Funding Concentration: Depending on Too Few Sources

Funding concentration is depending on too few sources of money. At Vindhya Commercial Bank Limited, invented, wholesale funding is Rs 19,920 crore of Rs 88,320 crore, breaching limit L10, and twenty depositors hold Rs 11,136 crore of deposits, breaching L12. Both are open at month 12 and both carry the same named risk owner.

The workbench opens on the invented bank's own month 12 figures, and the arithmetic behind them runs as follows. Total assets of Rs 96,000 crore less equity of Rs 7,680 crore give total liabilities of Rs 88,320 crore; deposits of Rs 76,800 crore and borrowings of Rs 8,400 crore leave Rs 3,120 crore of other liabilities that fund nothing. Wholesale funding of Rs 19,920 crore is 22.554 per cent of those liabilities against a limit L10 cap of Rs 17,664 crore, so utilisation is 112.8 per cent and the excess is Rs 2,256 crore. The twenty largest depositors hold Rs 11,136 crore of the deposits, being 14.5 per cent against a limit L12 cap of Rs 9,216 crore, so utilisation is 120.8 per cent and the excess is Rs 1,920 crore. Nine holders divide Rs 2,880 crore of certificates of deposit, the largest three at Rs 640 crore each and the other six at Rs 160 crore each, so the largest single holder is 22.22 per cent of that book.

Now the reasoning behind those two limits. A bank borrows short and lends long, and the shortfall this creates in the near buckets of its maturity ladder is the ordinary shape of the business rather than a fault. The shortfall is a fact about time. A second fact sits on exactly the same rupees and the ladder cannot show it: not when the money is due to leave, but who has to decide before it does. Funding concentrationDepending for funding on too few providers, too few types of funding, or too few dates, so that one decision elsewhere removes a large share of the money. answers that second question, and at this invented bank it is why two of the twelve limits are in open breach at the reporting date while the ladder limit beside them is comfortably within.

What is funding concentration, and why is it a separate risk from maturity mismatch?

Take the smallest bucket on the ladder at Vindhya Commercial Bank Limited, invented. In the first fourteen days, Rs 9,600 crore is due out and Rs 7,200 crore is due in, so the bucket carries a shortfall of Rs 2,400 crore. The bucket is a complete statement about time: the size of the hole and the fortnight it opens in, and nothing about it is incomplete on its own terms. The bucket cannot show whether that Rs 2,400 crore is owed to four hundred thousand people with savings accounts or to nine institutions with a treasurer each. The two banks have identical ladders and are not remotely the same bank.

Here is the same mechanism at household size, where it is easier to feel. Two households each need Rs 40,000 in the coming month and each expects Rs 38,000 to arrive, so both carry the same Rs 2,000 shortfall and on a ladder they look alike. In the first the Rs 38,000 arrives as one salary from one employer. In the second it arrives as takings from a stall that served about nine hundred customers last month. One conversation ends the first household's income; it would take a strike, a flood or a road closure to end the second. The shortfall is identical, the exposure is not, and no arithmetic about timing will ever separate them.

So a liability carries two facts about the same rupee at the same time. The liability has a date, and the maturity ladder records it. The liability also has a provider, and nobody records the provider unless somebody decides to. Retail fundingMoney from many small depositors, which behaves more slowly and is the reason a behavioural assumption exists at all. is the many-small kind: it moves slowly, rarely all at once, and that stickiness is why a bank is allowed a behavioural assumption about it. Wholesale fundingMoney raised from institutions and large depositors rather than from many small ones, which typically leaves faster and prices on the name. is the few-large kind: professionally managed, priced on the name rather than the branch, and withdrawable by a person whose job is to decide whether to keep it there. Concentration risk is depending too heavily on the second kind, on too few of the people who supply it, or on too few dates when it falls due.

Concentrated in what: the type of money, the people who provide it, or the date it falls due?

Concentration is not one measurement but three, answering different questions about the same liability side. The first is by type: how much of the funding is the fast-moving institutional kind rather than the slow-moving retail kind. The second is by provider: how much comes from the largest few names, whatever type they are. The third is by date: how much falls due inside one window, so a single bad week has to be got through with a single refinancing effort. A bank that measures one of the three has answered a third of the question, and the three can point in different directions on the same book.

At the invented bank the first two carry limits of their own: limit L10 measures the wholesale share of total liabilities and limit L12 measures what the twenty largest depositors hold as a share of total deposits. The third gets no funding limit at all and is seen only inside the maturity ladder, where limit L9 caps the negative gap in the first fourteen day bucket. A limit is a statement about what somebody decided to watch, so a dimension with no limit on it is watched by nobody between board meetings. The absence is worth noticing before any figure appears.

How is the first one measured, and what exactly sits inside the numerator?

The wholesale funding share is a share, so it needs a numerator and a denominator, and both are decisions before either is a number. At this invented bank the numerator is wholesale term deposits of Rs 11,520 crore, meaning individual deposits at or above the bank's own Rs 5 crore cut-off, plus the whole borrowing book of Rs 8,400 crore. The borrowing book itself has four parts: certificates of depositA short dated instrument a bank issues to raise wholesale money, named here and taught in the fixed income subject area. of Rs 2,880 crore, interbank call and notice money of Rs 1,440 crore, refinance from a financial institution of Rs 1,680 crore, and tier 2 subordinated bonds of Rs 2,400 crore. The four come to Rs 8,400 crore; with the term deposits added, the numerator is Rs 19,920 crore.

The denominator is total liabilities, being total assets of Rs 96,000 crore less equity of Rs 7,680 crore, or Rs 88,320 crore. The share is 19,920 over 88,320, or 22.554 per cent, and gets reported as 22.6. Limit L10 caps it at 20.0 per cent, and 20.0 per cent of Rs 88,320 crore is Rs 17,664 crore. The share is over the cap by Rs 2,256 crore, and the moment a measured amount exceeds a cap there is a breachA limit exceeded, which is a fact about a measurement rather than a judgement, and which the committee then accepts with a plan or refuses.. A breach is a fact about a measurement and not yet a judgement. Breach B3 was first crossed in month 11 at 21.4 per cent, with the cause written down as retail term deposits running off and being replaced with certificates of deposit.

The arithmetic rule that must not be broken: utilisation comes from the amounts

UtilisationWhat is running against a limit expressed as a percentage of the limit, computed from the underlying amounts and never from rounded percentages. is what a limit report publishes, and it is where a small carelessness turns into a number nobody can trace back to the book. There are two ways to compute it and they do not agree. From the amounts, Rs 19,920 crore over the cap of Rs 17,664 crore is 112.77 and reports as 112.8 per cent. From the percentages, 22.6 over 20.0 is 113.0 per cent. Because 22.6 is already a rounded 22.554, dividing it carries that rounding into the answer, so the second route gives a wrong number rather than a rounding difference.

The rule holds everywhere a limit is reported: divide the rupee figures, never the percentages. Two tenths of a point looks like nothing at the time and is not nothing later. A report that cannot be rebuilt from the book it came from turns into an argument about which of the two answers was wrong. The route selector in the workbench above produces both readings on demand, so the difference can be watched rather than taken on trust.

How is the second dimension measured, and what does an aggregate hide beneath it?

The top twenty concentrationThe share of total deposits held by the twenty largest providers, which is an aggregate and says nothing about the shape beneath it. measure asks a simpler question: of the Rs 76,800 crore of deposits this bank holds, how much sits with the twenty largest depositors? The answer is Rs 11,136 crore, or exactly 14.5 per cent of the deposits. Limit L12 caps that at 12.0 per cent, being Rs 9,216 crore, so utilisation is 11,136 over 9,216, or 120.83, and reports as 120.8 per cent with an excess of Rs 1,920 crore. Breach B4 was first crossed in month 10, when a single state undertaking placed Rs 1,440 crore.

Now the honest warning about the measure itself. Twenty names holding 14.5 per cent averages Rs 556.8 crore each, and that average says almost nothing about the shape underneath. The largest holds Rs 1,440 crore, being 1.9 per cent of all deposits on its own, and the other nineteen average Rs 510.3 crore. An aggregate over a fixed number of names is a threshold measure and not a shape measure: it moves when a name crosses into or out of the twenty, and it is silent about whether the twenty are even or wildly uneven. The bank publishes the total and not the distribution, so the case records no individual balance for the other nineteen.

Why does the definition of the category decide whether there is a breach at all?

Look back at the five components in the numerator and one of them behaves nothing like the other four. Tier 2 subordinated bonds of Rs 2,400 crore are long dated instruments the bank issued once, and nobody wakes up and decides to withdraw them. Wholesale funding is watched because it can leave on somebody's decision, quickly, and price on the name while it goes. A subordinated bond does not do that, and yet the bank's own definition puts the whole borrowing book inside the numerator, tier 2 included.

Take them out and the arithmetic moves a long way. Wholesale funding becomes Rs 17,520 crore, the share becomes 19.84 per cent of Rs 88,320 crore, and that is inside the 20.0 per cent cap at a utilisation of 17,520 over 17,664, being 99.2 per cent. One line item is the whole distance between a breach reported at 112.8 per cent and a limit sitting comfortably within. A limit is a definition before it is ever a number. The rule is general and holds for every ratio anybody reports: whether a number is too high cannot be argued until the person who computed it says what was counted.

Does removing the depositor that caused the breach cure the breach?

Breach B4 has a recorded cause, and it is a single event: in month 10 a state undertaking placed Rs 1,440 crore and the top twenty measure went over its cap. The obvious remedy is to reverse the event, so suppose that deposit is repaid tomorrow. A deposit that leaves is no longer in the top twenty and no longer part of total deposits, so the repayment comes out of the numerator and the denominator at the same time. The measure becomes 9,696 over 75,360, or 12.87 per cent against a limit of 12.0 per cent. The breach does not close, and the number that does not close it is the one everybody assumed would.

Worse, 12.87 per cent is a floor rather than an answer. Take one name out of the top twenty and the twenty-first largest depositor steps up into the list, so the measure after the repayment is at least 12.87 per cent and possibly higher, and nobody can say how much higher because the case gives no balance for that name. The floor does not contradict the recorded cause: the limit really was crossed in month 10 by that event, and by month 12 the breach is held open by the shape of the deposit book. A breach can begin with an event and be sustained by a structure, and once it has, reversing the event is no longer the cure. A breach held open that way is a structural breachA breach caused by the shape of the balance sheet rather than by one event, which does not close by itself when the event is reversed., and a structural breach is why the two open funding breaches behave nothing like the three that closed in days.

Why does a concentration measure get worse when the balance sheet shrinks?

Here is the property that makes a concentration ratio behave unlike anything else on the liability side. Suppose retail deposits run off by Rs 3,000 crore over a quarter, for ordinary reasons, and the bank raises not one rupee of wholesale funding to replace them. Nothing has been done on the concentrated side and wholesale funding still stands at Rs 19,920 crore, but total liabilities fall to Rs 85,320 crore, so the share rises to 23.35 per cent, the cap falls to Rs 17,064 crore and utilisation rises from 112.8 to 116.7 per cent. The run-off control in the workbench walks the whole path.

The measure got worse because the part of the book that leaves is the part that was not concentrated, so a shrinking balance sheet concentrates what is left. The household case says it again: if the second earner stops working, the household's dependence on the first salary rises to a hundred per cent without the first earner doing anything. The consequence for a treasurer is worth sitting with. A bank losing deposits in a difficult quarter watches its funding concentration limit deteriorate at the moment it has least appetite to raise more wholesale money, and every remedy available makes one of the two numbers worse.

How few providers is too few, and what does that do to a warning indicator?

The third place concentration shows up at this bank is in the paper itself, which is the book the provider ledger above opens on. Certificates of deposit outstanding are Rs 2,880 crore held by 9 counterparties, of which the largest 3 hold Rs 1,920 crore between them, being 66.7 per cent of the book. The other 6 hold Rs 960 crore, averaging Rs 160 crore each. The largest 3 average Rs 640 crore each, four times as much. The case does not record the three individually.

Now put an early warning indicator on top of that. The bank's early warning indicator W2 watches the share of certificates of deposit rolled at each auction, turning amber below 90 per cent and red below 75. Nine holders means the average holder is 11.1 per cent of the book. Amber fires on a shortfall of only 10.0 per cent. One average holder declining to roll gives 2,560 over 2,880, being 88.9 per cent, so the indicator is already amber; any one of the largest three gives 2,240 over 2,880, being 77.8 per cent, still amber; any two of them gives 1,600 over 2,880, being 55.6 per cent, and that is red. An indicator calibrated in percentages can never be finer than the lumpiness of the book it measures, and with nine holders this one cannot separate a market turning against the name from one counterparty rebalancing.

Who is accountable when a concentration breach stays open for months?

Three of this bank's twelve limits touch funding. Limit L9 caps the negative gap in the ladder's first fourteen day bucket and sits within at 89.3 per cent. Limit L10 on wholesale share is at 112.8 per cent and in breach since month 11, and limit L12 on depositor concentration is at 120.8 per cent and in breach since month 10. Both open breaches name the same risk ownerThe named person accountable for a risk and for the plan that addresses it, which is a person and never a function.: Devendra Achar, the head of treasury. The ladder reads clean and the funding does not, and one line of a limit report holds both the timing question and the provider question.

A risk owner is a person and never a function, and the reason is practical: a plan with a date on it needs somebody who can be asked in month 14 whether the date held. Note also what an open breach is not. An open breach is not, on its own, evidence that anybody failed. The bank's committee accepted its sector concentration breach in month 6 as a temporary excess with a remediation plan running to month 18. Accepting a breach is a decision and not an absence of one. But the two funding breaches share a property no limit report shows: both are held open by the shape of the balance sheet rather than by any single trade. The three breaches that closed in the year all closed within days, being positions that could be traded out or allowed to run off. A position closes itself. A structure does not.

Who actually reads a funding concentration number, and what do they do with it?

Four different people pick this number up and none of them reads it for the same reason. The treasurer reads it as a constraint on the next quarter's plan. At 112.8 per cent, every rupee of wholesale money raised before the next committee makes a reported breach deeper, so the funding plan has to be rewritten around a retail gathering effort that takes months rather than days. A concentration limit costs exactly that: it does not stop the bank funding itself, it changes which door the funding has to come through, and the slow door is the only one that helps.

The credit analyst at another institution, looking at this bank as a counterparty, reads the same number as a speed estimate. A high wholesale share says that a larger part of these liabilities is managed by people who read the news, and the share is therefore a way of guessing how quickly a bad headline turns into an outflow. The definition is the one feature that makes two banks comparable and the feature least likely to match between them, so the analyst attends far more to the definition than to the level.

The board member reads it as a question about the plan rather than about the number. Two funding limits in open breach, with one named owner, invites one question: is the remediation plan a plan for replacing the funding, or a plan for shrinking what needs funding? The household reader has the same lesson at a scale they can feel. If most of the money coming into a house arrives from one employer, that is a concentration, and the honest remedies are the slow ones: a second earner, a side income, a buffer big enough to bridge a search. Spending less does not reduce the dependence. Spending less concentrates the dependence, exactly as the invented bank's ratio worsens when its balance sheet shrinks.