The Contingency Funding Plan: What Happens When Funding Stops

A contingency funding plan is a written statement of what an institution does when funding stops arriving, built from stages, triggers and actions. At Vindhya Commercial Bank Limited, invented, the plan lists Rs 18,600 crore of actions. Run it through the name-specific stress it exists for and Rs 8,928 crore is actually there inside the first week, being 48.0 per cent.

Almost every institution of any size has one of these documents. The document is approved, it is on file, it is reviewed on a cycle, and in most places it has never been arithmetic. A plan is a list of things somebody could do, written in a calm room by people who were not frightened. The whole question is whether the things on the list would still be there on the morning they were needed. A calculator can answer that question, and the answer at this invented bank is that a little under half of the plan is real.

What is a contingency funding plan, and what is actually in one?

A contingency funding planA written statement of what an institution will do when funding stops arriving, made of stages, triggers and actions. is made of three things and no more. Take any of the three away and what is left is a description rather than a plan. The three are stages, triggers and actions, and the test of a document is not whether it contains sensible ideas but whether it contains all three of these in a form somebody could act on at eight in the morning with the phone ringing.

The first is the stageA named level of the plan, each with its own condition and its own set of steps, so a response is proportionate rather than all or nothing.. A plan with one setting is a switch, and a switch is the wrong shape for something that arrives gradually. Funding pressure does not turn up all at once. Pressure shows first as a slightly worse price on the bank's own paper, then as a slightly lower share of that paper being renewed, then as deposits leaving faster than they arrive. Stages let the response match the pressure. A bank noticing something odd is then not obliged either to ignore it or to declare an emergency.

The second is the triggerThe stated condition that starts a stage, specific enough that two people reading it on the same day reach the same answer.. The trigger is where most documents quietly fail, and the failure is easy to miss because the sentence reads well. A trigger has to be specific enough that two people reading it on the same day, with the same numbers in front of them, reach the same answer about whether the stage has started. Anything vaguer is not a condition, it is an invitation to a meeting, and meetings are what a plan exists to make unnecessary.

The third is the actionA step that raises or saves cash, stated with an amount, a lead time and the condition under which it would not be available., and an action is only an action if it carries a number. Without an amount and a time, a line saying that the bank will consider selling assets is not a step, it is an intention. With an amount and a time it becomes a quantity, and quantities can be added, compared against a window, and struck out one at a time. Everything interesting that follows depends on being able to strike them out.

The bank version is only the same shape at a larger scale, so the household version is worth holding on to. Consider a household running on one salary that suddenly stops. There is money in a savings account that can be moved this afternoon. There is gold that can be pledged in a couple of days. There is a cousin who has said, vaguely, that he would help. And there is a small plot of land in another district that could be sold in three or four months. The four of them written down are a contingency funding plan. The step almost nobody takes comes next: putting a rupee figure and a number of days against each of them, and then asking which of them still exists in the exact situation where they would be needed. The cousin who would lend money when things are fine is not necessarily the cousin who lends it when the whole street has heard about the lost job. The land takes four months, and the rent is due next week.

What are the three stages, and what starts each one?

The bank's plan has three stages, numbered P1 to P3, and each one carries its own condition. P1 is heightened monitoring, and it starts when any two early warning indicators turn amber. P2 is a name-specific stress, and it starts on a rating action, an adverse news cycle, or a failed auction of the bank's own certificates of deposit. P3 is a market-wide stress, and it starts when interbank rates move beyond the bank's own trigger, or when a peer fails. Every one of those conditions is the invented bank's own choice, and the three stages are not three severities of one thing. P1 is a level of attention. P2 and P3 are two different shapes of trouble that happen to be numbered in sequence.

The distinction between P2 and P3 matters more than the numbering suggests. A reader coming to this document for the first time sees P1, P2, P3 and reads a ladder: mild, worse, worst. It is not a ladder. P1 really is the mild one. But P2 and P3 are not a smaller and a larger version of the same event. The two stages are two different states of the world, and the actions that stop working are different in each. A plan that treats them as one ladder with a single list of steps has planned for one of them and hoped the other looks similar.

Why are a name-specific stress and a market-wide stress two different plans?

The everyday version is exact, so start there again. A name-specific stressA scenario in which the institution alone is in trouble, so the market is working normally and simply will not deal with this one name. is the street vendor whose stall was in the newspaper for a hygiene complaint. Every other stall on the road is doing fine. Money is available; suppliers are lending; the market is entirely healthy. The only difficulty is that nobody wants to be the person who extended credit to this particular vendor this particular week. A market-wide stressA scenario in which funding is scarce for everybody at once, so every institution reaches for the same actions on the same morning. is the week the whole market is shut for a flood. Everybody is in the same position, everybody wants to sell the same stock at the same time, and the vendor's own reputation has nothing to do with it.

The two shapes remove different actions, and that is the entire reason they cannot be treated as two severities of one thing. In the name-specific shape, the market works. Selling a government security still clears at a fair price. The buyer is buying a security and not buying the bank. Issuing the bank's own paper stops dead. Buying that paper is precisely the transaction in which the buyer is buying the bank. In the market-wide shape, the reverse pressure applies. Everybody's paper is expensive and this bank's is one of many, so the bank's own paper may still find a home at a price. The securities sale is the crowded trade. Every institution has reached for the same lever on the same morning, so what moves is not availability but price.

The bank's plan is written for the name-specific shape, and there is a specific reason why that is the relevant one here rather than a generic modelling preference. Its funding is concentrated. Wholesale funding runs at Rs 19,920 crore of total liabilities of Rs 88,320 crore, and the top twenty depositors hold Rs 11,136 crore, being 14.5 per cent of deposits. Its certificates of deposit of Rs 2,880 crore sit with nine counterparties, of which the largest three hold Rs 1,920 crore, being 66.7 per cent of the outstanding. Funding concentration is a subject of its own and is covered separately, but the consequence lands here: when a small number of people hold most of a bank's funding, an event that changes their view of that bank alone is the event that matters most. The name-specific test therefore runs here. The case does not say what a market-wide stress would remove, so the market-wide shape is covered separately.

What does an action have to state before anybody can count it?

An action needs four things on its face, and most written plans carry three. The three usually present are the step itself, how much it raises, and how long it takes. The fourth, and the one that does all the work here, is the condition under which the action would not be available. An action is not a resource until somebody else has agreed, and the fourth column is where that somebody is written down.

The clearest instance in this plan is action F4. Action F4 reads, on its face, as Rs 1,800 crore of interbank lines drawable on the same day, and same-day money is the most valuable kind there is. Then comes the fourth column. Of that Rs 1,800 crore, Rs 1,200 crore sits on uncommitted lineA facility the provider may decline, which is worth counting in ordinary conditions and worth nothing in the stress the plan exists for. facilities and Rs 600 crore on committed lineA facility the provider is contractually obliged to fund when it is drawn, which is a different object from one it may decline. ones. An uncommitted facility is one the provider may decline. An uncommitted facility is a perfectly reasonable thing to have, and in ordinary conditions it funds perfectly reliably. Reliability in ordinary conditions is exactly why it gets counted. But the state of the world in which a bank reaches for its contingency lines is precisely the state in which the provider on the other end reads the same news and decides not to fund. So the fourth column turns a Rs 1,800 crore action into a Rs 600 crore one, and it does that with a single word.

The full set as written, with all four columns. The last column is the one to read first: it decides how much of the plan is real.

Notice what the last row does to the arithmetic of the table. F1 to F5 add to Rs 18,600 crore, and F6 sits outside that total because it is a different kind of object. Stocks and flows are taken apart properly below. Notice also the pattern running down the final column: F1 needs a counterparty, F2 needs a buyer, F3 needs investors, F4 needs a lender who has not promised, F5 needs a purchaser and a legal process. Five of the six steps in this plan require somebody outside the bank to say yes, and only F6 does not. None of that is a criticism of the plan. The pattern is a description of what a funding plan is, and it is why testing one is not optional.

What happens when this plan is run through the stress it was written for?

Now do the arithmetic, one line at a time, and do it under a name-specific stress. A contingency funding plan exists for that stress. F1 stands: Rs 4,800 crore of repo against government securities held outside the buffer, and the counterparty is taking the security rather than taking a view on the bank, so it still clears on the same day. F2 stands, at a price: Rs 3,600 crore of securities sold at the assumed 2.0 per cent discount. The discount gives away Rs 72 crore and brings in Rs 3,528 crore. F3 goes, entirely: Rs 2,400 crore of certificates of deposit that nobody will buy. The whole definition of this stress is that the market has stopped wanting this name. F4 shrinks: Rs 1,200 crore of it was uncommitted and the provider declines, leaving Rs 600 crore. F5 is real, it is the largest single number on the list, and it arrives far too late. Lead times are taken on their own below.

F1 at Rs 4,800 crore plus F2 at Rs 3,528 crore plus F4's committed Rs 600 crore is what is actually there inside the first week. The three come to Rs 8,928 crore, or 48.0 per cent of the Rs 18,600 crore the plan appears to offer. And it reconciles to the rupee. A reconciliation to the rupee is what makes this an arithmetic finding rather than a judgement. Rs 18,600 crore less Rs 8,928 crore is Rs 9,672 crore, and that Rs 9,672 crore is exactly F5 at Rs 6,000 crore that arrives after the window, plus F3 at Rs 2,400 crore that nobody buys, plus the uncommitted Rs 1,200 crore of F4, plus the Rs 72 crore of discount on the F2 sale. Add those four: 6,000 plus 2,400 plus 1,200 plus 72 is 9,672. There is nothing left over and nothing missing.

Every combination, solved and written down

Two readings of that table are worth pulling out. The first is the diagonal: Rs 8,928 crore in the tested week is 71.3 per cent of the Rs 12,528 crore the same week offers when nothing is wrong, so more than a quarter of the plan's first week evaporates purely because the trouble is about this bank rather than about the market. The second is the last row. Under a name-specific stress with no window at all, Rs 14,928 crore eventually arrives, and Rs 6,000 crore of that is F5 turning up somewhere between about day 42 and about day 63. F5's money is real money and it is not contingency money, and the row is shown separately for exactly that reason.

Which actions arrive in time, and which arrive after the window has closed?

An amount without a lead timeHow long a step takes to produce cash, which is what decides whether it lands inside the survival window or outside it. is half a fact. The survival horizon is covered separately in its own right and is used here only as the window, so take it as given: on the Rs 14,400 crore buffer alone, under the bank's own severe scenario, this bank lasts 34 days. Thirty four days is the ruler. Every lead time in the plan gets measured against it, and the measurement is not a matter of opinion.

F1 and F4 are same day. F2 is two to three working days and F3 is three to five, so both sit comfortably inside the first week. F5 says thirty to forty five working days, and this is where a conversion is needed that the plan itself does not do. On a five day working week, thirty working days is six calendar weeks, or about forty two calendar days. Forty five working days is nine calendar weeks, or about sixty three calendar days. The horizon ends on day 34. So F5, the largest single action in the plan at Rs 6,000 crore, arrives at best about eight days after the money has already run out, and on the slow end about a month after that.

The mismatch is easy to miss, and the reason is plain. The plan states F5 in working days and the horizon in calendar days, so the two numbers sit in the same document in different units and never get compared. Thirty against thirty four looks like a comfortable fit when the two figures are read side by side without converting. The fit is not close. It is not even the same kind of number. A unit mismatch is the cheapest possible finding in a funding plan and it needs no analysis at all, only a column that says calendar days and a horizon printed next to it.

Why can a weekly saving not be added to the Rs 18,600 crore?

F6 is the odd one out and it is worth being precise about why. F1 to F5 are stock actions: each raises a fixed amount once, and once raised it sits in the pile. F6 is a flow actionA step that saves an amount per period for as long as it runs, so it reduces the daily burn instead of adding to the pile.: suspending new lending disbursals saves Rs 900 crore a week for as long as the suspension lasts. A stock and a flow cannot be added, and a plan that prints Rs 18,600 crore plus Rs 900 crore a week as though they were one total has produced a number that means nothing. The stock belongs in the numerator of the survival calculation and the flow belongs in the denominator. A presentation may show F6 either way but must not do both. Doing both counts it twice.

Convert it and the flow becomes readable. Rs 900 crore a week is Rs 128.57 crore a day. Now hold that against the bank's own modelled outflow path under its severe scenario. The path runs at Rs 720 crore a day on days 1 to 5, Rs 600 crore on days 6 to 10, Rs 420 crore on days 11 to 20, Rs 240 crore on days 21 to 30, and Rs 300 crore a day after that. F6 covers 17.9 per cent of the burn in the first five days, 21.4 per cent in the second five, 30.6 per cent through days 11 to 20, 53.6 per cent through days 21 to 30, and 42.9 per cent thereafter. The same action is worth three times as much late in the stress as it is early in it, purely because the thing it is measured against has shrunk.

The inversion is genuinely counterintuitive and it is worth sitting with. The instinct here is exactly backwards. The instinct says: act early, so do everything on day one. For a stock action the instinct is right. Cash raised on day one is cash held on every day after it. For a flow action the instinct is misleading. A saving of Rs 128.57 crore a day against a Rs 720 crore day is a rounding difference. The same saving against a Rs 240 crore day is more than half the problem. The household version is a person who stops eating out. In the first frantic week after the salary stops that saving is invisible against the rent and the school fees. By the second month, when the big bills have already been dealt with one way or another, it is a meaningful share of what is left going out.

What does the plan actually do to the survival horizon?

Put the two readings side by side and the size of the difference becomes hard to argue with. On the Rs 14,400 crore buffer alone, this bank lasts 34.0 days. Add the plan as written, Rs 18,600 crore, and the stock becomes Rs 33,000 crore. At the modelled Rs 300 crore a day beyond day 30 the stock reaches day 96.0. Add instead what actually arrives in a name-specific stress, Rs 8,928 crore, and the stock is Rs 23,328 crore. The same arithmetic reaches day 63.8. The plan as written looks like three months and the plan as tested is about two, and the whole of that difference is made of decisions taken by people who do not work at this bank.

Days added rather than days reached is the honest measure of what a plan contributes. Against 34.0 days on the buffer alone, the plan claims to add 62.0 days and actually adds 29.8 days, so 32.2 of the claimed days are not there. And a small point of arithmetic is worth noticing: 29.76 days over 62.0 days is 48.0 per cent, exactly the share of the rupees that survived. The match is not a coincidence and it is not an accident of rounding. Beyond day 30 the modelled outflow is a flat Rs 300 crore a day, so days and rupees in that region are the same measure divided by three hundred, and any percentage computed on one of them comes out identically on the other.

At month 12, with one indicator red and one amber, which stage is running?

The plan says P1 starts when any two early warning indicators turn amber. There are seven of them, W1 to W7, and each carries the bank's own amber and red trigger. W1 watches the cost of the bank's own certificates of deposit against the interbank rate. W2 watches the share of that paper renewed at each auction. W3 watches daily net retail deposit flow. W4 watches the top twenty depositor share. W5 watches how much of the undrawn committed lines customers have drawn. W6 watches the buffer as a share of thirty day net outflows. W7 watches the survival horizon.

At month 12, two of the seven are lit. W4 is red: the top twenty depositors hold 14.5 per cent of deposits against an amber trigger of 11.0 per cent and a red trigger of 12.0 per cent. W7 is amber: the survival horizon is 34 days against an amber trigger of 40 days and a red trigger of 30 days. Now read the P1 condition again, literally. A document written for a bad morning has to be read that way. The condition says two indicators turning amber. The bank has one red and one amber. The plan does not say whether a red counts as an amber for the purpose of the trigger, so on a literal reading one red and one amber is not two ambers, and P1 is not running.

Follow it through and the position is genuinely odd. If W4 were merely amber instead of red, that plus W7 would be two ambers and P1 would start. The bank is in a worse state than the state that would start the stage, and the stage has not started. Meanwhile P2 needs a rating action, an adverse news cycle or a failed auction, and none of the three has happened. P3 needs interbank rates beyond the bank's own trigger or a peer failure, and neither has happened. So at month 12, with one indicator red and one amber, no stage of this plan is running at all.

The finding needs stating carefully, both what it is and what it is not. The gap is not a failure by anybody, and it is not evidence that the bank is unsafe or that a person was negligent. It is a missing sentence in a document. Somebody drafting a trigger wrote amber because amber was the word in front of them and did not stop to ask what happens when an indicator jumps straight past amber. Fixing it costs nothing, changes no number anywhere in the case, and requires no approval to spend anything. The cost of the fix is what makes it the cheapest finding here and, arguably, the most valuable one.

Who reads a contingency funding plan, and what do they do with it?

Four people pick this document up, and they pick it up for four different reasons. Knowing the four reasons matters before writing one.

The treasurer reads it as a running order. Devendra Achar does not read it for insight; he reads it the way a pilot reads a checklist, for what to do first and who to telephone. He needs the actions in the sequence he would actually take them, the fastest ones at the top, and nothing on the list requiring a phone call he has not already had. The single most useful thing a treasurer can do with a plan is telephone the counterparties in it on a quiet Tuesday. An action nobody has ever discussed with the other side is a hypothesis rather than a step.

The board committee reads it as a headline and a horizon. The committee wants one number: how long the bank has. The single number is precisely why the difference between 96.0 days and 63.8 days matters so much more than it looks. A committee is not going to re-derive this. A committee will take the number in front of it, and the number in front of it should be the tested one, with the untested one shown beside it and labelled, rather than the other way round.

A credit analyst at another institution, looking at Vindhya Commercial Bank Limited as a counterparty rather than as an employer, reads it for the fourth column and nothing else. The analyst asks what proportion of the stated actions depends on somebody agreeing at the moment of stress. At this bank, five of the six. The ratio tells an outsider more about funding resilience than the headline total does. Everybody publishes the headline total and almost nobody computes the ratio.

And a household reads the same document without knowing it. The savings account is F1: available this afternoon, nobody's permission required. The gold is F2: two days and a haircut. The cousin who has said he would help is F4's uncommitted portion, and whether he is committed or uncommitted is a conversation nobody in the household has ever had explicitly. The plot of land is F5: real, large, and four months away from a bill that is due next week. Cancelling the wedding caterer is F6: it saves a fixed amount every week, it needs nobody's agreement, and it is worth relatively little in the first frantic week and a great deal by the second month. The mechanism does not change with the number of zeroes.

What could be changed, if there were no money to change anything with?

The useful exercise is deliberately constrained: assume the buffer cannot be raised by a rupee. Four changes are still available and all four are sentences. Say whether a red indicator counts toward a two-amber trigger. Mark every action as committed or uncommitted on its face, so nobody has to remember which is which at eight in the morning. Convert every lead time into calendar days and print the survival horizon beside it in the same units. And state for each action which of the two stress shapes it survives. None of the four raises a rupee. All four change the plan's worth to the person holding it.