured products, and partly because compression exercises removed redundant offsetting trades.

Which asset classes matter most

The OTC market spans five broad classes: interest rate (including inflation), foreign exchange, equity, commodity and credit derivatives. Interest rate products make up most of the outstanding notional, but notional is a poor guide to counterparty risk. Cross-currency swaps are long-dated and exchange principal, so they carry substantial counterparty risk even though most FX products are short-dated. CDSs combine high volatility with significant wrong-way risk.

Why notional exaggerates exposure

The exposure on a derivative is far smaller than on a loan or bond with the same face amount. An interest rate swap exchanges fixed for floating payments, so no principal is ever at risk, and on each payment date only the net difference changes hands. If the counterparty fails, the surviving party simply stops paying and the swap is unwound at some agreed valuation. Where the swap has negative value to the surviving party, it may lose nothing at all, provided the swap can be replaced at no extra cost.

So, the market value of derivatives is a better measure of risk than gross notional. It approximates the loss parties would bear if every counterparty failed and the contracts had to be replaced at current prices. There is a strong incentive to shrink gross market value, and central clearing does this; around 60% of notional reported by dealers, mostly interest rate contracts, is now centrally cleared. Collateral, which further cuts the risk, is not reflected in these gross market value numbers.

Exchange-traded derivatives are settled every day through margin. OTC derivatives are not settled that way, though many are collateralised, with the parties pledging cash or securities against the value of the portfolio so that the net exposure between them is neutralised. Collateral cuts counterparty risk but adds legal and operational risk, a funding cost because the collateral has to be sourced, and liquidity risk if the right amount and type cannot be found in time. (Margin is the term most associated with exchanges and collateral with OTC trades, though the two overlap.)

Three tiers of participant

The market is highly concentrated: the largest 14 dealers hold around four-fifths of total notional and provide most of the liquidity. Many end users have weak credit quality and cannot post collateral, so some counterparty risk is an unavoidable consequence of the derivatives market existing at all.

A special group of end users has exceptionally strong credit: sovereigns, supranational bodies and multilateral development banks. They have historically received collateral from banks without posting any, sometimes subject to rating triggers that would oblige them to post if downgraded. Those one-way arrangements now look expensive because of the funding and capital they consume. Alongside all of these sit third-party service providers offering settlement and margining, collateral management, software, trade compression and clearing, which help reduce counterparty risk and its associated legal and operational risks.

Grouping derivatives by how they are transacted and collateralised gives four categories, listed here from the least to the most complex and risky.

• Exchange-traded. Simple, liquid, short-dated contracts. Margin is posted daily against losses and the CCP guarantees performance of every member. The daily cash payment is called variation margin. This was long treated as the safest corner of the market, though the default of a clearing member at a Nordic commodity exchange showed it is not immune.
• OTC centrally cleared. Contracts too complex, illiquid or non-standard for exchange trading that are nonetheless cleared through a CCP. Regulation now requires standardised OTC trades to be cleared. The CCP usually takes daily collateral in cash, which is also called variation margin although it is not technically settlement.
• OTC collateralised. Bilateral trades, not centrally cleared, on which both parties post cash or securities to mitigate counterparty risk.
• OTC uncollateralised. Bilateral trades on which collateral is not posted, or is posted in smaller amounts or lower quality, typically because an end user such as a corporate cannot commit to it. With nothing mitigating the risk, these trades attract the most scrutiny.

Industry estimates put roughly a tenth of total notional on exchanges and the rest OTC. More than half the OTC market is already centrally cleared, and of the bilateral remainder about four-fifths (80%) is collateralised, leaving 20% under-collateralised.

An end user hedging with derivatives holds a portfolio that leans heavily one way, because its purpose is to offset an economic exposure elsewhere in the business. The mark-to-market (MTM) value of that portfolio swings a lot, and any collateral flows linked to it swing with it. The prospect of having to find large amounts of collateral at short notice is one reason many end users avoid collateral agreements altogether. A directional book also offers little netting benefit. In practice an end user spreads its trades across a handful of bank counterparties, depending on its volume and appetite for risk.

End users often hedge one-for-one, tying a swap’s terms directly to a specific bond issue instead of hedging interest rate exposure across the whole balance sheet. That has costs. Unwinding is awkward because the original dealer may quote poor terms. If the end user holds offsetting hedges, say receiver swaps (receive fixed, pay floating) against its lending and payer swaps against its borrowing, it pays more than it would by hedging the net exposure as a whole. And if a bank counterparty defaults, replacing each trade individually is slower and dearer than replacing a single macro hedge.

The loan and swap mismatch

Consider a company that borrows from a bank at a floating rate and enters a swap with the same bank, paying fixed and receiving floating, to lock in its funding cost. Economically the two pieces fit together, yet they may be treated quite differently. The company may account for them separately and collateralise only the swap. The bank holds the loan in its banking book and the swap in its trading book, so their capital treatment is separate too.

The classic bank set-up

A bank aims to keep its book flat from a market risk point of view, so it hedges each client trade, either across its whole book or directly with a back-to-back trade with another dealer. The hedges pass through the interbank market and may end with another end user.

The bank in this position has little market risk but counterparty risk to both A and B: if either fails, the bank is left with an open position against the other. Client trades are often uncollateralised while the hedges are collateralised, exchange-traded or centrally cleared, so the main counterparty risk sits on the client side, although the hedges still carry some. The asymmetry also causes funding strain. Dealers whose corporate clients mostly receive fixed, for example, hold the opposite side in their hedges, and when interest rates fall those hedges lose value and require heavy collateral postings while nothing comes in from the clients.

Standard legal documentation made the growth of the OTC market possible. The International Swaps and Derivatives Association (ISDA), a trade body for OTC practitioners, first introduced its Master Agreement in 1985, and it is now the documentation most market participants use for their OTC trades.

The ISDA Master Agreement sets up a bilateral framework governing every OTC transaction between two parties. Rather than each trade standing alone, all the trades covered are bound into one legal contract of indefinite term. The agreement has a standard core plus a schedule in which the parties adjust terms to suit themselves, covering matters such as netting, collateral, termination events, what counts as a default, and how close-out works. The economic terms of each individual trade go into a short trade confirmation that refers back to the Master Agreement for everything else. Negotiating the agreement can take a long time, but once it is signed the parties can trade without revisiting the general terms. English or New York law normally governs, though other jurisdictions are occasionally chosen.

Risk-mitigating features

From a counterparty risk point of view, the agreement does four things:

• it sets out the contractual terms for posting collateral;
• it defines events of default and termination events;
• it combines all covered transactions into a single net obligation, which is what makes netting enforceable;
• it specifies the mechanics of the close-out process.

Events of default

A default event ends the transactions before their scheduled maturity and starts the close-out. The agreement lists these events:

Failure to pay and bankruptcy are by far the most common triggers in practice.

Why documentation matters in a default

The Lehman Brothers bankruptcy led to years of litigation over whether different obligations could be offset against one another and how OTC derivative assets and liabilities should be valued. That experience shows how much depends on documentation that clearly defines what happens after a default. To settle fully with a derivative counterparty, three steps are needed: reconcile the full set of transactions, value each one, and agree a net settlement amount. Lehman operated through over 200 registered subsidiaries spread across 21 countries, around a million derivatives transactions, and more than 80 jurisdictions’ insolvency laws in play, so working through those steps took many years.

Credit derivatives transfer credit risk from one party to another. Their market expanded quickly in the decade before the GFC because banks and investors wanted an efficient way to move that risk around. The core instrument, the credit default swap (CDS), is simple: the protection buyer pays a regular premium and the protection seller compensates it if a reference entity suffers a credit event such as default. The CDS changed the way credit risk is traded.

A CDS can be used to hedge counterparty risk on other products, yet it carries counterparty risk of its own, and of an awkward kind. Spread changes drive its value, yet only credit events drive its payoff. The protection seller is most likely to be in trouble precisely when the reference entity is in trouble, particularly when both are financial institutions; buying protection on one bank from another bank is the textbook case. This link between the exposure and the counterparty’s default probability is wrong-way risk. So, credit derivatives can be highly efficient at moving risk, but used carelessly they can be counterproductive and even toxic. The market’s growth has stalled since the GFC.

The push to clear CDSs

The counterparty risk embedded in the CDS market was one of the main reasons regulators pushed standard OTC derivatives towards central clearing. A CDS used as a hedge of counterparty risk only works if the protection seller is unlikely to fail alongside the reference entity, and a CCP seems to offer that default remoteness. The catch is that the CDS is far less liquid and far riskier than the other products CCPs clear, so whether CCPs can cope with it in a real crisis is crucial and remains untested.

The case for derivatives is strong. Industry data show that they support global economic activity, that thousands of firms rely on OTC derivatives to manage business and financial risk, that users range from corporations and investment managers to governments, insurers, energy firms and banks, that they are traded in more than 30 currencies, and that they are used by most of the 500 largest companies in the world. Not every trade is socially useful, however; some exist to arbitrage regulatory capital, tax or accounting rules. Misused derivatives have caused large losses and even catastrophes; Hull discusses well-known cases such as Barings Bank in 1995, Orange County and Procter and Gamble in 1994, Long-Term Capital Management in 1998, Enron in 2001 and Société Générale in 2008.

Leverage

Most derivatives need little or no upfront payment relative to their notional, and that is what gives them leverage. An institution that expects US interest rates to fall can buy Treasury bonds, but the size of that bet is capped by the cash it can raise (a repo can lift the cap, creating leverage much like a derivative does). Entering a receiver interest rate swap gives roughly the same rate exposure with no initial investment, apart from initial margin and capital requirements. So, the size of the position must be limited by the institution, its counterparty or a regulator.

Concentration and systemic risk

A few dealers act as counterparties to huge numbers of end users and trade actively with one another. Before 2008 this was sometimes seen as stabilising, on the belief that none of the dealers would ever fail. It is now seen as a source of systemic risk, where the failure of one institution sets off a chain of failures that threatens the whole financial system. Even a heightened perception of losses, without actual losses, can trigger such a chain.

Buffett’s four criticisms

In 2002 Warren Buffett called derivatives “financial weapons of mass destruction”. His points can be grouped as follows:

Events from 2007 onwards bore these out, and several post-crisis regulations target exactly these problems. The criticism falls mainly on OTC derivatives: exchange listing requires volume, standardisation and liquidity to build up first, while OTC markets move faster, which encourages innovation but also lets large toxic risks grow. Before 2007 the OTC market was largely opaque and unregulated, and although industry initiatives have standardised it, it remains more decentralised, varied and less transparent than exchange markets.

Risk management seldom eliminates financial risk; more often it changes the risk into another form. Collateral, for instance, lowers counterparty risk but brings in market, operational, legal and liquidity risk. The new risk may be milder, but that is not guaranteed, and some of it stays hidden until it bites. Some risks are themselves blends of others, and counterparty risk is mainly a mix of market and credit risk.

Market risk

Market risk comes from short-term moves in market variables. It is linear when it arises from moves in prices such as stocks, rates, FX, commodities or credit spreads, and non-linear when it arises from exposure to volatility or to basis risk in a hedged position. It is the most studied financial risk, partly because of 1990s losses such as Barings and the 1995 amendment to Basel I that let banks use their own models for market risk capital. A position can be made market-neutral with an offsetting contract, but unless that contract is with the same counterparty (or through a CCP, or later compressed away) it creates counterparty risk; if either side fails, the position stops being neutral. So, market risk is a building block of counterparty risk, and uneven collateral and clearing arrangements across the market create funding costs as well. Most banks try to stay market-risk neutral, which keeps accounting volatility and capital down.

Credit risk

A debtor that cannot or will not pay or perform exposes its creditor to credit risk. Quantifying it requires the default probability over the whole life of the exposure and the recovery value (equivalently, the loss given default). Credit deterioration short of default, such as a downgrade, also matters because it raises future default probability and causes an MTM loss, so the term structure of default probability is central. For a bond or loan the exposure is fixed by the par value or notional. For a derivative the exposure is uncertain and driven by market risk. That is why counterparty risk combines credit and market risk.

Operational and legal risk

Operational risk comes from people, systems and internal or external events: trade entry errors, failed settlement or collateral processes, model risk, fraud by rogue traders, and legal risk. Small losses are frequent, but the big ones usually come from improbable combinations of events, which makes the risk very hard to quantify. Collateral management is operationally intensive, so mitigating counterparty risk with collateral inevitably adds operational risk. Legal risk, often treated as a subset of operational risk, is the risk that an expected legal treatment is not upheld because of faulty documentation, fraud, mishandled rights or surprising court rulings. A party may believe it is owed a certain amount by a defaulted counterparty, only for a court to set a lower value, as happened repeatedly in the Lehman litigation. Defaults are rare, so legal mitigants are seldom tested, and courts in different countries have ruled differently.

Liquidity risk

Funding liquidity risk is a main reason some end users refuse bilateral collateral agreements. The issue is the type of asset rather than the amount: every corporate owns property and machinery, but derivatives collateral usually has to be liquid securities, or even cash only, which few outside financial institutions hold in quantity. Banks, partly because liquidity regulation penalises less liquid collateral, prefer to receive cash and other liquid assets, while clients would rather post less liquid ones.

Counterparty risk as an intersection

Counterparty risk is the potential loss from the market value of transactions if a counterparty fails. Because that value moves over time and may come from a model, it holds both a credit element (the default) and a market element (the value at default). The interaction between them, the dependence between default probability and exposure, is wrong-way risk; recovery and collateral also matter. Crises usually involve several risks at once, yet risk types are rarely managed together, and mitigating counterparty risk creates still more risks to integrate.

Systemic risk is the danger of an uncontained crisis that brings down a whole market or financial system. Derivatives can add to it through their complexity, opacity and sheer volume. They can cause an episode, as when a single firm such as Barings suffers huge losses, or catalyse one, as Lehman did. The traditional answer was to make the default of any large participant less likely, using netting and collateral to limit counterparty risk. There is a tension here. Those same tools add complexity and allow the market to grow far beyond the size it would otherwise reach, so measures meant to reduce systemic risk can end up building a dense web of collateralised exposures that becomes a source of it.

The market has tried several structures over the years to shield participants from counterparty failure. Most are now largely obsolete, yet they help explain the strengths and weaknesses of central clearing.

Special purpose vehicles

An SPV (special purpose vehicle, sometimes called a special purpose entity) is a legal entity such as a company or limited partnership set up to ring-fence risk. A firm moves assets into it, or finances a large project through it, so that the rest of the firm and its counterparties are protected. Some jurisdictions require that the SPV is not owned by the entity it serves. In derivatives, the SPV’s purpose is to change the bankruptcy outcome: if the derivative counterparty becomes insolvent, the client is still repaid in full from the SPV ahead of other claims. Structured notes are the main use, where the SPV allows the principal to carry a very high credit quality, typically triple-A, better than the issuer’s own. Rating agencies examine the mechanics and legal details before assigning that rating.

Legal risk replaces counterparty risk

An SPV reorders priorities in bankruptcy, and favouring one group of creditors necessarily disadvantages others, so risk shrinks in one place and grows somewhere else. The SPV turns counterparty risk into legal risk. The most obvious danger is consolidation, where a bankruptcy court treats the SPV as essentially the same as its originator and merges their assets, making the ring-fence worthless. Whether that happens depends heavily on jurisdiction; US courts have a record of consolidating, while UK courts have done so reluctantly, mainly in cases such as fraud.

Lehman used SPVs to protect investors in highly rated structures such as collateralised debt obligations (CDOs) from its own credit risk, which was useful because the bank itself was not triple-A. The key protection was the flip clause, which put investors first in line if Lehman went bankrupt. The US Bankruptcy Court ruled these clauses unenforceable, contradicting UK courts, which ruled them enforceable, and many cases were settled out of court. Documentation that had gone untested for years failed when it was finally needed.

Long before any major dealer had failed, the bilateral, dealer-dominated OTC market was thought to be more exposed to counterparty risk than exchanges. The derivatives product company (DPC) was the industry’s response. A DPC is generally a triple-A rated, bankruptcy-remote subsidiary of a major dealer, set up by one or more banks and, unlike an SPV, capitalised separately so that it earns its own top rating. Most DPCs gained their rating through capital, though some simply relied on the sponsor’s rating. External counterparties were protected against the failure of the parent. Among the first were vehicles sponsored by Merrill Lynch, Salomon, Morgan Stanley and Lehman Brothers.

DPCs were made possible by better risk models and the growth of rating agencies. Each ran its own quantitative model of current credit risk, benchmarked against the triple-A standard, and most allocated capital dynamically to stay within it. The rating rested on three supports:

A DPC also defined in advance what would trigger its own failure, such as a downgrade of the parent, and how the workout would proceed. This pre-packaged bankruptcy took one of two forms. In a continuation structure a manager kept running and hedging the existing positions; in a termination structure the manager closed them out. Either way it was meant to be simpler, and less likely, than an ordinary counterparty bankruptcy.

Why the model broke down

DPCs let banks rated below triple-A trade long-dated derivatives, but the basic question was whether a triple-A subsidiary of a double-A or weaker bank was truly safer to face than the bank itself. The GFC supplied the answer. After Bear Stearns was rescued, J.P. Morgan wound down its DPCs and compensated clients for novating their trades. Two Lehman DPCs filed voluntarily for Chapter 11 protection, a move intended to protect their assets that instead showed how closely a DPC’s fate tracks its parent’s. Rating agencies responded to this lack of independence by withdrawing ratings. The lesson is that a DPC converts counterparty risk not only into legal risk, as an SPV does, but also into market and operational risk, and the conversion may not work. A DPC’s rating should be judged alongside its parent’s, though some DPCs do still exist.

From around 1998 the credit derivatives market grew explosively, and the demand for highly rated counterparties grew with it. A CDS is an unusual challenge because spread changes drive its value while only credit events drive its payoff, and the wrong-way risk in many CDSs made the seller’s credit quality even more important than for other OTC products. Senior tranches of structured finance CDOs carried still more wrong-way risk, and this created a strong demand for a default-remote seller of protection.

Monoline insurers were financial guarantee companies with strong ratings that sold credit wraps, meaning financial guarantees on other parties’ debt. They moved from their traditional business into single-name CDSs and structured finance in search of diversification and higher returns. AIG behaved in much the same way in credit derivatives and can be treated as the same kind of entity. Credit derivative product companies (CDPCs) extended the DPC idea with business models close to the monolines’. To hold top ratings, monolines and CDPCs kept capital linked dynamically to the losses possible on what they had wrapped, much like DPCs. Because of those ratings, they usually did not post collateral when their contracts lost value, at least in normal conditions.

The collapse

Through 2007 the monolines took large MTM losses on the protection they had sold, and doubts grew about whether their capital justified triple-A. Critically, their contracts required collateral if they were downgraded, in some cases merely below triple-A. In November 2007 ACA Financial Guarantee Corporation said that losing its single-A rating would oblige it to post collateral it could not provide. Once the agencies began downgrading, the need to post collateral pushed the monolines into default, and from November 2007 several, including XL Financial Assurance, AMBAC and MBIA, effectively failed. In 2008 AIG received a US government bailout of approximately $182bn. AIG was rescued and the monolines were not mainly because AIG’s exposures were larger and its troubles came so close to Lehman and the problems at Fannie Mae and Freddie Mac. In both cases downgrades, collateral calls and MTM losses fed a downward spiral.

CDPCs were similarly leveraged and typically posted no collateral. They fared somewhat better only because of timing: many were not fully operating until after the GFC started in July 2007, so they avoided the first wave of losses on protection sold, particularly on super senior tranches, whose spreads widened far more than credit spreads in general. Their resemblance to monolines was still noticed, and in October 2008 Fitch withdrew its ratings on the five CDPCs it rated.

The GFC had many causes, including macroeconomic events, government policy, looser lending standards and regulatory failure, but OTC derivatives received a large share of the blame. The market was huge, and the banks at its centre had grown through mergers: Citigroup (Citicorp, Travelers Group, Smith Barney, Salomon Brothers), JPMorgan Chase (built from Chase Manhattan and Bank One) and Royal Bank of Scotland (which bought ABN AMRO). Other firms, notably AIG, had become heavily exposed. Governments had no orderly way to handle institutions whose failure could spread through derivatives, so the Federal Reserve, the Bank of England and other central banks improvised, arranging sales of Bear Stearns and HBOS and putting capital into AIG, Bank of America, Royal Bank of Scotland and others. Lehman showed what happens without a rescue. The problem was not only “too big to fail” but “too interconnected to fail”, and regulators wanted a framework for this daisy-chain risk.

How OTC clearing developed

From the late 1990s CCPs began clearing OTC products. In 1999 LCH.Clearnet (now LCH Ltd) launched RepoClear for repos and SwapClear for plain vanilla interest rate swaps. After Enron’s failure in late 2001, demand for cleared energy derivatives grew, and Intercontinental Exchange (ICE) began clearing them in 2002; it now also clears CDSs. Trades are still negotiated privately and are then novated to the CCP after execution. Before the GFC, though, clearing stayed confined to certain products, suggesting its benefits do not always outweigh its costs.

What a CCP does

A CCP places itself between the two parties, becoming buyer to every seller and seller to every buyer, so that the original counterparty no longer represents a direct risk. It reallocates default losses through netting, margin and loss mutualisation. Its general role is:

• to set the rules and standards that clearing members (and partly their clients) must meet;
• to close out all positions of a defaulting member;
• to hold resources against a member default: variation margin that tracks market moves, initial margin to cover worst-case close-out costs beyond variation margin, and a default fund that mutualises losses in a severe default;
• to have a plan for the extreme case where initial margin and the default fund run out, such as further default fund calls, haircutting of variation margin, and selective tear-up of positions. Only the defaulter’s initial margin may be used.

When a member defaults, the CCP aims to cut its ties quickly and without loss, and it guarantees performance to surviving members, usually not by closing out at market value but by auctioning the defaulter’s positions to other members, which gives continuity. Loss mutualisation spreads losses across all members rather than concentrating them, and multilateral netting reduces how much has to be replaced, damping price impact. CCPs can also move client positions away from a weak member. Many banks and most end users, such as pension funds, clear through a clearing member rather than joining, mainly because membership means taking part in default “fire drills” and bidding in auctions.

Some OTC derivatives, most notably interest rate swaps, were centrally cleared before the GFC, and that part of the market held up well. Lehman’s default did not cause major problems at LCH.Clearnet, where its portfolio was $9trn notional, or at the Depository Trust and Clearing Corporation (DTCC), where it was $500m. Within hours the CCPs suspended the insolvent Lehman entities, preventing further risk building up, while solvent Lehman entities kept trading. Solvent client accounts were moved quickly to other clearing members.

SwapClear

LCH.Clearnet’s London-based SwapClear service cleared interest rate swaps for 20 large banks, Lehman among them. Its portfolio exceeded $100trn notional in 14 currencies, close to half of all interest rate swaps worldwide at the time. LCH.Clearnet had handled earlier defaults, Drexel Burnham Lambert (1990) and Barings (1995), but Lehman became the largest default in CCP history. Lehman Brothers Special Financing Inc. (LBSF) held a $9trn portfolio of tens of thousands of trades, far larger than Lehman’s exchange-traded positions. On 15 September 2008 LBSF failed to make its margin payments and was declared in default within hours. LCH.Clearnet held around $2bn of Lehman’s initial margin for exactly this event, and the task was to close out the 66,390 trades quickly without large losses or knock-on effects.

Where things went less well

Some client margin went to the administrator and was frozen for months, partly because of Lehman’s poor records and partly because UK law then lacked the customer segregation rules that existed in the US; even physical access to Lehman’s offices was delayed. Elsewhere, the Chicago Mercantile Exchange (CME) cleared Lehman positions in interest rates, equities, agriculture, energy and FX, and had to use gains in three asset classes to cover losses in the other two. It has been suggested that the three firms that won the CME auction of Lehman positions made a combined $1.2bn profit. In December 2008 HKEx (Hong Kong Exchanges and Clearing) disclosed a loss of HK$157m on closing out Lehman Brothers Securities Asia. Its clearing house, HKSCC, had to draw on resources including its HK$394m default fund and call extra contributions from its most active members, so surviving members bore default losses.

Before the crisis regulators did not require OTC trades to be cleared. Calls for clearing began after the Bear Stearns rescue. In the US, regulators and market participants set an agenda that included building a CCP for CDSs with robust risk management to reduce systemic risk, and the European Commission described a CDS clearing house as a pressing need. Lehman’s failure in September 2008 strengthened the argument. The underlying belief was that counterparty risk fears had made the crisis worse and that CCPs could contain them.

Regulators saw CCPs as shock absorbers. After a large default, many positions must be replaced quickly in an illiquid market, which moves prices and raises volatility and correlation; those moves can then inflict losses large enough to threaten other firms. CCPs promised transparency, risk reduction through margining, and a single body to set and enforce rules.

The G20 commitments

At Pittsburgh in 2009 the G20 leaders agreed four reforms:

Early debate focused on CDSs, but the mandate covers the entire OTC market, which is unsurprising since credit derivatives were less than one-tenth of OTC notional. Clearing needs reasonable standardisation and liquidity, so not every product qualifies, though the mandate has steadily increased the share that is cleared.

Margin on non-centrally cleared derivatives

A firm could sidestep the clearing mandate by trading products not standardised enough to clear. To remove that incentive, regulators made margin compulsory on bilateral trades as well. The bilateral rules roughly mimic CCP margining, and they can be more punitive, so there is little gain from avoiding clearing. At Cannes in November 2011 the G20 asked the Basel Committee on Banking Supervision (BCBS) and the International Organization of Securities Commissions (IOSCO) to draft consultation margin standards for OTC derivatives that are not centrally cleared by June 2012, with the Financial Stability Board reporting on progress. The BCBS and IOSCO published a consultative paper in 2012. The Working Group on Margin Requirements (WGMR), formed in October 2011 and run jointly by four bodies (the BCBS, IOSCO, the CPSS and the CGFS), developed the framework.

SPVs, DPCs, monolines and CDPCs all proved to have flaws, arguably fundamental ones, which is why few remain. Meanwhile the clearing mandate pushes as much OTC activity as possible into CCPs, so it is worth asking whether CCPs share those flaws.

A CCP is designed to absorb the periodic failure of its members in a controlled way. When a key participant defaults, the CCP guarantees every contract that member cleared, which calms the other institutions and discourages panicky actions that could deepen the crisis. Losses beyond the defaulter’s resources are shared among all members, much as insurance losses are shared among policyholders, instead of falling on the few firms most exposed. That loss mutualisation prevents a domino effect.

Comparison with the earlier structures

The first two rows show where CCPs inherit the old weaknesses; in a large bankruptcy some parties stand to gain a lot by challenging the priority of payments, as in the flip clause cases. The matched book and universal margin are the critical differences from monolines and CDPCs. Even so, CCPs, like those entities, effectively insure against systemic risk, and calling this systemic risk insurance is misleading because systemic risk cannot be diversified away.

Remaining dangers

Concentrating counterparty risk in one large entity creates a potential “too big to fail” institution. One specific danger is procyclical margin: monolines and AIG collapsed because collateral demands rose sharply in a crisis, and a CCP raising margin in a stressed market could create the same dynamic. Central clearing has also had failures, such as a clearing member default at Nasdaq Nordic’s commodity market in September 2018.

Derivatives risk management has become far more quantitative over the last twenty years, and value-at-risk (VAR) is the most significant tool. Built for market risk, it is now used widely to summarise risk in a single number. The potential future exposure (PFE) used for counterparty risk is defined in the same way as VAR.

VAR is the worst loss over a set horizon at a chosen confidence level. VAR at the α% level is a value exceeded with probability no more than (1 − α)%; for a discrete distribution, it is the smallest loss that is exceeded with at most that probability. For a continuous distribution it is simply a quantile.

In the illustrative distribution used to introduce the idea, the 99% VAR is −125 in return terms, a loss of 125, because a larger loss occurs with probability no more than 1%; with discrete outcomes the actual tail probability is 0.92%. VAR needs no distributional assumption such as normality, though some implementations assume one for simplicity.

What VAR misses

VAR says nothing about losses beyond the threshold. In a second distribution with the same VAR, the probability of losing 250 is 1% and no losses fall between 125 and 250, so the 99% VAR is still 125. Making that 250 loss even larger leaves VAR unchanged, because only its probability matters. Relying on VAR alone can create false confidence. VAR also falls short of being a coherent risk measure in the sense of Artzner et al., because it can fail sub-additivity: merging two portfolios should never produce more risk than their separate risks added together.

Expected shortfall (ES) is the average loss given that the loss is at least the VAR. It is less intuitive than VAR but reflects tail losses and is coherent, so the Fundamental Review of the Trading Book requires banks to use ES instead of VAR for market risk.

VAR and ES are most often computed by historical simulation. A history of several years of risk factor moves is applied, interval by interval, to the current portfolio to see how it would have performed. Four years of daily data gives around 1,000 scenarios (about 250 trading days a year times four). For longer horizons the one-day figure is often scaled with the square root of time rule; regulators let banks set the 10-day VAR equal to the one-day VAR multiplied by the square root of 10.

VAR models can be backtested by comparing actual outcomes after the fact with the model’s predictions. VAR suits backtesting because a 99% VAR should be exceeded about once in every 100 observations.

Why counterparty risk is harder

Historical simulation and backtesting work well for market risk because the horizon is short, 10 days at most. Counterparty risk and xVA look years ahead, which makes quantification far more difficult. Models remain indispensable: they reduce volatility, dependence and other market features to a few numbers, rank trades against defined metrics, and do it in seconds. Markets prize models in good times and blame them in bad times, when in truth a model is good or bad according to how it is used, a relationship MacKenzie describes well. PFE, the counterparty risk analogue of VAR, is central to counterparty risk management, but its modelling is complex and easily misread or misused. Regulators now require backtesting of counterparty risk models under Basel III, place more weight on stress testing to capture risks beyond those the models define, and keep xVA methods under continuous scrutiny.

Correlation and dependency

Dependency between financial variables is probably the hardest part of measuring risk. Historically estimated correlations may be a poor guide to the future, especially in volatile markets and crises, when correlations tend to become very large, and correlation itself is a restrictive way to describe dependency. Counterparty risk makes this worse in two ways. Correlations are unstable over time, and counterparty risk must be assessed over many years rather than the single day typical of market risk VAR. Correlation is also not the only measure of dependency; for wrong-way risk in particular, other measures of co-dependence matter. So, xVA calculations require careful treatment of the dependencies among credit risk, market risk, funding and collateral.