FMP 5: Exchanges and OTC Markets
An exchange is an organisation with members who trade with each other, and for a long stretch of history that was nearly the whole of it. The bargains struck on the early exchanges were relatives of the futures contracts quoted today. What the institution supplied was a meeting place, a set of standard contracts so that every member knew what was on offer, a way of resolving arguments, and the power to throw out anybody who reneged on an agreed deal. Protection against a member who simply failed to perform was not on the list.
Members supplied that protection themselves, and margin came first. Margin means assets handed from one trader to the other as cover against counterparty default, and the quantity handed over follows how far the market has travelled since the deal was struck.
A corn trade shows the mechanism
Trader A takes the buying side of 10,000 bushels of corn for September delivery, Trader B takes the selling side, and the agreed price is 400 cents per bushel. Corn then cheapens by 5 cents shortly afterwards. With a margin agreement in force, Trader B can call USD 500 from Trader A, being 10,000 multiplied by 5 cents. Another 10 cents of decline brings a second call, this time for USD 1,000, and the sequence repeats while the market runs against Trader A. Movement the other way reverses the flow, so that a rise of 20 cents leaves a cumulative USD 2,000 travelling from Trader B to Trader A.
What the arrangement buys is as much psychological as financial. A trader watching the market run a long way against a position has an obvious motive to disown it and deal at whatever better price is available. Margin removes the motive, since the loss has already been paid across in cash instead of waiting until delivery to crystallise. Measurement matters too: margin on a September obligation is worked out from the September price rather than the spot price of corn, because the September price fixes the value of the promise.
Under netting, a long and a short held in the same contract cancel one another, leaving only the difference between them to be settled. It rested entirely on earlier housekeeping, because obligations cancel only when they describe identical things. Standardising contracts by size, by the quality of the commodity and by delivery date is what made netting workable at all.
Two traders, two contracts, one payment
Trader A is committed to buying 10,000 bushels of September corn at 400 cents per bushel, with Trader B on the other side. Later the pair deal again in the opposite direction, Trader A selling the same quantity for the same delivery month at 380 cents per bushel. Nothing now requires corn to move, and the two obligations collapse into USD 2,000 passing from Trader A to Trader B in September, or into the present value of USD 2,000 paid when the second deal is struck.
Longer chains
Add a third party and the logic holds. Suppose the first contract stands as above, and Trader B separately buys 10,000 bushels for the same delivery date from Trader C at 400 cents per bushel. Trader B is flat, having bought and sold identical quantities at identical prices, so the middle link can be removed and one contract survives in which Trader C sells 10,000 bushels to Trader A at 400 cents per bushel.
Trader A buys 10,000 bushels of corn from Trader B at 400 cents per bushel. Trader B covers itself by buying the same quantity from Trader C at 380 cents per bushel, same delivery date.
Modern exchanges do far more than organise trading. Each runs a clearing house, known as a central counterparty or CCP, through which every transaction between members is cleared, and that is what converts a promise between two firms into an obligation owed to the exchange.
How a trade moves to the clearing house
One corn contract at the Chicago Mercantile Exchange covers 5,000 bushels. Member A and Member B deal in one such contract for September delivery, priced at 400 cents per bushel and therefore worth USD 20,000. The clearing house then inserts itself into the middle. It buys the 5,000 bushels from Member B at the agreed price and sells the identical quantity on to Member A at the identical price, so a single agreement between two firms has become two agreements, each of them facing the exchange.
Neither member has any reason to investigate the other after that. They may have dealt on the floor or electronically, and they need never learn who was on the far side. Every member ends up facing one counterparty, the same one for everybody.
Exiting a position becomes easy
The second gain concerns exits. Member A holds a long position in one contract of September corn, September corn meaning corn for delivery in that month. To close out, Member A sells one September corn contract to whichever member is willing to take the other side. That trade also lands at the clearing house, so Member A finishes with two positions against the CCP that cancel to nothing.
Before central clearing the exit was awkward. Member A could try to negotiate a settlement with Member B, who might have no interest in unwinding anything, or sell the same quantity to a third member. That second route left Member A watching the creditworthiness of two firms rather than one, so closing a position multiplied the exposures being carried.
Sitting between every pair of members concentrates credit risk rather than abolishing it, so an exchange that establishes a clearing house must manage what it has absorbed. Four instruments do the work together: netting, variation margin with daily settlement, initial margin, and default fund contributions.
Netting at the clearing house
Here netting means combining the long and short positions of a member to arrive at the net exposure the clearing house is running to that member. Member X sells one September corn contract, and both the sale and everything afterwards face the CCP. Member X then buys four September corn contracts, which also face the CCP. The two collapse, leaving Member X net long three September corn contracts, an obligation covering 15,000 bushels for September. Margin is then calculated against three contracts and not against five.
The order in which losses are absorbed
Default fund contributions sit at the back of the structure. Members pay into the fund, and it is reached only when the margin of a failed member runs out. Where initial margin proves too small, the contribution of the failed member itself is consumed next. Anything still outstanding after that reaches the contributions of the surviving members. Only a loss larger than the margin of the defaulter plus every contribution in the fund can travel further, and at that point the equity of the CCP is what remains.
The four tools divide the job neatly. Netting shrinks the exposure, variation margin clears losses as they arise, initial margin covers the gap between the last settlement and the sale of a defaulted position, and the fund catches whatever remains.
Settlement on a futures contract does not wait for maturity. It happens in instalments, one for each day of the life of the contract, which is the feature separating futures most sharply from forwards. Members carrying a loss on the day pay it over to the clearing house and members carrying a gain collect it, those payments being the variation margin. Once a day is the usual rhythm, though highly volatile conditions can push an exchange into calling for it more frequently.
Where the money goes
Suppose the settlement price climbs between two consecutive closes. Cash travels through the CCP out of the accounts of members who are net short and into the accounts of members who are net long, and a falling price sends it the other way. Because every long contract has a short contract facing it, the amount collected and the amount paid out are identical, so on a settlement day the clearing house is a conduit and not a participant.
Trader X is net long three September corn contracts, a commitment covering 15,000 bushels. September corn settles at 400 cents per bushel on Day 1, at 395 cents per bushel on Day 2, and at 405 cents per bushel on Day 3.
Why daily settlement simplifies an exit
A second benefit is easy to miss. Someone unwinding a position need not know when it was opened or at what level, because everything up to the previous close has already been paid or received. Take a long contract exited at 11 a.m. through a short trade at 375 cents per bushel, with the previous close standing at 372 cents per bushel. Only the extra three cents per bushel needs adding to the account of that member. The rest of the history has already been settled away.
Variation margin settles what has already happened. Initial margin is a deposit of funds or marketable securities lodged with the clearing house on top of the daily settlements, held against the possibility that a member stops paying altogether. Long or short makes no difference to the figure demanded per contract, since the price can travel either way.
A member is net long 20 September corn contracts, each covering 5,000 bushels. Variation margin of USD 7,000 falls due and the member does not pay, so the clearing house sells 20 September corn contracts to unwind the position.
What determines the level
Exchanges set these requirements themselves. Two things drive the number: the volatility of the futures price, and how long an unwind of a defaulted position would take. A contract that can travel far while it is being sold needs a bigger cushion. Nothing about the level is permanent, since the exchange keeps the right to reset it whenever market conditions shift.
Interest, haircuts and portfolio effects
No interest attaches to variation margin, because daily settlement makes that money the permanent property of whoever received it. Initial margin still belongs to the member who lodged it, so the clearing house pays interest on it. A member unhappy with the rate may be permitted to lodge securities such as Treasury bills instead of cash, in which case the CCP writes their value down by a set percentage to reach a cash equivalent. That write-down is the haircut, and a more volatile asset attracts a bigger one.
Holding several contracts on the same commodity changes both requirements. A member long one September corn contract and short one December corn contract usually collects variation margin on one leg while paying it on the other, so the settlements partly cancel without anybody arranging it. Exchange rules go further and shrink the combined initial margin below the sum of what the two legs would attract separately.
Everything so far describes accounts between a clearing house and its members, and a retail trader never reaches the CCP directly. Placing a futures trade through a broker obliges the trader to lodge margin with that broker, and brokers outside the membership pass the trade to a member, which creates a second margin account between broker and member. Margin runs along a chain, and every link is collateralised.
Two thresholds instead of one
The retail account differs in carrying a maintenance margin alongside the initial margin. Settlement is daily, exactly as it is for a member, with gains added to the balance and losses taken out. Anything above the initial margin requirement belongs to the trader and can be withdrawn. Once the balance dips under the maintenance level, enough new funds must arrive to lift it back to the initial margin figure rather than merely over the line it crossed, and funds that do not arrive leave the broker to unwind the position with an offsetting trade. The maintenance margin is typically 75% of the initial margin. Exchanges publish minimum levels for both, and a given trade attracts a larger initial margin from a retail trader than from a clearing member.
A trader instructs a broker to sell wheat futures covering 5,000 bushels at 600 cents per bushel. Initial margin is USD 3,000 and maintenance margin is USD 2,000. Settlement prices then follow the path below. Interest on the balance is ignored.
| Day | Settlement (cents) | Gain or loss (USD) | Balance before any call (USD) | Action |
|---|---|---|---|---|
| 0 | 600 | 0 | 3,000 | Position opened |
| 1 | 604 | -200 | 2,800 | None |
| 2 | 610 | -300 | 2,500 | None |
| 3 | 618 | -400 | 2,100 | None |
| 4 | 623 | -250 | 1,850 | Call for 1,150 |
| 5 | 615 | +400 | 3,400 | 400 withdrawable |
Source: margin levels and contract terms follow the chapter. The price path and the resulting balances are constructed for this lesson.
To buy on margin is to fund part of a purchase with money borrowed from a broker. No clearing house appears anywhere in the arrangement. The broker advances the balance of the price, keeps the stock against the loan, and watches the position against two percentages it has set itself. Equity in the position, which is what the margin account balance measures, is the market value of the holding less the debt owed on it.
A retail trader takes 1,000 shares at USD 60 each on margin. The broker works to 50% initial margin and 25% maintenance margin. Interest on the loan is ignored throughout.
One contrast is worth carrying forward. A retail futures trader answering a call must restore the initial margin level, while the stock buyer here climbs back only to the maintenance level of 25%. Where an account is restored to varies by product, and it changes the size of the payment demanded.
Shorting a stock means borrowing shares from somebody who owns them and selling those shares on in the ordinary way. Later they are bought back and returned to the lender. Losses on a short position have no natural ceiling, so the margin arrangement is heavier than for a purchase. In the U.S. a retail short normally carries a requirement of 150% of the stock price, measured when the position is opened. Sale proceeds cover two-thirds of that, since 100% divided by 150% is two-thirds, leaving the trader to find the remaining 50% of the stock price.
From there the account moves with the market. A falling share price lifts the balance, and a rising one erodes it. Maintenance margin is typically 125% of the stock price, and a balance that slips beneath it draws a demand for enough funds to reach the maintenance level again.
Working the numbers
Someone shorts 100 shares at USD 30. Selling raises USD 3,000, which stays with the trader, while the requirement is USD 4,500 at 150%, so the sale proceeds go in together with a further USD 1,500.
Now let the stock climb to USD 35, unwelcome for a position built to profit from a decline. The borrowed shares are worth USD 3,500 and the requirement is 1.25 multiplied by that, or USD 4,375, which the USD 4,500 on deposit still covers. Push the price to USD 40 and the requirement becomes 1.25 multiplied by USD 4,000, or USD 5,000, leaving USD 500 to be found. An unmet call ends the position. Throughout, the balance belongs to the trader, and interest on it is due to the trader.
Setting 125 multiplied by the share price against the USD 4,500 on deposit shows exactly where the trouble starts: the demand appears once the price passes USD 36. Note also where the top-up lands. A short seller climbs back only to the maintenance level, whereas a retail futures trader in the same predicament must climb all the way to the initial margin level.
Option positions divide by whether they can generate an obligation later. Someone net long an exchange-traded stock option owes nothing later, having paid the premium upfront and holding a right that may or may not be exercised, so no margin is asked for. Someone net short is in the opposite position, since exercise would force the writer into a purchase or a sale on unfavourable terms, and that potential liability is collateralised with the clearing house.
The Chicago Board Options Exchange calculation
The Chicago Board Options Exchange recalculates the requirement every day as the larger of two quantities. Both start from 100% of the value of the option. One then adds 20% of the underlying stock price and deducts however far the option sits out-of-the-money. The other adds 10%, taken from the stock price for a written call and from the strike price for a written put. Whichever comes out higher is the margin to be maintained.
Take a trader who writes 100 calls at USD 5 each while the stock changes hands at USD 47. A strike of USD 50 puts those calls USD 3 out-of-the-money, so the first branch gives 5 plus 0.2 multiplied by 47 less 3, and the second gives 5 plus 0.1 multiplied by 47. The larger of the two is 11.4 per option, and the whole written position therefore needs USD 1,140.
Neither input stays still, so the requirement is recomputed as the stock price and the option price move, and a writer may be asked for more funds at any point. Anyone who fails to supply them has the position closed out. Options are usually settled at maturity rather than settled daily, which changes the ownership question: margin lodged against a written option remains the property of the writer, and the clearing house pays interest on a cash balance.
Over-the-counter trading of derivatives is an old practice, and in its traditional form it means two firms dealing directly and writing whatever terms suit them. The population splits into end users and dealers. A dealer supplies whatever an end user needs, sometimes runs derivative positions of its own, and lays the risk off with other dealers, frequently through interdealer brokers. Nothing is standardised by an exchange, which is the whole appeal, because a contract can be shaped around a particular hedging problem instead of around a template. Exchange trading offers the opposite bargain, since fixed terms are what make netting, central clearing and one uniform margin regime possible.
Clearing has traditionally been bilateral. The two firms settle between themselves how the trade will be cleared, which netting arrangements govern it, and whether collateral moves. Clearing houses are not new here, but their use expanded sharply after the 2007-2008 global financial crisis.
Scale is where the two structures part company. Measured by the notional value of the underlying assets, the June 2019 OTC market was more than five times the exchange-traded market, while the transactions themselves summed to under 2% of that notional. Interest rate products dominate, at roughly 82% of the market by underlying and roughly 73% by the value of outstanding derivatives. Swaps make up most of that, an interest rate swap being an agreement to trade a fixed rate against a floating rate on the same notional principal, which never changes hands and is notional for exactly that reason.
| Asset class | Share of total underlying | Underlying assets (USD billions) | Transactions (USD billions) | Transactions as a ratio of underlying |
|---|---|---|---|---|
| Interest Rate | 81.81% | 523,960 | 8,806 | 1.68% |
| Foreign Exchange | 15.40% | 98,651 | 2,229 | 2.26% |
| Credit Default Swaps | 1.31% | 8,418 | 235 | 2.79% |
| Equity | 1.10% | 7,046 | 579 | 8.22% |
| Commodity | 0.33% | 2,114 | 198 | 9.37% |
| Other | 0.04% | 253 | 14 | 5.53% |
| Total | 100.00% | 640,442 | 12,061 | 1.88% |
Source: Bank for International Settlements, June 2019 (www.bis.org). Rows are ranked by size of underlying and the share column is calculated for this lesson.
Two cautions attach to those figures. Underlying value is notional principal on outstanding transactions, which dwarfs what the transactions are worth. And a deal worth X to one party is worth minus X to the other, so only the positive X reaches the market value column.
Credit risk was always the weak point of trading away from an exchange, which answers it with clearing houses and margin. Early OTC trades were cleared bilaterally with very little protection attached: two firms agreed a set of contingent future cash flows and left the matter there, so trouble at one of them, severe enough to stop payment, normally meant a loss for the other.
Size that exposure carefully, though. Put an interest rate swap beside a bond of identical principal and the credit exposure on the swap is far smaller. Where nothing has been collateralised, what a trader stands to lose is the value of the contract when that value is positive, and nothing at all when it is negative.
How much default is expected to cost across a book held with one counterparty turns on how long the derivatives run, because two things stretch with the horizon: the counterparty has more opportunity to fall into difficulty, and the market variables driving value have more room to wander.
What a master agreement does on a default
Netting arrived in the bilateral market early. Two firms sign a master agreement covering every derivative between them, and a default by either side then converts the whole book into a single transaction.
| Transaction | Value to Company A (USD million) | Value to Company B (USD million) | Running net to Company A (USD million) |
|---|---|---|---|
| 1 | +40 | -40 | +40 |
| 2 | -30 | +30 | +10 |
| 3 | +20 | -20 | +30 |
| 4 | -10 | +10 | +20 |
Source: transaction values follow the chapter. The running net column is calculated for this lesson.
Company B fails. Absent a netting clause its liquidators repudiate Transactions 1 and 3, the ones that cost them money, while keeping Transactions 2 and 4 or selling those on, which leaves Company A facing USD 60 million. Netted, the four become one contract standing at minus USD 20 million to Company B, and the exposure of Company A drops to USD 20 million.
Reverse the failure and the clause protects the other side. Should Company A collapse, Transactions 2 and 4 alone would expose Company B to USD 40 million. Netted, that exposure disappears and Company B in fact owes USD 20 million on the closing settlement. Practice is a shade more involved, since the surviving side must replace the transactions it has lost and may value them at bid or at ask, whichever suits it.
The aggregate effect is substantial. Allowing for legally enforceable netting agreements, the Bank for International Settlements put the total exposure of derivatives market participants at about USD 2.7 trillion in June 2019, under 25% of the market value of transactions outstanding.
Collateral has spread widely through OTC derivatives trading. Inside the master agreement sits the credit support annex, or CSA, which fixes how the required collateral is computed and which securities may be delivered. Daily valuation is the norm, with the net figure across the outstanding book driving how much extra has to move. Exchange vocabulary has bled across, so what is posted under a CSA is frequently just called margin.
Companies A and B illustrate the flow. One day the book shifts USD 1 million in favour of Company B, which is the same USD 1 million going against Company A. Under its CSA, Company A may then be obliged to deliver USD 1 million of collateral across. The transfer tracks the change in value, exactly as variation margin does at a clearing house.
Where collateralisation differs from exchange margining
Two features of a CSA soften the comparison. Collateral is often called only once the value standing to one side passes a threshold, so exposure beneath it stays uncovered by design. A minimum transfer amount is usually specified too, which keeps small movements from generating administrative cost and leaves another sliver uncovered between transfers. An exchange tolerates neither gap, settling variation margin in full daily and holding initial margin on top.
The remaining differences follow from who writes the rules. Exchange levels are set centrally and apply to every member alike, while under a CSA the two firms negotiate their own terms, including eligible securities and the haircuts applied to them, so the protection reflects bargaining strength and credit quality. In substance the systems have converged on the same machinery: daily valuation, daily transfer, haircuts on anything that is not cash, and a close out when the assets fail to arrive.
Netting and collateral are not the only devices tried here. Three others are worth knowing, each having failed in an instructive way.
Special purpose vehicles
Special Purpose Vehicles, also called Special Purpose Entities, are built by a parent so that credit risk stays on the far side of a legal wall. A large project can be run inside one without exposing the sponsor to it. Where Company Y sets up Company X as such a vehicle, assets move from Y into X and control over them usually does not, and some jurisdictions forbid Y to own X at all. Bankruptcy at Y should therefore leave X still able to meet its obligations, and the wall works in both directions. A AAA credit rating is normal here, though rating agencies grant it only after picking through the legal arrangements and the operating mechanics. Such structures are used heavily to manufacture derivatives out of pools of mortgages or other loans. The sponsor is not on the hook for the payoffs, and a buyer is exposed to defaults inside the underlying loan portfolio without being exposed to the credit standing of the vehicle.
Derivative product companies
A derivative product company, or DPC, was a well-capitalised subsidiary engineered to earn a AAA rating. Clients dealt with the DPC rather than with its dealer parent, so a dealer whose own rating was poor could still put a AAA counterparty in front of a client. Market risk was designed out almost entirely, since every client trade was matched by an offsetting trade with the parent, which then managed the exposure. Credit risk was not. The subsidiary might be close to riskless, but the parent was not, and a failure there would have reached the clients anyway, typically through a sale of the DPC to somebody else or a close out of every transaction at mid-market prices. Since the credit crisis of 2007-2008 DPCs have all but disappeared, and the spread of collateral had already made it redundant well before then.
Credit default swaps
A credit default swap behaves like insurance. The protection buyer pays a premium at regular intervals to the protection seller, and default by a named entity, which is neither of them, triggers a payment from seller to buyer. Volumes grew quickly between 2000 and 2007 and have fallen away since. Monolines are firms with good ratings whose principal business was writing this protection, and several insurers wrote it alongside their ordinary underwriting. AIG is the famous case, its subsidiary AIG Financial Products having sold protection in enormous size on instruments built out of mortgage portfolios. Monolines failed in numbers through the 2007-2008 crisis. Losses on credit default swaps at AIG were severe enough to require a USD 180 billion bailout from the U.S. government, funds since repaid, and Citigroup and Merrill Lynch, both buyers of monoline protection, lost several billion dollars.
None of these lasted. Risk in the OTC market is increasingly handled the way an exchange handles it: netting, central clearing, margin requirements and default fund contributions.