CR 17 – Margin (Collateral) and Settlement

A long-dated derivative can begin with a small, comfortable exposure and grow, over years of market moves, into a large claim on a counterparty whose credit may have weakened along the way. The most direct defence is a contractual right to demand security as that exposure builds. Margin puts that right into practice: as the value of the trades moves, one party hands cash or securities to the other, and the receiver can keep them if the giver defaults.

Two words circulate for the same idea. Strictly, margin refers to how much value has to be delivered, and collateral to the cash or bond that is physically handed over. OTC documentation traditionally spoke of collateral and exchanges of margin, which is why the OTC “independent amount” and the exchange “initial margin” describe much the same thing. Usage is converging on margin, and the bilateral regulations are themselves called margin requirements, so this lesson uses margin and keeps “collateralised” and “uncollateralised” to say whether margin is present.

Settlement on exchanges, security in OTC markets

The same payment can mean two different things. On an exchange, margin is a daily settlement: the gain or loss is paid away and the position effectively restarts, which is the classic difference between a futures contract and a forward. In bilateral OTC derivatives, margin is security. The receiver becomes the permanent owner only if the giver defaults, when it may keep the margin and set it against losses on its positive portfolio value, provided that right is legally enforceable, as with netting.

Security is flexible, since cash in several currencies, many securities and in principle even non-financial assets can be accepted, but the holder does not own what it holds and faces legal challenge over ownership. Settlement is simpler and its value is realised at once, but it must be cash, most naturally in the trade currency. Survey data put margin supporting bilateral derivatives at about $5trn, with cash gaining share.

Under a two-way agreement either party may post when the portfolio has a negative value from its own side. Trades are marked to market regularly and margin is called or returned in blocks under pre-agreed rules, not continuously, to keep operating costs down.

Margin is never free. It cuts counterparty risk but brings in market, operational, legal and liquidity risk, and it raises funding questions: an end user hedging a loan with a swap must find assets to post on the swap while the loan pays it no margin at all.

Before margin became routine, simpler tools were used to cap long exposures, and some of them behave like crude, infrequent margining. A break clause allows early termination of a trade. It may be mandatory, optional, or triggered by an event such as a downgrade, and may protect one side or both. Banks once demanded breaks mainly from counterparties that did not post margin; after the global financial crisis (GFC), asset managers and pension funds began asking for breaks tied to the banks’ own ratings, because banks had proved less safe than assumed.

The ISDA Master Agreement lets the parties define an additional termination event (ATE), usually in the schedule at counterparty level. A downgrade below a set rating, often investment grade, is the most common trigger; for unrated entities such as hedge funds, a fall in net asset value or market capitalisation, or the loss of a key person, can be used instead. Breaks written into individual confirmations are known as mutual puts, and these usually ignore credit events: the trade either ends on the break date or one side may end it then. A 15-year swap could, for instance, be breakable after five years and every second year from then on. Mandatory breaks sometimes exist because credit limits rule out a long trade, with an informal expectation that a new trade will be booked at the break.

A triggered break need not end the trade. The affected party may instead find a replacement counterparty, bring in a third-party guarantor, post margin, or terminate at current replacement value. Termination is normally at the prevailing mark-to-market (MTM), often set by polling dealers, which as a base valuation avoids most, though not all, of the disputes over close-out values that include xVA.

Which breaks actually reduce risk

A bracketed entry means the benefit is accepted only with caution. For credit limits every break helps somewhat. A CVA desk will usually price only the mandatory break, because it controls neither the exercise of an optional break nor the firing of a rating trigger. The LCR forces banks to pre-fund what rating triggers could demand. Capital relief is realistic only for mandatory breaks, and even then a record of waiving them can undermine it; optional breaks might qualify only if shown to be rigorously exercised, and Basel rules explicitly deny relief for rating-based triggers.

A break that is never exercised protects nobody. Using one can sour a client relationship, so salespeople resist it, particularly when the client is still sound, while the xVA desk and risk managers want breaks used whenever possible. The front office can cite the break to get a trade approved and then argue against exercising it, a moral hazard that banks counter with clear policies on optional breaks, on the benefit credited to them, and on who may exercise them. Banks have become much readier to exercise optional breaks, mindful of xVA and capital, and clients now expect it, so a case can be made that optional breaks deserve some CVA credit after all.

Four weaknesses exposed by the crisis

• Weak against systematic stress. A single counterparty’s downgrade can be handled, but a broad decline across many names, most likely among large financial institutions, is much harder to escape.
• Slow ratings. A break must be used early, before credit quality or exposure has moved far. Ratings react sluggishly, so by the downgrade the counterparty may be beyond rescue, as with monoline insurers in the GFC.
• Cliff-edge effects. If many counterparties hold similar triggers, a small event such as a one-notch downgrade can set them all off at once. AIG’s near failure is the standard example.
• Modelling difficulty. Rating transition probabilities cannot be implied from market prices, so scarce, coarse historical data must be used and there is no natural hedge. A mandatory break is the exception, since a model can treat it as certain.

Rating triggers inside margin agreements

The same flaws apply when ratings drive margin terms. Before the GFC, monoline insurers and other triple-A names traded under one-way terms in which they posted nothing unless downgraded. The downgrade tended to come late, and the sudden duty to post then deepened the credit problem it was meant to address.

AIG is the leading case. Its subsidiary AIGFP had written large volumes of credit default swaps on terms that tied its margin to AIG’s rating. In September 2008, downgrades of AIG’s bonds produced calls for $20bn of extra margin and left the group essentially insolvent. Because AIG mattered to the whole system, the Federal Reserve Bank extended up to $85bn of secured credit so that AIGFP could pay what it owed and the group would survive.

So rating triggers can fail as protection because the rating moves too late, and they can push the downgraded party into a liquidity spiral. Regulation reflects this asymmetry: capital rules give no benefit for triggers and assume default can come without a prior downgrade, while liquidity rules make a bank pre-fund what its own downgrade would require it to post.

A reset agreement periodically readjusts the parameters that drive a trade’s value, so that neither party drifts deep into the money. Resets may fall on payment dates or be triggered when the value crosses a level.

In a cross-currency swap, most of the MTM comes from FX moves acting on the final exchange of notionals, particularly where both legs pay floating rates. A resettable version pays that MTM across in cash on each reset date and resets the FX rate, usually to spot, which changes the notional on one leg. That is equivalent to closing the swap at MTM and replacing it at market terms, so exposure returns to zero at each reset, typically every quarter.

Resets compared with breaks and with margin

Because resets happen on fixed contractual dates whatever the market or the counterparty does, they are much simpler to value than break clauses: the model just applies the reset on each date. A reset is a weak form of margining, since margin usually moves daily and resets only a few times a year, and it is a true settlement, whereas OTC margin is held as security. In the same spirit, a break clause resembles a single margin payment followed by termination, and a reset an occasional settlement that clears the exposure built up so far. Margining is the most dynamic and general of the three.

Variation margin follows the value of the underlying trades, which may be positive or negative to either side. It is computed from the base valuation, the value without xVA. That is not the true economic value, but both parties can agree on it daily because it is roughly symmetric and easy to compute, and including xVA would be circular since margin itself changes xVA.

In an actual default, variation margin usually falls short for two reasons: the exposure can grow during the delay between the defaulter’s last posting and the declaration of default, and replacing or rehedging the defaulted trades costs money. Together these define the margin period of risk (MPoR), covered later.

Initial margin is extra margin taken to cover that shortfall. ISDA documents historically called it the independent amount, but the exchange and clearing term is now standard. Variation margin tracks the portfolio’s current value, while initial margin is a cushion sized for what could happen during the delay and close-out after a default.

Surveys find that cash accounts for about three-quarters of variation margin and slightly more than half of initial margin, with government bonds next. Bilateral markets historically relied almost entirely on variation margin, whereas initial margin is routine at exchanges and central counterparties (CCPs); the bilateral margin rules are now spreading it.

Funding asymmetry

Margin is also a funding matter. A bank that pays fixed to a corporate on an uncollateralised swap and hedges by receiving fixed with another bank under a CSA must post on the hedge whenever the client trade moves in its favour, yet receives nothing from the client. Funding that one-sided flow is the subject of funding value adjustment (FVA), and it explains why some end users have been unable or unwilling to post.

OTC margin usually moves directly between the parties rather than through a custodian, although regulation is changing that for initial margin. Legally, the transfer takes one of two forms.

• Security interest. Ownership stays with the giver; the receiver gains an interest it can enforce only on defined events such as default. This is typical under New York law.
• Title transfer. Ownership passes to the receiver, perhaps with limits on use, so the receiver can generally use the assets freely and its rights are stronger. English law agreements usually work this way.

Survey data show the New York pledge as the most common form, at 46.8% of non-cleared margin agreements, with the English title transfer at 30.1%. Title transfer favours the receiver, who holds the assets and faces less legal risk. Security interest favours the giver, who keeps ownership and is better protected if the receiver fails while holding more than it is owed, since under title transfer that excess may vanish into the receiver’s bankruptcy estate.

Income and interest

Coupons, dividends and other income on posted securities must be passed back, unless doing so would immediately trigger a margin call, in which case the holder may keep enough to stay properly collateralised. Cash margin earns interest at a rate set in the agreement, typically an overnight rate (OIS) such as EONIA in Europe or Fed Funds in the US; exchanges pay nothing because their margin is settlement. Some receivers, usually sovereigns or institutional investors, pay below OIS to discourage cash, which they must reinvest or redeposit, and a few pay above OIS to attract it. An overnight rate fits margin that may be held only briefly, but on long-dated trades large amounts can stay posted for years, and a party borrowing at well above OIS to fund them suffers negative carry, one source of FVA. The value of choices embedded in the agreement, such as which currency or security to deliver, is captured by a collateral funding adjustment (ColVA).

Substituting collateral

A giver may want securities back to deliver under another agreement or a repo, to hold them over a coupon or vote, or to post whatever is cheapest to deliver. It then requests a substitution: it sends a notice, the receiver consents where consent is required, the giver delivers new eligible assets after haircut, and the receiver returns the originals (or equivalent securities of the same issue) by the settlement date. Without a consent requirement the request can be refused only if the new asset is ineligible, although the old one need not be released until the new one arrives. Free substitution affects the funding value of the cheapest-to-deliver option, and the short overlap in which the receiver holds both sets of assets is a brief exposure similar to settlement risk.

Rehypothecation is the reuse of margin by its receiver, for example by posting it to another counterparty or lending it in a repo. Segregation ring-fences margin so it is not reused. Which one suits depends mostly on whether the margin is variation or initial margin.

Why reuse suits variation margin

Dealers hedge. If dealer X trades with client A and hedges with dealer B, both under margin agreements, the two trades have equal and opposite values, so margin A posts to X is exactly what X owes B, and passing it straight on is natural. Had the trades been settled, the cash would simply have netted, so variation margin behaves as quasi-settlement. Reuse lets margin flow through a market full of offsetting hedges and reduces demand for high-quality collateral, which lowers funding costs. Cash and title-transfer margin are reusable by nature; other margin needs an explicit right of rehypothecation, and such rights are widely used.

The counterparty risk cost is small for variation margin. It answers a negative valuation, so on default it can be set off against what is owed. Two minor exposures remain: margin due back after a value change, which daily exchange keeps small, and the slight excess delivered because of haircuts.

Why initial margin should be segregated

Initial margin is overcollateralisation, an amount not owed, so reusing it leaves the giver exposed to the receiver for its full value. Lehman Brothers and MF Global, where reused assets and co-mingled variation and initial margin were not returned, made this clear, and rehypothecation fell sharply after the crisis, mostly of independent amounts; hedge funds, large posters of initial margin, became reluctant to permit it. Segregation protects margin if the receiver becomes insolvent, either through legal rules that give priority to returning unneeded margin or by placing it with a third-party custodian. It is incompatible with rehypothecation and raises funding costs.

Segregated margin can sit with the receiver itself, with an agent acting for one party, or with a tri-party custodian contracted to both sides. All three give legal segregation, but only the last two give operational segregation. MF Global customers had legal without operational segregation and lost margin through fraud. Cash is fungible and hard to ring-fence on a receiver’s books, which favours a dedicated tri-party account, at the cost of limited investment returns and concentration risk with a few custodians.

Segregation is safest when the margin is pledged rather than transferred by title (title transfer leaves the giver an unsecured creditor), when the custodian is unaffiliated with the counterparty, and when reuse is forbidden. The bilateral rules follow this logic, requiring segregation of initial margin but not of variation margin, and the funding cost of replacing margin that could otherwise be reused is the heart of the trade-off between counterparty risk and funding.

Exchange variation margin is settlement: cash in the contract currency with no interest. Bilateral OTC margin traditionally is not, as the interest paid to the giver and the freedom to post bespoke securities or currencies show. Cleared OTC derivatives sit between.

Cleared and collateralised bilateral derivatives are both collateralised to market rather than settled. Bilaterally the cause is the bespoke terms, though these are becoming more standard. A CCP margins like an exchange but pays price alignment interest (PAI), designed to reproduce the economics of a bilateral contract.

Because CCP variation margin is cash in the derivative’s currency, it is already settlement in substance, and European regulators have stated that such cash should carry no haircut because it simply settles a claim. So some cleared contracts have been redocumented as settle-to-market (STM), so each payment legally settles the exposure. That gives a firmer legal position on default and lower regulatory capital, including leverage ratio costs, because for regulatory purposes the trade’s maturity collapses to a single day or the shortest bucket available. Interest is still paid to preserve equivalence with bilateral contracts.

The settlement view also explains why bilateral margin is not netted against trade cash flows, even in the same currency. A large coupon changes the trade’s value at once while the margin adjustment arrives later, so exposure briefly jumps, a collateral spike. On an exchange, settlement nets the two automatically.

The valuation agent is normally whichever party calls for margin or its return, and it carries out the calculations. A large dealer facing a small counterparty may be the agent for everything; often both parties act as agent and each calls whenever its value has become less negative, which makes disagreement likely.

The agent must establish the portfolio’s current value after netting, revalue the margin already held and apply its haircuts, work out the resulting uncollateralised exposure, and finally set the credit support amount, the margin either party must now transfer.

Where disputes come from

• Trade population: which trades belong in the portfolio.
• Valuation methodology for the trades.
• Application of CSA terms, such as thresholds or eligible collateral.
• Market data used, including the market close time.
• The value placed on collateral already posted.

Cleared trades avoid these disputes because the CCP is the sole valuation agent, although it must keep its methods and requirements market standard, transparent and robust.

Reconciliation

Bilaterally, reconciliation heads off disputes by agreeing trade populations, values and margin requirements in advance. It is becoming more frequent, and surveys show most portfolios, above all large ones, are now reconciled every day. In the US, Dodd-Frank, and in Europe, the European Market Infrastructure Regulation (EMIR), both demand tighter and more regular reconciliation, and unresolved disputes can raise capital requirements. Electronic messaging is replacing manual steps, and third-party firms offer valuation, margin transfer, dispute management and reconciliation services to cut operational risk.

No OTC derivative obliges either party to post margin on its own. The obligation comes from a credit support annex (CSA) attached to the ISDA Master Agreement, now the standard bilateral margin document; 87% of margin agreements in use are ISDA agreements, and in one survey of 41 member firms, mostly banks and broker-dealers, 34% held more than 3,000 margin agreements. ISDA keeps legal opinions in many jurisdictions on the enforceability of CSA terms.

A CSA usually covers the same trades as the Master Agreement, and their net value sets the margin. The exception arises when parties put a new CSA in place for future trades and leave older trades under the original, so two CSAs can sit under one Master Agreement, which is how most banks implemented the bilateral margin rules.

What the CSA specifies

The terms fixing the credit support amount cover valuation method and timing; how margin is calculated; transfer mechanics and timing; eligible cash currencies and securities; substitution; dispute resolution; interest on margin; any initial margin; haircuts; whether reuse is allowed; and triggers, such as downgrades, that tighten the terms. These are fixed at signing and change only by mutual amendment, whereas a CCP can alter its terms on its own, as several have done in illiquid or volatile markets. Once live, the trades are revalued, usually daily, and margin recalculated, with procedures for disputes and periodic reconciliation. The parameter choices largely trade operational effort against protection, with funding effects now part of the balance.

Why some entities will not post

Many corporates cannot carry the operational and liquidity burden of posting. Highly rated sovereigns, supranationals and multilateral development banks often receive margin but refuse to post, relying on their generally triple-A standing. Such entities typically prefer, or are forced, to pay the bank’s xVA charges for counterparty risk, funding and capital (KVA) instead. Other motives include internal or external constraints (monoline insurers kept triple-A ratings only by not posting) and the view that uncollateralised trading is cheaper once liquidity costs are counted.

Three broad arrangements appear in practice, though none is a contractual label; each results from how the parameters are set.

• No CSA. One side is unable or unwilling to sign up to posting. The usual case is a bank and a corporate whose treasury would find the liquidity demands of a CSA hard to manage, though some large corporates do post.
• Two-way CSA. Both sides post. This is normal between financial institutions, and in the interbank market such agreements typically run with thresholds of zero, to the counterparty risk benefit of both.
• One-way CSA. Only one side receives margin. The giver is then worse off than with no CSA, because it carries the receiver’s risk on what it posts and gets nothing back. Typical receivers are triple-A sovereigns and supranationals facing banks, and banks have imposed one-way terms on some hedge funds.

A one-way CSA is really a CSA with an infinite threshold for the non-posting party, often only while its rating stays strong. Funding and capital costs have made these agreements especially troublesome for banks, and a sovereign under a one-way CSA is, from the bank’s side, uncollateralised.

Source: ISDA (2015). Rows sorted by share; the last column is derived as 100% less the share under a CSA.

Banks nearly always trade under CSAs, other financial firms usually do, and non-financial firms usually do not, which is the corporate liquidity problem in numbers.

Four CSA parameters decide how closely margin tracks exposure.

The margin call frequency is the contractual interval between calls. Longer intervals save effort and allow time for valuations, and some small firms find the daily calls of large dealers hard to meet; less frequent calls can suit calm markets, but daily calls are now standard in most OTC markets and intraday calls are common for repos and cleared derivatives. The call frequency is just one ingredient of the MPoR: a CSA may call daily while a model assumes an MPoR of 10 days.

The threshold is the exposure below which nothing can be called, so it deliberately leaves the portfolio undercollateralised; above it, only the excess can be called. It sacrifices protection to save operational and liquidity cost, and a lender may accept unsecured exposure up to the credit limit it would grant anyway. A zero threshold means margin is always posted (subject to the MTA); an infinite one means never, as in a one-way CSA. Thresholds are increasingly zero and may be positive only for parties outside the bilateral margin rules.

The minimum transfer amount (MTA) is the smallest transfer allowed, avoiding repeated movement of trivial amounts of possibly non-cash margin. The MTA and threshold add up in one sense, because the exposure must exceed their sum before a call is possible, but the MTA cannot be folded into the threshold, since once a call is possible the full excess over the threshold is due. Transfers may also be rounded to convenient sizes, a minor effect. MTAs and rounding matter mainly for non-cash margin and are generally zero for cash-only arrangements such as regulatory variation margin and CCPs.

Initial margin as a negative threshold

Initial margin is posted regardless of portfolio value, giving overcollateralisation as a cushion against delays and close-out costs, and it depends on how much the value could move, not on the value itself. Bilateral independent amounts were historically rare and one-directional, usually from the weaker credit to the stronger. Exchanges and CCPs always require initial margin, and the bilateral rules now demand it whatever the credit quality, unless a party or product is exempt. It is no longer static: it changes with the portfolio’s perceived risk, often through complex dynamic models. Because it acts as a negative threshold, the two are not combined: either a threshold with zero initial margin, or initial margin with a zero threshold.

Bilateral CSAs have often tied thresholds, initial margin and sometimes MTAs to credit quality, usually the rating, occasionally other measures such as traded credit spreads. The aim is a light workload while the counterparty is strong and tighter terms as it weakens. The schedule below is typical and could apply to one or both parties.

The notch column is derived.

At AAA the party never posts; each notch lower shrinks the threshold until it is zero at A+, and below A+ initial margin is added. Two parties under such a schedule post symmetrically only if they share a rating, which is how a one-way CSA often arises. The weaknesses of rating triggers apply in full: the downgrade comes late and the margin it releases lands when the downgraded party can least afford it. Capital rules give no credit for receiving such margin, while liquidity rules make a bank pre-fund what its own downgrade would require.

The CSA lists the assets that may be posted. Cash, mainly dollars and euros, dominates: one survey found it was 74.9% of margin received against non-cleared trades, with government securities at 14.8% and the other 10.3% spread across agency, supranational, municipal, covered and corporate bonds, equities and letters of credit. Posting non-cash assets helps liquidity, but cash keeps gaining share, helped by cash-only variation margin at CCPs.

A haircut discounts the credited value of an asset whose price can move, leaving a cushion if it must be sold lower. A haircut of x% means each unit posted counts only (1 – x)%, the valuation percentage, so the giver must post more. Cash in the termination currency (the one used to close out the portfolio) usually has none, though other currencies may take an FX haircut. Bonds move with rates and spreads and need a few percent; equities and gold need far more. Volatile assets are acceptable with a large enough haircut, but assets with heavy credit or liquidity risk are generally excluded, since no practical haircut covers a default or a forced sale at a deep discount.

Bilateral haircuts are fixed percentages that change only by agreement, whereas a CCP can revise them. Foreign-currency securities may take extra haircuts, and hedging that FX risk must be dynamic. A posted security downgraded below the agreed level must usually be replaced at once. Haircuts should reflect liquidation time, the volatility of the value drivers, default risk, maturity, liquidity, and any link between the asset and the counterparty or the exposure, which is wrong-way risk. That link is the hardest to legislate: a CSA accepting sovereign bonds may still let a bank post its own sovereign’s bonds.

Deriving a haircut

A natural rule is to cover a severe fall in value over the few days needed to sell. With normal returns:

With a short liquidation period even fairly volatile collateral adds little exposure once haircut. The real danger is wrong-way collateral, such as a sovereign posting its own bonds, or a bank receiving euros from European sovereigns or banks. Even then, more margin can be called as the sovereign weakens, and a sudden default still leaves the bonds’ recovery value plus an unsecured claim.

In ISDA documents the credit support amount is how much margin can be demanded at a particular moment; thresholds, MTAs and initial margin make it differ from the portfolio value. For a party that only receives margin, allowing for the threshold and initial margin alone, the margin it should hold is:

The threshold is usually fixed while initial margin can vary, and the expression shows again that they pull in opposite directions, so with initial margin the threshold is normally zero. For a two-way agreement, the variation margin amount is:

A positive result lets the calculating party call margin or ask for its own back; a negative one means it must post or return margin. Either way a transfer happens only above the MTA, possibly rounded. Initial margin depends on none of these terms and is handled apart: it is not netted against variation margin, may be segregated, and when both sides post it the two amounts are paid gross.

With an MTA, the margin held depends on the route the value took, not only on where it ends. Take an MTA of 50, a zero threshold and no initial margin, and three scenarios that share the same values at t2 (110) and t3 (150) but differ at t1. Margin moves only when the required change, value less balance, is at least 50.

The call at t3 depends on the balance at t2, which depends on t1. Such path dependency makes collateralised exposure harder to model than a function of today’s value alone.

Dynamic initial margin with a zero threshold

At t1 the receiver holds 350 of variation margin and 50 of initial margin against a value of 350, so it is overcollateralised. At t2 the value has risen to 375, leaving a variation margin shortfall of 375 – 350 = 25, which is called; until it arrives the shortfall sits inside the initial margin cushion, though whether close-out costs would also fit is another matter. Initial margin has fallen to 45 because the measured risk fell, not because the value changed, and that drop of 5 is not netted against the 25 call.

Margin falls short of perfect protection for two reasons. A threshold leaves part of the exposure permanently uncovered, although thresholds are increasingly zero. And margin arrives with a lag and in steps set by the call frequency and MTA, so some movement is always uncovered between calls.

Initial margin works as a negative threshold and can, in principle, cover that lag in advance and reduce the exposure to zero. Margin use has grown strongly over the past decade, and comparing total margin with gross credit exposure gives a rough share of exposure that is collateralised, although the aggregate mixes uncollateralised, collateralised and overcollateralised portfolios. The bilateral rules will push it higher.

Bilateral vs centrally cleared margin practices

Bilateral margining is bespoke, sometimes one-way, and full of delivery options across currency, asset class and maturity. Clearing grew from exchange practice: cash variation margin in the trade currency plus initial margin keeps the CCP free of FX and liquidity risk and limits its counterparty exposure.

The ISDA standard credit support annex (SCSA) tried to bring bilateral practice closer to clearing, removing embedded options and encouraging standard OIS discounting. Where a typical CSA nets a multi-currency portfolio into one amount, settles in several currencies and securities and uses non-zero thresholds and MTAs, the SCSA allowed only cash for variation margin (initial margin could still be securities), restricted currencies to the five whose OIS curves are deepest (USD, EUR, GBP, CHF and JPY), set thresholds and MTAs to zero, and computed a separate requirement for each currency, putting cross-currency products in the USD bucket. That currency silo problem, a cash call in every currency, kept it unpopular, and regulatory margin rules have largely replaced it.

Regulators now require most large participants trading non-cleared derivatives with each other to exchange margin under the uncleared margin requirements (UMR), set out by BCBS-IOSCO. The rules aim to reduce systemic risk, by making margin available when a counterparty defaults and so limiting contagion, and to encourage central clearing, by bringing bilateral costs, especially initial margin, into line with cleared costs. Variation margin is meant to cover current exposure and initial margin potential future exposure.

The rules bind covered entities trading with each other; if one side is not covered, the other need not comply in that relationship. Exempt everywhere are the Bank for International Settlements, central banks, sovereigns and multilateral development banks, with other exemptions differing by region.

The 10-day horizon sits on top of daily variation margin and loosely matches the MPoR. Reuse of initial margin is permitted only in a narrow case: the trade hedges a client position, the client agrees and has a priority claim in insolvency, and reuse happens once. Because one client trade sets off a chain of interbank hedges, this exception is of little practical use.

Delivery and segregation

Margin is generally due the same business day as the calculation, with limited cases allowing two days for variation margin. The collector must protect initial margin from its own failure, through a third-party custodian or another legally binding arrangement, so that it is promptly available to the poster; it must offer the poster separation from other clients’ assets; and initial margin must be given by security interest, not title transfer. Variation margin may still be reused. The logic is that posting extra margin should not create extra counterparty risk, and it forces firms to open custody accounts and build connectivity.

Third-party custody is the strongest protection, though the few custodians could concentrate systemic risk, and risk-sensitive models make firm dispute procedures essential.

The rules cover all OTC derivatives except FX swaps and forwards. The exemption, like its clearing counterpart, is contested: these products are often short and carry more settlement than counterparty risk, but FX can be volatile and tied to sovereign risk, and cross-currency swaps tend to be long. In the EU, physically settled FX forwards still carry variation margin. Scope depends on a group’s average aggregate notional of non-cleared derivatives at the month ends of June, July and August, including physically settled FX forwards and swaps.

Each phase runs to 31 August of the next year, with Phase 6 open-ended; Phase 5 was revised and Phase 6 added after the original rules, and local start dates sometimes slipped (the EU began on 4 February 2017). Initial margin applies only to trades executed after the relevant date, and a party may agree an initial margin threshold of up to €50m, shared across its consolidated group, posting only the excess. A firm below €8bn of notional, or under €50m with each counterparty, can stay outside the initial margin rules. Cross-currency swaps are awkward: the FX component that drives most of their exposure is exempt from initial margin but not from variation margin.

The standardised schedule

Initial margin may come from a validated portfolio model, in-house or third-party, or from a standardised schedule, with no cherry-picking between them, although different asset classes may use different methods consistently. Under the schedule, each trade’s notional is multiplied by a rate and summed into gross initial margin.

Portfolio effects enter through NGR, net replacement cost after enforceable netting divided by gross replacement cost:

The schedule is conservative and handles diversification badly, and trades without clear notionals (variance swaps) or maturities (callable structures) need cautious assumptions. Even models may not offset currency/rates, equity, credit and commodities, which are added. Because each pair of firms must agree numbers, ISDA built a common model, the ISDA SIMM (Standard Initial Margin Model).

Regulatory initial margin should be highly liquid and keep its post-haircut value under stress. Haircuts must be risk sensitive, and margin should avoid excessive credit, market or FX risk, including a mismatch between collateral and settlement currencies. It must not be wrong-way, so a counterparty cannot post its own bonds or shares. Acceptable examples are gold, equity in major indices, high-quality corporate and covered bonds, high-quality government and central bank securities, and cash. US rules limit variation margin to cash in dollars, another major currency or the settlement currency, except that financial end users may use initial margin assets.

Haircuts may also come from an approved model. The FX add-on has a precise scope: cash variation margin takes none even in another currency, while non-cash variation margin outside the settlement currency, and initial margin of any kind in a different currency, take 8%. Haircuts should reflect market, liquidity and credit risk in normal and stressed markets and avoid procyclicality, meaning sudden increases in a crisis. Their horizon is not specified but could be shorter than for initial margin, perhaps 2 to 3 days, because collateral alone can be sold faster than a portfolio can be closed out.

Implementation and early impact

Because initial margin applies to new trades once both parties are in scope, the rules phase in gradually, though parties may agree to cover all trades. ISDA’s Variation Margin Protocol let firms document compliance with many counterparties at once, by amending existing CSAs so they cover old and new trades, by copying the CSA and amending the copy for new trades only, or by signing a new standard CSA where none existed. New or amended CSAs must settle thresholds and MTAs, eligibility, haircuts, calculations, timings and deliveries, dispute resolution, and initial margin calculation and segregation.

A survey of 18 of the 20 Phase 1 firms, with an estimate for the other two, found about $50bn of regulatory initial margin posted, and a similar amount received. By 31 March 2017 the same firms had also handed over roughly $16bn of discretionary initial margin while taking in roughly $61bn, the gap reflecting smaller counterparties posting before falling in scope. Most variation margin was cash, and government bonds led for initial margin. Firms entering scope must understand ISDA SIMM, the cost of initial margin through margin value adjustment (MVA), and the capital relief from receiving it.

Margin is usually judged by how much it reduces exposure to a failed counterparty, but a wider view shows it shifts losses: derivative creditors are paid more in default at the expense of everyone else. Like netting, margin redistributes risk rather than removing it, and it adds residual risks of its own: operational, legal and liquidity risk, wrong-way risk, credit risk on margin securities, and FX risk on margin in another currency.

What A gains, the other creditors lose. They become more exposed, shifting risk into markets such as lending, and they react to their loss of effective seniority, for example by charging more to lend. The redistribution may be beneficial, but total risk has not fallen.

Residual risk comes partly from contract terms, thresholds and MTAs, which are increasingly zero or tiny, and from the call frequency, where daily calls now dominate. Even then there is an unavoidable lag in receiving margin, and market moves after the counterparty’s last successful posting are market risk. The margin period of risk (MPoR) is the effective time from that last posting until all its trades are closed out and replaced or hedged; any rise in exposure in that window, close-out costs included, is uncollateralised. It is purely a counterparty risk concept and plays no part in assessing funding costs.

Before default the clock runs through the valuation and call (computing the MTM and margin held, confirming and making a valid call, including the wait for the next call date), receiving collateral (slower by fax or email than by electronic messaging, and longer if disputed), settlement (cash may settle intraday, while government and corporate bonds can take one and three days in some markets), and a grace period, or cure period, before a missed call becomes a default. After default it runs through macro-hedging of key sensitivities, close-out into a single MTM claim, rehedging or replacement (one-to-one replacement, which end users need to avoid market risk and accounting mismatches, or a cheaper portfolio hedge preferred by banks), and liquidation of collateral, whose cost belongs in haircuts.

How long the MPoR is assumed to be

Margin normally arrives quickly, but that is irrelevant: it protects against counterparty risk only after a default, when conditions may be far from normal. So Basel II set a minimum MPoR of 10 business days for OTC derivatives with daily calls (any extra days between calls are added), Basel III uses a stricter 20 days or more in some cases, and CCPs assume around five business days. The ISDA Resolution Stay Protocol, signed by the 18 major global banks, can suspend default rights for 24 or 48 hours, which argues for a longer MPoR.

Models usually take one step with 100% of the risk throughout, though real risk declines after default, arguing for a shorter modelled MPoR, while volatility may jump after a default, arguing for a longer one. The MPoR is a catch-all estimate of the effective period of market risk rather than a literal timetable, and initial margin horizons are set consistently with it.

Real results are worse than that estimate, so only initial margin can bring counterparty risk to negligible levels, and the MPoR is the main driver of how much is needed.

Holding margin creates liquidity risk if it must be sold after a default: the survivor pays bid-offer costs and bears price moves while turning securities into the cash needed to rehedge. Haircuts buffer prices falling before the sale, but selling a holding that is large relative to traded volume can push the price down and cause a loss well beyond the haircut. Wrong-way risk arises if the collateral’s value is tied to the counterparty’s default, as when an entity posts its own sovereign’s bonds. Before accepting collateral, it pays to ask how big the issue or market capitalisation is relative to what is posted, whether its value is linked to the counterparty’s credit (perhaps only visible in correlations; LTCM’s huge holding of Russian government bonds made them poor collateral), and how its liquidity would behave after the counterparty failed. A concentration limit of 5 to 10% may be set to curb liquidation risk.

Collateral denominated in a currency different from that of the portfolio creates FX risk, and even euro cash posted by a European bank creates a worrying link and wrong-way risk. Unsurprisingly, margin is increasingly cash in major currencies or low-risk securities: among major institutions one survey found 70.9% cash, with government securities making up 20.7% and other securities 8.3% (99.9% in total because of rounding). Smaller users still often post less liquid assets under bespoke terms.

Operational and legal risk

Collateral management is intensive and dynamic, so operational risk arises from missed calls, failed deliveries and human error, especially at large banks running thousands of non-standard agreements and moving billions of dollars of margin a day. Basel III recognises this by requiring a longer MPoR where a netting set exceeds 5,000 transactions, holds illiquid collateral or exotic trades that are hard to value under stress, or has a history of disputes. A longer MPoR means more capital, which rewards better operations.

Legal risk arises if the right to net margin against the portfolio, or the portfolio’s valuation, is challenged, as in the dispute between Citigroup and Lehman Brothers. The holder must be confident that a defaulted counterparty’s administrator cannot reclaim the margin, and rehypothecation and segregation each carry legal risk of their own.

MF Global, a major derivatives broker, failed in October 2011. Segregation should have kept customer margin safe, but before the bankruptcy the firm had illegally moved $1.6bn of supposedly segregated customer money out to third parties to cover overdrafts and its own margin calls. By June 2013, 89% had been found and returned. Private lawsuits followed, along with a civil action by the US Commodity Futures Trading Commission against the former chief executive. A firm fighting insolvency may take extreme, even illegal, steps, so segregation rules must be clear and enforceable.

Margin also provides funding. A bank owed money on an uncollateralised trade is not paid until settlement, yet may have to post on the hedge in the meantime, which is why reuse rights matter. Counterparty protection matters only in default, while funding matters in every scenario, and two extremes show the difference:

• Reusable wrong-way margin. A counterparty posting its own bonds offers little protection against its own default, yet if the bonds can be reused they still fund the receiver.
• Good but non-reusable margin. Corporate bonds that cannot be repoed or reposted protect against default but leave the funding need in place.

The Totem xVA consensus service uses this split by asking banks to price trades in which the counterparty’s margin goes into a ring-fenced account; since Totem excludes capital costs, the price isolates funding. Margin management has become an optimisation exercise, choosing the cheapest-to-deliver cash or securities after haircuts, the OIS rate paid on each currency, and the repo rate earned by reusing securities.

From counterparty risk to funding liquidity risk

Here funding liquidity risk is the risk of struggling to raise funding later, especially when margin is segregated or cannot be reused. Margin agreements create it by demanding payments at short notice in amounts tied to future values. Moving from no CSA, to a two-way CSA, to bilateral initial margin under UMR, to central clearing with default funds and intraday calls, steadily lowers counterparty risk and steadily raises funding liquidity risk, a conversion that is cheap in calm markets and can be crippling when liquidity vanishes.

End users feel this most. Their hedges are directional, such as paying fixed against floating-rate debt, so a big rate move brings big margin calls with nothing offsetting them, and many hold little cash; pension funds hold bonds but limited cash. Ashanti, a Ghanaian gold miner now part of AngloGold Ashanti, lost $450m on gold forwards and options when gold rose in September 1999, and its counterparties were owed about $280m of cash variation margin. It had gold to deliver but no cash, and escaped only by negotiating freedom from posting for just over three years.

An end user unable to meet a call on time in eligible assets could default while solvent; its bank may then waive the call, turning the risk back into uncollateralised exposure and a funding need, and agencies may view agreeing to post negatively (monolines kept triple-A ratings only by not posting). Banks run hedged books, so their strain comes from hedging uncollateralised client trades under CSAs: they fund the hedge margin when client trades gain and benefit in reverse, which is one way to explain FVA and why banks want more clients on CSAs. Contingent needs from extra initial margin or rating-linked margin must be pre-funded under the LCR, so LCR costs belong in FVA and margin value adjustment (MVA). Pushing towards clearing minimises counterparty risk but builds opaque funding liquidity risk, and lower CVA and KVA have been matched by funding terms such as MVA.

Gathered in one place, the regulatory rules that touch margined derivatives show how capital and liquidity requirements shape incentives.

Capital rules are deliberately sceptical of protection that may fail when it is most needed, which is why rating triggers earn nothing and why MPoR floors far exceed the time margin normally takes to arrive. Liquidity rules work the other way, treating a bank’s own triggers as real outflows to fund now. The margin rules complete the picture by pushing bilateral trading towards clearing economics: lower counterparty risk, CVA and capital on one side, and higher funding liquidity risk, FVA and MVA on the other.