Contagion: How One Institution's Stress Reaches Another
Contagion is a loss, or a demand for cash, arriving at a second institution because of what happened at a first, along a connection that already existed. Five connections carry it: a claim held on the other, the same kind of asset held by both, the same funders standing behind both, an outsider who cannot tell the two apart, and a service nobody else can supply by tomorrow.
Everything in this guide rests on one property of a financial claim, and it is worth stating before anything else. A claim is one thing written on two sets of books. Whatever one institution counts as an asset, another institution carries as a liability, and the two entries are the same rupee seen from opposite sides. The two-sided nature of a claim is why stress does not have to travel in the ordinary sense at all. The moment a borrower's position changes, the lender's asset has already changed. No news has to reach anybody for that to be true, and nobody has to react to anything.
Almost every wrong mental picture on this subject comes from imagining stress as something moving along a wire between two buildings. Hold on to the two-sided claim instead. Some of it does move. Most of what matters was already in both places.
What does contagion actually mean, and what is it not?
The loose sense of the word has to go first. Left standing, it will spoil everything that follows. In ordinary conversation contagion means worry spreading. Somebody hears something, tells somebody else, and a mood travels. Contagion in this subject is not a mood, and a reader who keeps the loose meaning will look for a mood and miss the mechanism.
Contagion is a loss, or a demand for cash, arriving at a second balance sheet because of what happened at a first. Contagion is a quantity, not a feeling. A loss has a size, it has a direction, and it arrives along something that can be named and drawn. A mood cannot be divided by anything. A loss can be divided by the net worthWhat is left of everything an institution holds once everything it owes has been taken off. It belongs to the people who put up the institution's own funds. of the institution it lands on, and that division is most of the subject.
A named episode is the wrong unit for this subject. The channel stands in its place, worked properly, with the arithmetic shown. A channel is not a smaller thing than a story. A channel is the part that transfers: a story records what happened once, and a channel shows what to look for every time. The story was never the teachable part.
How Contagion Can Spread Through Financial Institutions: which five connections carry it?
The connections are set out in one place before any of them is worked. The shape of the list is then visible, rather than each one arriving cold. Five connections carry stress from one institution to another, and all five are separated here because each stops in a different way. A claim held by one institution on the other. The same kind of asset held by both and priced by the same market. The same people supplying funding to both. An outsider who cannot tell the two apart and therefore acts on both. And a service one institution supplies that the other cannot replace by tomorrow.
Three questions apply to every connection: does it need a link between the two institutions at all, how fast does it carry, and what would stop it. The three questions are a routine rather than a fact, and a routine is worth more than a list of names. Run on any pair of institutions, in any market, they get further than a memorised list of five names.
The answers are not what most readers expect. Only two of the five need a link between the two institutions. Three of them work on institutions that have never dealt with one another, have no claim in either direction and share no funders in the ordinary sense. A person checking for connections almost always checks a list of who owes whom, and a list of who owes whom will never show three of the five. The finding is worth sitting with.
Why does a claim on another institution put one loss in two places at once?
Take the direct connection first. The claim one institution holds on another is the easiest to state and the easiest to underrate. When one institution lends to another, the lender records an asset and the borrower records a liability. The asset and the liability are not two facts. The asset and the liability are one fact, written twice, in two ledgers, on the same day. A counterpartyThe other party to a dealing. From either side, the counterparty is the one who has to perform. is not somewhere the trouble might go later; it is somewhere the same line is already sitting.
A loss on the borrower's side is simultaneously a loss on the lender's side, and the word to hold on to is simultaneously. Not afterwards, not once it is discovered, not once anybody reacts. The second entry was always there. The lender's asset is defined by the borrower's ability to pay and by nothing else. Nothing has to be carried anywhere for that asset to be worth less.
Here is the everyday version, and it is exact rather than an analogy. A supplier delivers stock to a shop on credit and has not been paid. The shop runs into difficulty on Tuesday. The supplier does not learn about the shop's difficulty on Friday and then acquire a problem. The supplier's unpaid bill is the shop's difficulty, written on the other side of the same dealing. Friday is only the day somebody found out.
A borrower's position worsens on Monday and its lender hears about it on Friday. On which day did the lender's asset change?
How can two institutions with no dealings between them still be hit together?
This is the connection that needs no link at all, and it is the one readers miss because it looks like coincidence when it is nothing of the kind. Two institutions have no claim on each other in either direction. The two share no owner. The two have never done any business together. Both of them hold the same kind of asset, and both of them are valued off the same price on the same morning.
When that price moves, both institutions move with it, and neither has touched the other. There is no messenger, no arrangement and no dealing of any kind. The connection is the price, and the price belongs to neither of them.
The absence of a link changes the way connections have to be looked for. There is nothing to inspect, and that makes the shared holding the hardest channel to see coming. A register of who owes whom is a real document with real entries, and it will come back empty here every single time. The link is not on anybody's books, so no amount of care in reading books will find it.
The everyday version. A row of workshops sits along one lane. None of them has any arrangement with the others, none supplies the others, and they compete if anything. The lane is dug up for a week. Every workshop on it loses the same mornings, and not one of them did anything to any of the others.
Two institutions have no claim on each other, no shared owner, no dealings of any kind and no funder in common. Which connection can still reach both?
How does an institution lose its funding without a single figure on its own books changing?
The third connection travels faster than either of the first two, and it works in a way that catches people out because they keep looking in the wrong place for the cause. The people who put money into one institution usually put money into others of the same kind. The funders are not choosing between two names; they are choosing whether to be in that kind of institution at all. If they decide to stop supplying that kind, every institution in it is short at the same time.
An institution can lose its funding without a single figure on its own books having changed. The absence of any internal cause is the whole reason this channel is separated from the first two, and it is worth reading twice. Nothing inside went wrong. No borrower failed, no holding fell, no cost rose. The decision was taken outside the institution, about a category rather than about it. The cause was never inside, so looking inside for it will find nothing to look at.
The everyday version. Every stall in a night market is stocked on credit by the same wholesaler. The wholesaler decides to stop giving credit to night-market stalls this season. Every stall is short on the same evening, and not one of them has done anything different from what it did last season.
Rukmini Finance Limited, an invented finance company, is funded this way rather than by taking deposits, and that is the whole of the difference between it and a bank. Its borrowings are Rs 14,400 crore against assets of Rs 18,000 crore, or 80.0 per cent of its assets, and that is the size of the position this connection acts on. Which lenders supplied that Rs 14,400 crore is not something a balance sheet total shows, so no lender is named.
An institution loses its funding and not one figure on its own books has changed. Is that possible?
What is somebody outside an institution actually able to do?
Treatments of this subject usually start insulting somebody at exactly this point, and the insult is also wrong.
A person outside an institution can see what that institution publishes and nothing else. Published information is the whole of an outsider's material. If two institutions look alike from outside, and something makes an outsider cautious about one of them, they have no basis in anything they can see for excluding the other. So they act on both.
Acting on both is not panic and it is not ignorance; it is the only reading the available information supports. Refuse the moral here. The moral is what stops the analysis. The less an outsider can see, the wider the group they are forced to act on, and the grouping is the connection. The remedy for this channel is therefore information rather than reassurance. Reassurance is what gets offered, and information is what would work.
The everyday version. Two contractors bid for the same work, and all that is available on each is the same one-sheet profile. If something makes the client cautious about that trade, nothing in front of them separates the two, so they go slower with both. Nobody has panicked. One sheet of information supports exactly one reading, and that is the reading taken.
Somebody outside two similar institutions treats both the same way. Which reading does that support?
What happens when a service cannot be replaced by tomorrow?
The fifth connection is short to state and is pointed onward deliberately. Where one institution depends on something another institution does, and nobody else could do that thing quickly, the dependence is a connection whether or not a single rupee is owed in either direction. SubstitutabilityWhether anybody else could do tomorrow what an institution does today. Where nobody could, a dependence exists whether or not any money is owed. is the property that decides how much of a connection it is.
The test is two questions long: ask what would go dark if this went dark, and then ask how long it would take somebody else to switch it back on. If the answer to the second question is measured in hours, the dependence is mild. If it is measured in months, the dependence is a connection as real as a loan, and it will not appear anywhere in a list of amounts owed.
The everyday version is a goods lift serving a whole office building. Nobody in the building owes the lift company anything. Every floor stops moving anything heavy on the day it breaks, and stays stopped until somebody with the right parts arrives. How central an institution is, and what makes it central enough for that answer to matter across a whole system, is covered separately.
The same proportional fall in the value of assets arrives at one institution with net worth at 10.0 per cent of its assets and one at 20.0 per cent of its assets. Commit before reading on: how do the two readings compare?
Why does the same proportional fall land differently on two balance sheets?
This is the spine of the argument: one illustration, held completely still, meeting two different balance sheets. The illustration is a fall of 5.00 per cent in the value of assets. The fall does not change between the two institutions. Nothing about it is bigger at one than at the other.
Suvarna Commercial Bank Limited, an invented bank, has total assets of Rs 2,40,000 crore and net worth of Rs 24,000 crore. Net worth is 10.0 per cent of assets, so its assets are 10.0 times its equity. A fall of 5.00 per cent in the value of assets is Rs 12,000 crore, and Rs 12,000 crore against net worth of Rs 24,000 crore is 50.0 per cent of that absorbing layer.
Rukmini Finance Limited has assets of Rs 18,000 crore and net worth of Rs 3,600 crore. Net worth is 20.0 per cent of assets, so its assets are 5.0 times its equity. The same fall of 5.00 per cent in the value of assets is Rs 900 crore, and Rs 900 crore against net worth of Rs 3,600 crore is 25.0 per cent of that absorbing layer.
| The same fall, worked twice | Suvarna Commercial Bank | Rukmini Finance |
|---|---|---|
| Total assets | Rs 2,40,000 crore | Rs 18,000 crore |
| Net worth, and its share of assets | Rs 24,000 crore, 10.0 per cent | Rs 3,600 crore, 20.0 per cent |
| Assets as a multiple of equity | 10.0 times | 5.0 times |
| A fall of 5.00 per cent of assets | Rs 12,000 crore | Rs 900 crore |
| That fall, as a share of net worth | 50.0 per cent | 25.0 per cent |
The arriving fall is identical and the two readings are not. The difference lies in what each fall met. The sentence to carry away is worth more than the five names above it. The link decides the reading, not the shock. And the two readings stand in exactly the ratio of the two leverageWhat an institution holds, written as a multiple of its own funds. Assets of ten rupees standing on one rupee of own funds is ten times. figures: 50.0 per cent against 25.0 per cent is two to one, and 10.0 times equity against 5.0 times equity is two to one as well. The matching ratio is not a coincidence and not a law of nature either; it follows directly from dividing one number by two different denominators.
The same one litre of water is poured into two vessels. In a two litre jug it comes halfway up. In a four litre bucket it comes a quarter of the way up. Nothing about the litre changed between the two pours, and anybody calling the water more serious in the jug has described the jug.
The fall of 5.00 per cent is an assumption, and the status of an assumption has to be stated plainly. Neither institution has had a fall of 5.00 per cent in the value of its assets. The fall is held still so that one figure can be divided twice. A division of that kind shows what leverage means, and it reports nothing about either institution.
Hold the fall completely still and move the layer instead
One control, and it does not move the trouble. The fall in the value of assets is held at 5.00 per cent at every single setting, and what moves is the absorbing layer the fall arrives at, written as a share of assets. Watch the upper bar. Its length never changes at any setting.
20.0 per cent of assets is the layer, and 25.0 per cent of that layer is consumed
With the absorbing layer set at 20.0 per cent of assets, a fall of 5.00 per cent in the value of assets consumes 25.0 per cent of the layer. The fall is held at 5.00 per cent of assets at every setting, and the only thing that moves is the layer it arrives at.
Educational illustration. Invented institutions and invented arithmetic. The fall in the value of assets is held at 5.00 per cent at every setting, and holding it still is the assumption doing all the work on this screen. One balance sheet meets one arriving figure, so nothing is passed on to anybody else at any setting. No institution is described as failing at any setting, and no likelihood of anything is stated. The two marks are the reported net worth of Suvarna Commercial Bank Limited and of Rukmini Finance Limited as a share of their own assets, both invented, and no real institution is measured anywhere on this screen.
Why are two institutions funded the same way not levered the same way?
The block above opens a hole and this one closes it. Deposits are Rs 1,92,000 crore of Suvarna Commercial Bank Limited's Rs 2,40,000 crore of assets, or 80.0 per cent of assets. Borrowings are Rs 14,400 crore of Rukmini Finance Limited's Rs 18,000 crore of assets, or 80.0 per cent of assets as well, identical to the decimal. Both institutions are funded to the same share by somebody other than their own owners.
And yet net worth is 10.0 per cent of assets at the first and 20.0 per cent of assets at the second, so assets are 10.0 times equity at one and 5.0 times equity at the other. Two institutions with the same funded share and twice the leverage at one of them. The pair looks like a contradiction until what sits underneath the top band is taken into account.
Suvarna Commercial Bank Limited carries a third layer of liabilities, sitting beneath its deposits and above its net worth, and Rukmini Finance Limited does not. At the finance company, borrowings of Rs 14,400 crore plus net worth of Rs 3,600 crore come to Rs 18,000 crore, the whole of its assets. There is nothing else, and that is why an 80.0 per cent funded share leaves 20.0 per cent as net worth. At the bank the two named lines do not reach the total, so something else is there.
The next point has to be stated plainly. Neither balance sheet names that third layer or says what is in it, so no amount and no share can be given for it. The gap yields a habit rather than a figure. When a lender's deposits and net worth do not add up to its assets, the gap is a real layer of somebody's claims, and finding out what sits there is the first question to ask of any lender's balance sheet.
Deposits are 80.0 per cent of Suvarna Commercial Bank Limited's assets and borrowings are 80.0 per cent of Rukmini Finance Limited's. Are the two levered the same way?
A lender short of cash sells what it can sell quickly. Which institutions feel that?
How Bank Liquidity Stress Can Spread Through the System: which channel switches on second?
Everything above has taken the connections one at a time. Here is the one case where two of them run at once, and it is worth watching as a sequence rather than as a description. The second connection switches on partway through and reaches somewhere the first one cannot.
A lender is short of cash today. Its first move is to stop supplying cash to other lenders. The funding connection is now acting directly on the institutions it deals with. Still short, its second move is to sell whatever can be sold quickly. The selling sets a price, and the price is lower than it was. Every institution holding the same kind of thing is now valued off that lower price. The common holding connection is acting on institutions that never dealt with the first lender at all.
The second connection reaches institutions the first one cannot, so the circle widens without the original trouble widening at all. Nothing worse has happened to the lender that started short. The lender simply did the ordinary thing a short institution does, and the ordinary thing set a price. The widening is the property that makes the pair worse than either connection on its own, and it is why a defence that works against the first connection does nothing whatever about the second.
Which part of a balance sheet this sequence can act on quickly is worth being precise about. Suvarna Commercial Bank Limited holds Rs 60,000 crore of investments against Rs 1,44,000 crore of advancesThe loans a lender has already made. They run for the term that was agreed, so a lender cannot call them all back on the morning it wants cash.. The investments are the part that can be sold in a morning and therefore the part that can set a price. The advances are not: they run for the term agreed with each borrower, and a lender cannot recall them on the day it wants money. So the selling happens in one part of the balance sheet and the markingRestating what a holding is worth using the price the market is showing today, rather than the price that was paid for it when it was bought. that follows lands on everybody else who holds that same part.
The everyday version. A shopkeeper who needs cash this week first stops extending credit to the other shopkeepers, and then sells stock at whatever it will fetch this afternoon. The second move sets a price that every shop holding the same stock is now judged against, including shops that have never dealt with them.
What has to be true for stress to stop at the first institution?
The question worth carrying away is the one almost nobody asks. The interesting-sounding question is why the trouble started. Ask instead what would have had to be different for the next institution to be untouched.
There are three answers, and the reason there are three rather than a long list is that between them they exhaust the possibilities. Either the absorbing layer at the next institution is larger than what arrives, so the loss is met there and nothing is passed on. Or the connection is not there at all, so nothing arrives in the first place. Or somebody stands between the two and takes what would otherwise arrive. Something arrives and is met, nothing arrives, or somebody else takes it. There is no fourth case hiding behind those three.
Asking why it started reveals something about one institution, and asking what would have had to be different for the next one to be untouched reveals something about the connection between them. Only the second question is about the system. The second question is the whole reason the habit is worth forming, and a reader who leaves with nothing but that habit has got the valuable part.
The third answer is a large subject in its own right. Who stands between two institutions, on what terms, and what it costs to be that party, is covered separately and is not previewed here.
The question is whether stress stops at the first institution. Which of these settles it?
The failure: reading the rupees and calling it the pressure
The failure on this subject is reading the rupee amount instead of the share of the absorbing layer, and then concluding that the larger institution took the harder hit. The failure is worth stating at length because the wrong reading is the one that is printed and the right one has to be worked out.
Put the two figures from the worked illustration side by side. Rs 12,000 crore at Suvarna Commercial Bank Limited and Rs 900 crore at Rukmini Finance Limited. The first amount is more than thirteen times the second, and a reader who stops there concludes that the bank is the one under pressure and the finance company got off lightly. Read each against what it has to be absorbed with and the ordering reverses: 50.0 per cent of net worth at the bank against 25.0 per cent of net worth at the finance company. The bank's reading is twice the finance company's rather than thirteen times it, and the comparison that matters points the other way from the one that was printed.
Who makes this reading: everybody, at first, and there is nothing careless about it. The rupee amount is the figure in print and the share of the layer has to be computed by hand. The cost: attention spent on the institution with the biggest number rather than on the one with the thinnest layer. The biggest number is the wrong place to be looking. The correction is a single division, and it is the same division every time: what arrives divided by what is there to absorb it, with the base named in the same breath as the answer.
How does somebody reading two lenders actually use this?
The ten minutes an analyst spends before forming any view at all
The routine runs in one direction and produces questions rather than verdicts. First, find net worth and write it beside total assets, then divide one by the other and say the answer out loud with its base attached: net worth is such and such per cent of assets. The single division turns a balance sheet into a statement about how much can go wrong before somebody other than the owners is affected.
Second, whatever fall is under consideration is applied to assets and the result divided by net worth, rather than read off as a rupee figure. The rupee figure gives the size of the institution and the share of the layer gives the size of the pressure, and confusing those two is the single most common error on this whole subject. The same proportional fall is applied to both institutions being compared. The only thing varying between them is then what the fall met.
Third, list the connections rather than the amounts. Who holds a claim on this institution, what does it hold that others also hold, who funds it and would those funders leave the whole category at once, would an outsider be able to tell it apart from its neighbours, and what does it supply that nobody could replace quickly. Two of those five are answerable from a balance sheet and three of them are not, and that is why the exercise takes ten minutes rather than one.
Fourth, and this is where the discipline lies, the routine stops. The routine shows where the pressure would land and how it could arrive. The routine does not establish that either institution is safer, better run or more likely to be in difficulty. A wider absorbing layer is not virtue and a narrower one is not recklessness; they are different businesses making different arrangements, and neither has been given an outcome to be judged by.
Who sets the limits on any of this?
Six figures behind the connections above are not open to anybody's choosing. An authority sets each one, each is revised, and each is the sort of figure a reader would copy into a note and rely on a year later. Each one is drawn here as a labelled row with the authority printed inside it and nothing in the value column at all.
The list carries information even while it is empty. Every connection worked above has a limit sitting on top of it somewhere. A limit on how much of one institution's paperA written promise to pay, issued by one institution and held by another, which the holder can usually sell on to somebody else. another may hold acts on the claim connection. A limit on how much of one kind of thing an institution may hold acts on the common holding connection. A requirement to hold liquidity in advance acts on the funding connection. The rows teach which limit exists and who sets it even while blank, and that is the part that does not go out of date.
Six requirements named here, each with its value left blank
| What is set | The value here | Who sets it |
|---|---|---|
| The liquidity an institution holds in advance, and what counts towards it | Left blank | Reserve Bank of India at rbi.org.in |
| The limits on how much one institution may hold on another | Left blank | Reserve Bank of India at rbi.org.in |
| The capital an institution holds against its assets, and every buffer above it | Left blank | Reserve Bank of India at rbi.org.in |
| How much of one kind of thing an institution may hold | Left blank | Reserve Bank of India at rbi.org.in |
| What a depositor is covered for | Left blank | Deposit Insurance and Credit Guarantee Corporation at dicgc.org.in |
| The order in which a clearing arrangement's resources are used when a member fails | Left blank | Securities and Exchange Board of India (SEBI) at sebi.gov.in |
Each of these is set by the authority printed inside its row and each of them moves. A written-out value would be a wrong figure rather than a stale one, and a wrong figure is the worse thing to hand anybody. The sheet is meant to be taken to the addresses named and the middle column filled in there.
Six blanks, and the address of each answer
| Where the blank requirements are set | Who settles it | Site | Address checked |
|---|---|---|---|
| The liquidity an institution holds in advance, and what counts towards it | Reserve Bank of India | rbi.org.in | 25 August 2026 |
| The limits on how much one institution may hold on another | Reserve Bank of India | rbi.org.in | 25 August 2026 |
| The capital an institution holds against its assets, and every buffer above it | Reserve Bank of India | rbi.org.in | 25 August 2026 |
| How much of one kind of thing an institution may hold | Reserve Bank of India | rbi.org.in | 25 August 2026 |
| What a depositor is covered for | Deposit Insurance and Credit Guarantee Corporation | dicgc.org.in | 25 August 2026 |
| The order in which a clearing arrangement's resources are used when a member fails | SEBI | sebi.gov.in | 25 August 2026 |
Suvarna Commercial Bank Limited and Rukmini Finance Limited are invented.
Educational material. Not advice on any investment, tax, budget or market position.