Counterparty Risk vs Credit Risk: A Known Amount or Not

No. Credit risk is the borrower not paying an amount that is already known: counterparty C1 owes Rs 1,680 crore of funded exposure and anybody can read the figure off a record. Counterparty risk is the other side of a contract failing when the amount owed is not known in advance and moves with the market. C1's netted derivative current exposure is Rs 96 crore today. The estimating is the whole difference.

Everything that separates these two comes down to one thing, and it is worth naming before any of the detail arrives. Can the amount at risk be read, or does it have to be worked out? A loan produces a number that exists before anybody measures it. A contract whose value moves produces a number that only comes into being once somebody has chosen a valuation, a moment and an assumption about how far the thing could travel between now and the failure. The difference is not one of degree. The two amounts are different kinds of object, and every other difference set out below follows from that one without effort.

Most people already run both arrangements without any of the vocabulary, so the household version comes first. A cousin has been lent Rs 40,000/- for a scooter and is repaying it monthly. The balance is written in the back of a notebook and it changes only when a payment lands. If he stops paying, the amount lost is known precisely. Now suppose instead an agreement with a neighbour to swap one household's share of the electricity bill for the other's share of the water tanker for the next twelve months. Whether that arrangement ends up worth something to one side or to the other depends entirely on what the tanker and the power cost over the year, and neither party can say today which way the balance will fall. If the neighbour moves out in month seven, the amount lost is not written anywhere. Somebody has to work it out. The scooter loan is credit risk and the bill swap is counterparty risk, and the difference between them is not the size of the amount but whether the amount already exists.

The worked case throughout is Vindhya Commercial Bank Limited, an invented mid-sized Indian commercial bank, and its ten largest single-name exposures, numbered C1 to C10. The largest of them, counterparty C1, is Nirjhar Industries Limited, an invented steel and alloys maker. Every figure below is in Rs crore, and every one is the bank's own choice rather than a published rule. C1 is the one name in the book that carries both kinds of risk at once, on the same day, inside the same reported figure. Holding the two side by side on that single counterparty shows they are not the same animal wearing two collars.

What exactly is credit risk, and where does its number come from?

Credit risk is the risk that a borrower does not pay an amount it has already been lent or already committed to. The defining feature is not the borrowing and it is not the interest. The amount is settled in advance and sits in a record. The question at any moment is only whether it will be paid, never how much it is.

Vindhya Commercial Bank Limited has lent C1 Rs 1,680 crore. The Rs 1,680 crore is the funded exposure, and it is a fact about a ledger. Somebody in the loan operations team can pull it up, and it was very nearly the same number yesterday. Only three things move it: a disbursement, a repayment or the ordinary amortisationThe predictable reduction of a loan balance over time as scheduled repayments land. Slow, known in advance, and the reason a loan balance is nearly the same figure next month as it is today. of the facility, all of which are scheduled events with dates on them. Estimating is what happens when a number does not already exist, so nobody has to estimate C1's funded exposure.

There is a second known amount beside it. C1 also has an undrawn committed line of Rs 720 crore. The bank has promised to lend it and C1 has not yet taken it. Nobody knows how much of it will be drawn before a failure, so the commitment carries a judgement, and this bank applies its own 50.0 per cent assumption, giving Rs 360 crore. Notice carefully what kind of judgement that is. The judgement is about how a borrower in trouble behaves, and it is settled by looking at what borrowers have historically done. No price enters it. The Rs 720 crore itself is exact; only the fraction is assumed.

So credit risk yields two figures for C1 that a person can read and one assumption sitting on top of the second. Rs 1,680 crore funded, Rs 720 crore committed and undrawn, and the bank's own 50.0 per cent drawdown view producing Rs 360 crore. Added together they give Rs 2,040 crore of exposure at default coming from the lending relationship. Every rupee of that has a document behind it.

What exactly is counterparty risk, and why does its number have to be built?

Counterparty risk is the risk that the other side of a contract fails when the amount owed under that contract is not known in advance and moves with the market. Both halves of that sentence carry weight. Counterparty risk is still a failure to pay. The size of what would go unpaid is itself a moving quantity, and on any given day it is whatever somebody's valuation says it is.

What a swap or a forward actually does is covered separately, so C1's two derivative trades facing Vindhya Commercial Bank Limited are described here only by what they are worth and how large they are. One is worth plus Rs 132 crore to the bank. The other is worth minus Rs 36 crore to the bank. If it closed out today, the bank would be the one paying. Both figures came from a mark to marketValuing a contract at what it is worth now rather than at what it cost. The step that creates a current exposure figure where no figure existed before. exercise, and before that exercise was run there was no number at all.

The two trades sit under one netting agreementA signed document saying that whatever the two sides end up owing each other under the contracts it covers, only the difference is payable., so the bank's claim if C1 fails today is Rs 132 crore less Rs 36 crore, being a netted derivative current exposure of Rs 96 crore. Then there is the part that has not happened yet: between today and a failure, the value of those contracts can travel. The bank sizes that travel by taking its own invented add-on factor of 1.5 per cent on the trades' notionalThe reference amount a contract is written on, used to size how far its value could move. It is not an amount anybody owes. of Rs 1,600 crore, giving a potential future exposure of Rs 24 crore. Rs 96 crore plus Rs 24 crore is a derivative exposure at default of Rs 120 crore.

Counting the judgements gives the difference exactly: the Rs 1,680 crore took one lookup, and the Rs 120 crore took a valuation, a view on whether the netting agreement would hold, a notional and an add-on factor. Change any one of those four and the Rs 120 crore changes. The four judgements are not a criticism of the method but the nature of the object, and a bank that pretends otherwise is the one to be worried about.

Which six questions actually separate the two?

Both sides are now defined, so they can be laid against each other properly. Six criteria do the whole job, and between them they will classify anything anybody presents. The set is worth learning whole. The table reads downwards rather than across: the first criterion is the one that generates the other five.

KC1 does the real work and the other five are its consequences. Because the amount can be read, it comes from a record rather than a model, and that is KC2. Because it is a lending balance, it cannot be negative, and that is KC3. Because it moves only when a payment lands, it is stable between one day and the next, and that is KC4. Because there is nothing to set it against, only security helps, and that is KC5. And because the amount is fixed, the only uncertain thing left is the borrower, and that is KC6. With KC1 settled on any exposure, the rest can be derived at the table.

Can the amount ever be owed in the other direction?

KC3 is the criterion people skip, and it is the one that explains a whole technique. A loan is owed in one direction. C1 owes Vindhya Commercial Bank Limited Rs 1,680 crore and there is no arrangement of events, however strange, in which the bank ends up owing C1 on that facility. The number can fall to zero, but it cannot go through zero.

A contract is different in kind. On today's valuation, C1's second trade is worth minus Rs 36 crore to the bank and the bank is the one that would pay. The obligation is real, it sits on the bank's side of the relationship, and a document C1 signed created it. A derivative position is therefore a bilateral exposureAn exposure that can be owed in either direction depending on how prices have moved, so that either side of the contract can be the one at risk.. Either party can be the one at risk, and prices rather than lending decide which one.

The chance of owing in either direction is the entire reason netting agreements exist, and it is also the reason netting is meaningless for lending. Netting sets amounts owed one way against amounts owed the other. If nothing is ever owed the other way, there is nothing to set against, and no document can create an amount that was never owed. A request to net a home loan against something leaves nothing on the far side of the sentence.

How does each one move between one day and the next?

KC4 is where the difference stops being philosophical and starts costing money. C1's funded exposure of Rs 1,680 crore will be almost exactly Rs 1,680 crore next week. The balance moves only when a scheduled event happens, and scheduled events are on a calendar somebody can look at. The repayment profile is written into the facility, so even a year out the number is knowable within a narrow band.

C1's netted derivative current exposure of Rs 96 crore has no such property. The Rs 96 crore was struck on the morning somebody valued the two trades. Nothing about C1 has to happen for it to be a very different figure at the next valuation, and nothing about the bank has to happen either. Prices move, the two trades revalue, and the netted derivative current exposure follows. The bank's claim on C1 under those contracts can change materially without a single decision being taken by anybody at either institution.

Sit with how strange that is compared with lending. In a lending relationship, if the exposure grows, somebody caused it: a drawing was requested, a limit was raised, a facility was renewed. There is a person and a date. In a contract relationship, the exposure can grow overnight with nobody to name and nothing to approve, and the first anybody knows of it is the next valuation run.

What reduces each of them?

KC5 is where the two lists stop being the same length. Against a lending exposure the bank has two instruments. Security changes what is recovered after a failure, and repayment changes the balance itself. Against a contract exposure it has three. Netting arrives first and reduces the claim before any measurement is taken, and security then works on whatever is left. How a charge is valued and discounted, and how netting is credited, are covered separately and used here as given.

The ordering is the part worth carrying. Netting operates on the amount owed and it operates before the measurement; security operates on the recovery and it operates after the failure. Because netting cannot reach a lending exposure at all, a bank facing a name that is mostly lending has one lever, and a bank facing a name that is mostly offsetting contracts has a different and much more powerful one. Neither technique is generally better. Which one reaches the money is a property of the shape of the relationship, not a property of the technique.

Who has to be watched, and is it the same list for both?

KC6 is the shortest of the six. Credit risk requires watching one thing: whether the borrower can pay. Counterparty risk requires watching two. The borrower's ability to pay matters exactly as much, and on top of it the amount owed is itself moving. A counterparty whose accounts, grade and behaviour are all unchanged since last quarter can owe the bank twice what it owed then, and every question normally asked about the name would return the same answer it returned before.

A credit review that covers the borrower thoroughly and stops there is therefore complete for a lending relationship and incomplete for a contract relationship. Such a review has covered one of the two things that decide the loss.

Can one counterparty carry both kinds at once?

C1 carries both, on the same day, inside one reported figure, and that is the single most useful thing in this guide. The distinction runs inside a single counterparty rather than sorting counterparties into two piles.

Work the exposure at default through and the split falls out. Rs 1,680 crore of funded exposure, plus Rs 360 crore being the bank's own 50.0 per cent assumption applied to the Rs 720 crore undrawn committed line, plus Rs 120 crore of derivative exposure at default. The three add to Rs 2,160 crore. Of it, Rs 2,040 crore is lending and Rs 120 crore is contract. On the largest name in this bank's book, 94.4 per cent of the exposure at default can be read and 5.6 per cent has to be estimated.

Two of those figures are close enough to merge by accident, so keep them apart. The 94.4 per cent is the share of C1's Rs 2,160 crore exposure at default that comes from lending. Limit L1 shows C1 at 94.5 per cent utilisation, and that fraction is a completely different one: Rs 2,496 crore of reported totalFunded exposure plus the undrawn committed line plus derivative current exposure, which is the basis this bank measures its single-name limit on. against a limit of Rs 2,640 crore. Two fractions one tenth of a percentage point apart, describing two different things, on one counterparty.

How much of the money is the part that has to be estimated?

Very little of it, and this is where a reader can take away exactly the wrong conclusion, so it is worth being blunt. Across the bank's ten largest single names, the funded exposures total Rs 9,960 crore, the undrawn committed lines total Rs 2,880 crore, and the derivative current exposures total Rs 228 crore. Together that is the reported total of Rs 13,068 crore. The contract part is 1.7 per cent of it and the lending part is the other 98.3 per cent.

Now watch what a reasonable person does with that number. If contracts are under two per cent of the exposure, surely they deserve under two per cent of the attention. The share of the money and the share of the measurement problem are two entirely different quantities, so the inference is wrong. The Rs 12,840 crore of lending needs one thing: a balance. The Rs 228 crore of contracts needs a valuation of every trade, a view on whether each netting agreement would hold, a notional for each contract and an add-on factor to apply to it.

And here the case is honest about its own limits in a way worth noticing. Of the five names carrying derivative current exposure, only C1 has a notional recorded. C1 is therefore the only counterparty on which the derivative exposure at default can be completed at all. For C3, C5, C7 and C10 the record holds a current exposure and nothing to build a potential future exposure from, leaving the exposure at default unproduced. Their combined derivative current exposure is Rs 132 crore, being C3 at Rs 24 crore, C5 at Rs 12 crore, C7 at Rs 36 crore and C10 at Rs 60 crore. C1's un-netted derivative current exposure is also Rs 132 crore and is a completely different object. Do not confuse the two: one is a sum across four counterparties, the other is what a single counterparty's two trades come to before netting.

How much of this bank's book moves with the market?

Exactly half of it, by name. Five of the ten largest single names carry derivative current exposure and five carry none at all, and the split is clean enough to be worth learning as a fact about this book. The list below is ranked by funded exposure, this bank's own published basis. On any other basis the ranking changes.

Take the two halves seriously as two different objects. C2, C4, C6, C8 and C9 are made entirely of amounts that are already known, so no price anywhere on earth can change their exposure figures by a rupee. C1, C3, C5, C7 and C10 each carry a part that is a valuation, so each of their totals is partly a statement about the morning it was struck. Both halves sit in the same table, in the same font, with no marking to tell them apart, and that is exactly how the confusion gets in.

Which counterparty is hardest to measure, and is it the largest one?

Ask which name in this book takes the most measurement judgement and almost everybody answers C1. C1 is the biggest, carries the largest derivative current exposure in rupees, and is the only name with a netting agreement recorded. Every one of those statements is true and the answer is still wrong.

Measure it as a share of each name's own total and the order inverts. C1's Rs 96 crore is 3.8 per cent of its Rs 2,496 crore total. C3 is 1.5 per cent, C5 is 1.1 per cent, C7 is 3.4 per cent. And C10 Betwa Speciality Chemicals, an invented speciality chemicals maker, carries Rs 60 crore of derivative current exposure against a total of Rs 780 crore. The share is 7.7 per cent, more than twice the proportion at C1. The most contract-heavy name in this book is the smallest one on the list, and no report sorted by size will ever point at it.

The everyday shape of that is close at hand. A vegetable seller with a stall worth Rs 3,00,000/- who has taken Rs 20,000/- of stock on a price that has not been fixed yet is carrying more of his own business in an unsettled number than a wholesaler with a Rs 40,00,000/- shop and Rs 60,000/- of the same. In rupees the wholesaler has three times the unsettled amount. In proportion to what he is, the stall holder has far more. Both facts are true. Only one of them identifies whose figures could be badly wrong.

What happens to a ranked list when only half of it can move?

Here is the cleanest demonstration this case can offer. Take the ten names ranked by total exposure, hold every funded exposure and every undrawn committed line completely still, and multiply only the derivative current exposures. Nothing else changes. No borrower does anything. No limit is raised. Five of its members are made of numbers that a market can move and five are not, so the list rearranges itself.

Ranked on total exposure at the book as it stands, the order is C1 at Rs 2,496 crore, C2 at Rs 1,920 crore, C3 at Rs 1,584 crore, C4 at Rs 1,320 crore, C6 at Rs 1,200 crore, C5 at Rs 1,092 crore, C7 at Rs 1,056 crore, C8 at Rs 960 crore, C10 at Rs 780 crore and C9 at Rs 660 crore. Note that this is not the same order as the list ranked by funded exposure, where C5 sits above C6 and C9 sits above C10. One book, one day, two ranking bases and three names in different places, before any market has moved at all.

Now let the multiple run from one to ten and eight separate moments arrive at which two names draw exactly level. At a multiple of 2.50 C7 draws level with C5 at Rs 1,110 crore each. At 4.00 C10 draws level with C8 at Rs 960 crore. At 5.00 C7 draws level with C6 at Rs 1,200 crore. At 7.50 C10 draws level with C5 at Rs 1,170 crore. At 8.00 C10 draws level with C6 at Rs 1,200 crore. At about 8.33 C7 draws level with C4 at Rs 1,320 crore. And at 10.00 two happen at once: C5 draws level with C6 at Rs 1,200 crore and C10 draws level with C4 at Rs 1,320 crore. C1, C2 and C3 change place with nobody anywhere in that range.

When does the difference actually change a decision?

The honest answer is that it changes one decision more than any other, and it is not a decision anybody thinks of as a risk decision. The decision is how often a name gets looked at.

Sorted by size, this book gives a sensible-looking answer: the big ones get watched closely and the small ones less closely. Sorted instead by whether the exposure figure can go stale, this book gives a completely different list. The second list decides whether the number in front of the committee is still true. Size gives how much is at stake, and only the kind of exposure gives how quickly the answer rots. A bank that sets a single review frequency for everything above some rupee amount has sorted its book on the wrong attribute, and it will have done so in perfectly good faith. Size is the attribute the report puts in front of it.

Notice also that nothing on a standard exposure report marks the difference. C2 at Rs 1,920 crore and C10 at Rs 780 crore appear in the same table, in the same typeface, one above the other. The first is entirely amounts already known. The second is 7.7 per cent a valuation. Nobody reading down the column would guess.

What is worth carrying away from all of this?

One sentence holds all of it. Credit risk is a failure to pay an amount that already exists, and counterparty risk is a failure to pay an amount that somebody had to construct, and every other difference between them is a consequence of that one. The direction can reverse only in the second. The number moves between days only in the second. Netting reaches only the second. And in the second, the market decides the size of the loss, so the market has to be watched alongside the borrower.

The second thing to carry is a proportion and a warning attached to it. On this bank's largest counterparty, 94.4 per cent of the Rs 2,160 crore exposure at default can be read and 5.6 per cent has to be estimated. Across the ten largest names, derivative current exposure is Rs 228 crore of a Rs 13,068 crore reported total, being 1.7 per cent. Almost every technique in counterparty measurement, netting, the add-on, the current exposure method, lives inside that small share. A bank spends most of its measurement effort on the smallest part of its exposure, and the smallest part is the only one that can move on its own.

A third risk sits next to both of these and is neither of them. Settlement riskThe risk that one side of an exchange pays and the other does not on the same day, which is a risk about a moment rather than about a period. is the risk that one side of an exchange pays and the other does not on the same day. Settlement risk is a third object with its own measurement, distinct from both credit risk and counterparty risk, and it is covered separately.