Systemic Importance: What Makes an Institution Central

An institution matters to the system by what stops if it stops, not by how large it is. Trouble spreads three ways: other parties hold claims on it, others depend on a service only it provides, or others hold the same things it holds. A small party with no substitute can matter more than a large one that several others could replace.

Every requirement in this subject has an owner, and the owner is an authority rather than an institution. Whether an institution is identified as central to the system, what extra capital that identification carries, how exposed one party may be to another, the margin and position limits a member works under, the order in which a clearing corporation's resources are used, the cover a depositor has and the solvency margin an insurer holds are every one of them set by an authority and every one of them revised. Each appears below as a labelled row, with the authority and the site it publishes on standing where the value would otherwise be. The authority outlasts the value it sets. A requirement copied out anywhere else stops being right the day the authority moves it.

Reading one institution's condition is covered separately: a single balance sheet taken apart to ask whether the party holding it is sound. Systemic importance asks a different question, and it is the one an authority, a treasury desk and a careful analyst all end up asking: if this party got into trouble, how many other people would find out?

Soundness and importance are not the same question, and the second is not a bigger version of the first. A party can be in perfect condition and still be the thing everybody depends on. A party can be in poor condition and matter to almost nobody outside its own building. Soundness is a reading of what is inside an institution, and importance to the system is a reading of what runs between it and everybody else.

The everyday version is exact rather than merely illustrative, so start there. Picture a street with fourteen shops selling roughly the same groceries. One of them closes on a Tuesday. Thirteen others are open and the customers simply walk a little further, so nobody on that street goes without rice on Tuesday evening. Now picture the single bridge into the same town. The bridge is far smaller than the fourteen shops put together, it employs nobody, and it holds no stock at all. Close it on a Tuesday and nothing comes in, so every shop on that street has an empty shelf by Thursday. The bridge is not important because it is large. The bridge is important because there is exactly one of it and everything passes over it.

What has to be true for one party's trouble to become everybody's?

Three conditions, and each of the three is tested in turn below. Other parties have to hold claims on the party in trouble. Its loss then lands directly on somebody else's books. Or other parties have to depend on something only it does. Its stopping then stops them. Or other parties have to hold the same things it holds. Whatever moved against it then moves against them at the same moment.

Not one of those three is a statement about how large the institution is. Size does not appear in any of them, not even as a hidden variable. The first is about who lent to whom. The second is about whether a substitute exists. The third is about what sits on several balance sheets at once. A very large party can satisfy none of the three, and a very small one can satisfy all three at once.

Why is size the weakest of the four tests?

Because size measures what is inside one institution and says nothing at all about what passes through it, who depends on it, or how quickly somebody else could do the same job. Size is the most available number, and that is exactly why everybody reaches for it. Every party publishes a total. Almost nobody publishes a map of who would be stuck if they stopped.

Work it against the parties used in this guide. Suvarna Commercial Bank Limited carries total assets of Rs 2,40,000 crore, the largest figure anywhere in this record. Rukmini Finance Limited carries assets under managementThe total of what a lender or a manager has out working, whether that is loans it has made or holdings it looks after for other people. It is the figure the business is run against. of Rs 18,000 crore. The clearing corporation of Kaveri Stock Exchange Limited is a separate entity from the exchange, and it holds margins of Rs 11,000 crore against a settlement guarantee fundA pool a clearing corporation keeps aside so that a trade still completes when one member cannot pay. It exists to be used, not to be admired. of Rs 2,750 crore, so those margins are 4.00 times that fund. And Setu Payments Limited has no balance sheet in this record at all: what it has is a flow of 1,200 crore payments a year carrying Rs 3,60,000 crore.

A party with no balance sheet at all is already the argument. One of the four parties cannot be placed on a ranking by size, not because the number is missing by accident but because what makes that party matter is not a stock of anything. It carries. A ranking that cannot even seat one of its four candidates is a poor instrument for deciding which of them deserves attention.

There is a second reason to distrust a ranking built on one number, and this record was built to make it unmissable. Suvarna Commercial Bank turns a profit of Rs 2,250 crore on net worthWhat is left of an institution for its shareholders once every liability is met. It is the cushion that absorbs a loss before anybody else's money is touched. of Rs 24,000 crore, a return on equity of 9.375 per cent for the year. Rukmini Finance turns Rs 337.50 crore on net worth of Rs 3,600 crore, again a return on equity of 9.375 per cent for the year. The same number, at two businesses that are not remotely alike.

Take each apart and the reason appears at once. The bank earns 0.9375 per cent on its Rs 2,40,000 crore of assets and holds assets 10.0 times its net worth. The finance company earns 1.875 per cent on its Rs 18,000 crore of assets and holds assets 5.0 times its net worth. Half the return on every rupee of assets, twice the leverage, identical answer. Two businesses that are not remotely alike can land on the same headline ratio, so one ratio hides the whole difference between them. Neither route is better than the other, and a return on equity is not the number that would settle it.

One warning about those figures, and it has caught readers before. Multiplying the rounded limbs, 0.94 per cent by 10.0 times, gives 9.40 per cent rather than 9.38 per cent. The mismatch looks like an arithmetic mistake and is not one. The product only ties at the exact limbs, 0.9375 and 1.875, and the reported figure is the exact one rounded for reading. The exact limb is worth carrying beside the rounded one wherever the decomposition is written out.

What does it mean for a party to have no substitute?

Substitutability is the strongest of the four tests, and one question with a time limit attached settles it. Ask what stops if this party stops. Then ask who else could do that same job by tomorrow morning. The time limit is not decoration. Almost everything has a substitute eventually; the question is whether the substitute arrives before the damage does.

Run it on a lender. If Rukmini Finance stops lending tomorrow, new credit to whoever it was lending to stops. Other lenders can take that business, and over the following months some of them will, imperfectly and at different prices, and some borrowers will not be picked up at all. The loss of that credit is a real cost and it lands on real people, but it lands slowly and it lands unevenly. Crucially, nobody's payment fails on Tuesday because Rukmini Finance stopped on Monday.

Now run it on the clearing corporation of Kaveri Stock Exchange Limited. The clearing corporation is a separate entity from the exchange, and it exists because somebody has to stand between the two sides of every matched trade and become the counterpartyThe party on the other side of a claim or a trade, the one who has to perform for the other to be paid. Every trade has two of them, and each is the other's. to each of them. If it stops, every trade matched on that exchange stops the same day. Ask who else could do that by tomorrow and the honest answer is nobody. There is no second one waiting. Building one needs not capital so much as everybody's agreement to route through it, and that agreement takes years rather than days.

Substitutability comes to this: a lender that stops is replaced slowly and imperfectly, and a party every trade must pass through is replaced by nobody at all. The two answers are not different in degree. The difference is one of kind, and no balance sheet total shows it.

How do claims between parties carry trouble from one to the next?

A claim is the most direct route of the three, and it needs nobody's confidence to be shaken and nobody's opinion to change. Institutions hold claims on one another as a matter of ordinary business, and a claim is one party's asset sitting on top of another party's liability. When the party underneath cannot pay, the loss does not spread by rumour. The loss arrives on the other party's books as an entry.

Rukmini Finance makes this concrete because of what it does not do. Rukmini Finance takes no deposits at all, and taking no deposits is the whole difference between it and a bank. Funding comes from the market instead: borrowings of Rs 14,400 crore set beside Rs 3,600 crore of net worth, so its borrowings are 4.0 times that net worth and its assets under management of Rs 18,000 crore are 5.0 times that net worth. Every rupee of that Rs 14,400 crore is somebody else's financial asset, and often that somebody is another institution.

Now notice the direction of travel. The claim runs both ways, and most readers only see one of them. If Rukmini Finance runs into trouble, the loss lands on the parties that lent it the money. The loss landing on the lenders is the obvious direction. The second direction is that if those parties simply decide to stop lending, Rukmini Finance has a serious problem while every single loan on its own books is still being repaid on time. Nothing has to be wrong with its borrowers for it to be in difficulty. A party funded entirely in the market has a different vulnerability from a party funded by deposits, not a smaller one.

The same arithmetic quietly names an absence in the bank's figures. Suvarna Commercial Bank has assets of Rs 2,40,000 crore, deposits of Rs 1,92,000 crore and net worth of Rs 24,000 crore. Deposits and net worth account for Rs 2,16,000 crore of the total, and the rest is never broken down. Some of the bank's funding is therefore neither deposits nor its own capital, and the record does not say what it is.

How can parties fail together without being connected at all?

Here is the test readers miss, and it is missed because it has no arrow to follow. Two institutions with not a single claim between them can still be in trouble at the same moment, simply by holding the same kind of thing. Nothing passes from one to the other. Nothing needs to.

Suvarna Commercial Bank holds Rs 60,000 crore of market instruments, or 25.0 per cent of its total assets of Rs 2,40,000 crore. If other lenders hold instruments of the same kind, and there is every ordinary reason why they would, then a single repricing of those instruments reaches every holder on the same day. Each of them records the loss in the same session. Not one of them lent anything to any of the others.

Common exposure is a separate test rather than a version of interconnection, and the difference is not a subtlety: with a claim the path can be traced and the exposure counted, and with common holdings there is no path to trace. A supervisor mapping who owes what to whom would find nothing between these two lenders at all, and would still be looking at two parties that move together.

The household version is a street of ten shops in one shopping arcade. None of them has lent a paisa to any other. The arcade closes for repairs and all ten have a bad month simultaneously, not because any shop affected any other shop, but because all ten were standing in the same place. Common exposure is standing in the same place.

Why does the speed of a failure matter as much as its size?

Because how fast a failure is felt decides how much anybody can do about it. The same word covers events measured in hours and events measured in reporting periods, and treating those as the same kind of thing is how supervision ends up prepared for the wrong one.

Take the two ends of the range from this record. Setu Payments moves 1,200 crore transactions in the stated year carrying Rs 3,60,000 crore, or Rs 300.00/- on the average payment. The average payment is the teaching point. A system whose typical payment is three hundred rupees is doing something quite different from one whose typical payment is three lakh: it is carrying a very large number of small, ordinary, immediate payments. If it stops, the effect is measured in hours, and what is felt is not a hole in anybody's balance sheet but a shop that cannot take money and a fare that cannot be paid.

At the other end, a lender's deterioration is measured in reporting periods. Loans go bad over quarters. The figures move slowly enough that people can argue about them. There is time to raise capital, time to sell assets, time for an authority to act, and time for the whole thing to be discussed at length before anything irreversible happens.

The finding is uncomfortable: the parties felt within a day are usually the ones that carry and settle rather than the ones that judge, so attention aimed only at the largest balance sheets is aimed at the slow half of the system. The slow half matters. The slow half is simply the half where there is time.

What follows once a party is identified as central to the system?

Identification exists as a formal thing. An authority does it, rather than the institution deciding it is important or the market deciding it for them, and consequences follow. The shape of those consequences can be set out in full, and the values attached to them belong to the authorities named below.

There are four classes of consequence, and naming the classes is genuinely useful even with every value left out. More capital than would otherwise apply. A party whose trouble reaches everybody should absorb more of its own loss before anybody else's money is touched. Closer supervision too, and the reason for watching is no longer the ordinary one. Planning, in advance, for how the institution could be wound down without stopping the service everybody depends on it for. And limits on how exposed any party may be to any other, the interconnection test turned into a rule.

Every one of those four is set by an authority and every one of them is revised. How an institution is identified and what follows, the extra capital that comes with it, how exposed one institution may be to another and the route by which a failing institution is resolvedThe formal route by which a failing institution is dealt with, rather than simply being left to stop. Who decides, and how, is set down in advance. all belong to the Reserve Bank of India at rbi.org.in. The marginMoney or securities a member places with a clearing corporation against positions it has open, so that something is already there if it cannot pay. and position limits capping what a member may carry, and the order in which a clearing corporation's resources are used when a member fails, belong to the Securities and Exchange Board of India (SEBI) at sebi.gov.in. Cover for a depositor belongs to the Deposit Insurance and Credit Guarantee Corporation at dicgc.org.in, and the solvency margin an insurer holds above its policy liabilities belongs to the Insurance Regulatory and Development Authority of India (IRDAI) at irdai.gov.in.

On the last of those there is one part that does not date, and it is the part worth learning. When a member of a clearing corporation fails, the losses are met in a stated order: a fixed sequence of resources, worked out and written down before anybody is in trouble. Every step in that order and every threshold inside it is set by SEBI at sebi.gov.in, and every one of them is revised. The reason for settling that sequence while there is still nothing to argue about does not move: everybody then knows where they stand before the difficulty arrives, and a sequence negotiated during the difficulty would protect nobody. The negotiation itself is the emergency.

What does the expectation of rescue change before anything has happened?

The tests so far have been about structure. The expectation of rescue is about behaviour, and it is the only one of the tests that is.

Suppose too much depends on a particular party for anyone to let it stop, and everybody comes to believe it will be kept going whatever happens. Notice what that belief does to the people funding it, and notice that it does it immediately, with nothing having occurred. Because the answers matter less to them, the funders ask fewer hard questions. Believing they are lending to something safer than its own condition would suggest, they accept less for lending to it. And so the party raises money more cheaply than a party nobody expects to be rescued, on identical figures.

The expectation is itself a subsidy: it is paid by everybody, it is received by whoever is expected to be rescued, and it operates whether or not any rescue ever happens. Arriving before anything happens is what makes it a cost rather than a contingency. A contingency sits in the future. The expectation is already in the price.

Now the honest counterweight. The alternative is letting a party stop when everything depends on it, and that cost lands squarely on people who had no part in the decision and no way to see it coming: the customer whose payment does not go through, the counterparty whose trade does not settle, the household whose money is somewhere it cannot reach. Neither option is free. Both are real, and the argument about which is worse is genuinely old. The idea of a lender of last resortThe idea that somewhere in the system there is a party that will lend when nobody else will, so that a shortage of cash does not by itself finish an otherwise sound institution. belongs to Bagehot, Lombard Street, 1873, and the idea itself is set out separately under the lender of last resort.

What happens when four parties are put through the same four questions?

Here is the run, written out so that the test can be seen doing the work rather than its conclusion simply being handed over. Every figure below is that party's own, and no run turns on a failure that happened, a cycle or a second period. Each run asks what would follow if a party stopped, and asks it while the party is still standing. The question is a counterfactualA question about what would follow if something were removed, asked while it is still there. It needs no event to have occurred and produces no prediction that one will. throughout.

Ranked by the figures this record supplies, the order runs Suvarna Commercial Bank, then Rukmini Finance, then the clearing corporation, and Setu Payments cannot be seated on that ranking at all. Ranked by what stops and who could replace it, the order runs the clearing corporation, then Setu Payments, then the bank, then Rukmini Finance. Two of the four move from one end to the other. Both orders are arithmetically correct, and only one of them answers the question anybody actually cares about.

Notice also what the third column did. On three of the four rows it says something like slowly, imperfectly, over time. On one row it says nobody. The word nobody is the strongest reading anywhere in the grid, and it comes from a party whose own figures are a small fraction of the bank's. The answer to the second question outweighed every figure in the first column, and it did so without any of those figures changing.

Who actually runs this test, and on what?

Four kinds of reader use these questions, and they use them differently. Seeing how each one uses them stops the subject reading as an abstraction.

A treasury desk placing money overnight uses the fourth question and the third one together. Before it lends to a party for a night, it wants to know who else has lent to that party and what those lenders would do on a bad morning. A funder who leaves at nine tomorrow turns a manageable problem into an immediate one. Rukmini Finance is the shape of that concern: Rs 14,400 crore of borrowings against Rs 3,600 crore of net worth, all of it somebody's decision to keep lending.

An analyst covering an institution uses the first two questions to sort what she reads. A soundness reading tells her whether the party can absorb its own losses. The substitutability reading tells her what would happen around it if it could not. The second reading is a separate matter, and it changes how much anybody else's numbers are worth.

An authority uses all four to decide where to put attention, and this is where the speed test earns its place. Attention aimed at the largest balance sheets is aimed at the part of the system where there is time to act. The part where there is no time is the part that carries and settles, and that part is often small on paper. Four questions are run rather than one because they disagree, and where they disagree is where the interesting party is.

And a household runs the same test without calling it anything. When a salary, savings, a loan repayment and rent all move through one arrangement, that arrangement matters to the household in a way nothing on its balance sheet would tell them. The arrangement is the bridge again. Nobody thinks about the bridge until the morning it is shut, and that invisibility is precisely what makes it worth thinking about beforehand.

Which four questions are worth keeping?

The whole subject collapses into a routine, and the routine is short enough to carry. For any party at all, these four are asked in this order.

1. What stops if it stops? Name the actual thing that would not happen. Not a loss, a thing: a payment, a settled trade, a loan being made.
2. Who else could do that by tomorrow? The time limit stays on. If the honest answer is nobody, the rest of the balance sheet can be left unread.
3. Who holds claims on it? Ask it in both directions, and remember that whoever is funding it can also walk away.
4. Who holds the same things it holds? The fourth question has no arrow to follow, and that is why people forget it.

The routine needs no failure to have happened, no series of past readings and no threshold, and that is exactly why it can be applied to a party met four minutes ago. Systemic importance is those four questions, fully assembled, and nothing more.