How to assess Concentration Risk: A Seven Step Method

Seven steps in a fixed order. Choose the dimensions and test the list for what is missing, fix the ranking basis before sorting anything, apply the group definition to surface what no single name report shows, measure each level against its limit, turn the question round to test whether a concentration could actually reach the capital floor, and state the basis beside every figure. At this bank the answer is one sector.

One uncomfortable fact about concentration decides how every step of the method is read. The arithmetic of concentration is trivial. The exposures in one bucket are added up, divided by something, and compared with a limit. A school pupil can do it. Every real finding in the bank worked through below came from somewhere else entirely: from deciding what the rows were, what they were sorted on, what was not in the table at all, and which two rows were secretly the same borrower. A concentration measure is only ever as good as the list it was computed on, and six of the seven steps below are ways of interrogating that list rather than the sum.

The worked institution is Vindhya Commercial Bank Limited, an invented mid-sized Indian commercial bank with a balance sheet of Rs 96,000 crore and gross advances of Rs 58,800 crore. Every figure below belongs to that bank or to its counterparties, and every limit, floor and loss assumption is one the bank set for itself rather than a requirement imposed on it.

In what order should concentration be assessed?

Concentration riskToo much exposure sitting in one name, one group, one sector or one anything, which is a fact about the shape of a book today rather than a forecast. is not a forecast and it is not a probability. Concentration is a fact about the shape of a book on one day: too much of the book is in one place. No model is needed to see it, and that makes concentration unlike almost everything else in credit. A table is what is needed, along with the right questions asked of that table before it was drawn.

The mechanism is identical at every scale, so start with the household version. A household running on one salary is concentrated. Nothing has gone wrong; the arithmetic of the monthly budget is perfectly sound; every bill is being paid. But one event, the loss of that one job, takes the income to zero rather than reducing it. Now put a second earner in the house and the same event halves the income instead. Nothing about the household got cleverer. The shape changed. Concentration is a statement about how many separate things would have to go wrong, and a book with one big exposure needs only one.

The seven steps run in a fixed order because each one changes the table the next one reads. Choosing dimensions before sorting is obvious enough; fixing the ranking basis before applying the group definition is less obvious, and is exactly the step most often skipped.

Which dimensions are chosen, and what shows the list is complete?

A dimensionAn attribute a book can be concentrated along. The chosen list is always shorter than the possible list, which is the whole difficulty. is any attribute a book can be concentrated along. Vindhya Commercial Bank Limited chooses six: single name, borrower group, sector, currency, tenor and collateral type. The choice of six dimensions is not administrative. A dimension list is a claim that these are the places where concentration could hide, and a claim is a thing that can be wrong.

Consider the household again. An income concentrated in one job is plainly a risk. Is a household whose entire savings sit in one bank, whose home and workplace are in one town liable to flood, and whose two earners work for the same employer also concentrated? All four are true at once, and only the first is the one anybody thinks to name. The others are dimensions nobody chose to look along.

Step CA2 is the test, and it has only one form: go and look for what is not on the list. At this bank the answer is immediate. Three of the six dimensions carry a limit and are reported monthly. Three of them carry no limit at all. There is no cap on how much of the loan book sits in one currency, none on how much of it reprices in one stretch of the maturity profile, and none on what kind of asset secures it. A book heavily secured on one class of collateral would show green on every concentration report this bank produces, not because the report is wrong but because nobody is looking along that line.

Why must the ranking basis be fixed before anything is sorted?

The ranking basisThe column a table is sorted on. It changes which rows appear on a top ten, not only the order they appear in, which is why it is chosen first. is the column the table is sorted on, and it sounds like the least interesting decision in the whole method. It is not. Vindhya Commercial Bank Limited publishes its ten largest single name exposures ranked on funded exposure, meaning money actually out of the door. The bank could equally have ranked them on total exposure, meaning funded plus the undrawn committed line plus the current exposure on derivative trades.

Both bases are correct. Both describe the same book on the same day. The two lists disagree.

Read the two columns against each other. C6 Manjeera Housing Finance Limited rises above C5 Tapti Agro Processing Limited because it carries a larger undrawn committed line. C10 Betwa Speciality Chemicals Limited rises above C9 Lohit Valley Tea Estates Limited because it carries Rs 60 crore of derivative current exposure and C9 carries none. And once the group definition of the next step admits Nirjhar Alloys Private Limited at Rs 672 crore, C9 leaves the ten altogether. Three names change place and one drops off the list, and not a single rupee of exposure moved: only the sort column changed.

How is an exposure that appears on no report found?

A borrower groupA parent and the entities it controls, treated as one exposure because they tend to fail together rather than separately. is a parent and the entities under its control, treated as one exposure because they tend to fail together. Step CA4 applies that definition to the list, and at this bank it produces the single most useful number in the case.

The Nirjhar group has three entities. E1 Nirjhar Industries Limited is the parent and is counterparty C1, the bank's largest single name at Rs 2,496 crore of total exposure. E2 Nirjhar Alloys Private Limited is a wholly owned subsidiary carrying Rs 672 crore, and the bank reports it inside a group line rather than as a name of its own, so it appears on no row of the published top ten. E3 Nirjhar Trading FZE is an overseas trading entity and the bank's book records no exposure to it at all.

Connect E1 and E2 and the group exposure is 2,496 plus 672, being Rs 3,168 crore. The Rs 3,168 crore exists on no report anybody reads until somebody applies the group definition, so this bank can be Rs 3,168 crore into one economic borrower while correctly reporting its largest exposure as Rs 2,496 crore. Both numbers are right. Only one of them states what would be lost if the group failed, and against tier 1 capital of Rs 6,600 crore that group exposure is 48.0 per cent.

The street version is a shopkeeper who has extended credit to a wholesaler and, separately, to the wholesaler's brother who runs the loading business next door. Two ledgers, two names, two amounts, both accurate. One phone call from the bank that funds the wholesaler and both ledgers stop being paid on the same afternoon.

What is each level measured against?

Step CA5 is the one piece of arithmetic, and it has a rule: each level is measured against its own limit and never against another level's. Vindhya Commercial Bank Limited holds three, each set by the bank itself. Limit L1 caps single name exposure at Rs 2,640 crore. Limit L2 caps borrower group exposure at Rs 3,960 crore. Limit L3 caps any one sector at Rs 7,056 crore, being 12.0 per cent of gross advances.

Now read the table the way a risk committee should read it. The largest single name is comfortable. The largest borrower group is comfortable. The only level actually over its limit is the sector, and at Rs 7,644 crore it is more than three times the size of the largest single name in the book. A reader who checked only the names would have found nothing, twice.

The infrastructure and power position is an open breachA limit that has been crossed and is still crossed. Here it has been accepted by the board committee with a dated plan rather than left unnoticed., numbered B1, first crossed in month 5 at 12.2 per cent of gross advances and standing at 13.0 per cent at month 12, an excess of Rs 588 crore. The breach has been accepted as a temporary excess with a remediation plan running to month 18. Who accepted it and on what authority belongs to the governance subject. The shape of the breach is what matters for the assessment: three other breaches at this bank closed inside days because a position was traded out or ran off. A concentration breach is unwound loan by loan as facilities mature, so a breach caused by the structure of the balance sheet does not close by itself and is read on a different clock from a trading one.

What separates a concentration that is dangerous from one that is merely large?

The first five steps establish how big things are. None of them establishes whether any of it matters. A book can be lumpy and perfectly survivable; another can look smooth and hold one exposure that would finish it. Step CA6 separates the two, and it does it by turning the question round.

A forward question asks what happens if a scenario occurs. A reverse testTurning the question round to solve for what would have to happen to reach a fixed outcome. The method belongs to another subject and only its answer is used here. fixes the outcome first and solves backwards for the cause. How that method is designed and run is covered separately. Only the answer matters for a concentration assessment, and the arithmetic to get there is two multiplications.

Fix the outcome as the bank's own capital floorThe lowest capital ratio the institution is willing to run at. Here it is the bank's own 11.0 per cent internal figure and not a regulatory minimum. of 11.0 per cent. The bank set that floor for itself, and no requirement of any kind imposes it. Total capital is Rs 9,000 crore and risk weighted assets are Rs 60,000 crore, so the floor needs Rs 6,600 crore of capital held. The other Rs 2,400 crore is loss capacityThe most that could be lost before the capital floor is reached. It is arithmetic rather than a choice, and it ignores the fall in risk weighted assets as losses are written off.. Loss capacity is arithmetic and not a choice. The subtraction also ignores the fall in risk weighted assets as losses are written off, so Rs 2,400 crore reads conservatively. One warning on the figures: the Rs 6,600 crore the floor consumes and the Rs 6,600 crore of tier 1 capital used to size limits L1 and L2 are two different objects that happen to share a number in this bank.

Now solve backwards. At the bank's own assumed loss given default of 40.0 per cent, losing Rs 2,400 crore takes Rs 6,000 crore of exposure actually defaulting. Divide 2,400 by 0.40 and the answer is 6,000. Rs 6,000 crore is 10.2 per cent of gross advances. And the infrastructure and power sector holds Rs 7,644 crore. Six thousand crore is 78.5 per cent of one sector, so a single sector that is already over its limit is large enough on its own to take this bank to its own capital floor with nothing else going wrong.

The finding claims less than it appears to. A reverse test states arithmetic capacity and says nothing about likelihood. No probability has been computed anywhere in the seven steps, and a sector default is neither probable nor implausible on this evidence. The reverse test shows only that the bank holds enough in one place for the question to be worth asking.

What has to be written beside every figure?

Step CA7 is the shortest and the most often skipped. Every figure in a concentration assessment travels with three lines above it: what the rows are, what they are ranked on, and what each is measured against. Three lines, before any row.

Take them away and look at what Rs 2,496 crore at 94.5 per cent could mean to a reader who supplies the missing assumptions from habit. The Rs 2,496 crore could be a funded figure, a total figure, or an exposure at default figure, and it could be measured against any of three limits sitting at different levels. A concentration number without its basis is not one number read three ways, it is a different number in every reader's hands, and the reader cannot tell which one they are holding.

What can this method not do?

Two limits, and both of them are structural rather than fixable with more effort.

The first is correlation. The method shows infrastructure and power at Rs 7,644 crore and stops there. The method counts what sits in one bucket and assumes the bucket means something, and whether the borrowers inside it would actually fail in the same month is a different question that this method does not ask and this bank has not answered. A sector label is a classification, not evidence of shared fate. The case records the sector total and does not record which counterparties sit inside it, so C1 to C10 cannot be mapped into that Rs 7,644 crore, and any assessment that quietly did so would be inventing the very evidence it needs.

The second is the dimension list again, arriving from the other end. The method cannot find a concentration along a dimension nobody chose. Nothing is measured along currency, tenor or collateral type, so the method has no mechanism for discovering that any of the three mattered. Step CA2 is a discipline, not a detector: it makes somebody go and look, and the looking is human.

One thing this method is also not: it is not the assessment of a single counterparty, and it is not wrong way risk. Wrong way risk is the exposure growing in exactly the state of the world that weakens the counterparty. At this bank it sits at C7 Wainganga Textiles Limited and is a different fault settled elsewhere. Concentration needs no state of the world at all. A concentration is already true this morning.

Who picks this assessment up, and what do they do with it?

Three readers, three different uses, and the differences are worth seeing because they explain why step CA7 exists at all.

An independent director on the board committee reads it for the level. She does not need the ten names; she needs to know which of the three levels is over and whether the one that is over is the kind that closes by itself. At this bank the sector level is over, at 108.3 per cent, and it is the kind that does not close by itself. A sector position runs off loan by loan over years. The whole board reading is that pair of facts, and it fits in two sentences.

A credit analyst at another institution, looking at Vindhya Commercial Bank Limited as a counterparty rather than as an employer, reads it for the reverse test. He cannot see the loan book and does not need to. One question is enough: is any single reported concentration large enough to reach the bank's own floor? Rs 6,000 crore against a Rs 7,644 crore sector says yes, and that changes how he sizes his own line to this bank without saying anything about whether it will happen.

The mechanism does not change with the scale, so a household version lands in the same place. A household running a food stall outside one office building has a customer concentration of 100 per cent along a dimension nobody wrote down. The reverse test for that household is one sentence long: how many weeks of that building being shut takes the savings to zero? Notice that the useful question in both cases was not how large is the exposure but how much of it would have to go wrong. Step CA6 asks that question and no ranked list ever does.