The Discount Window: Emergency Funding and Its Stigma
The discount window is a standing arrangement under which a bank borrows settlement balances from the central bank on its own initiative, pledging securities and paying a price set above what banks charge each other. Its stigma is that asking is visible: a bank seen at the window is assumed to have been turned away elsewhere, so the facility goes unused exactly when it is needed.
Where the system's cash actually sits, and what a facility does to it, is covered separately. So is the difference between a bank short of cash and a bank short of capital: two banks that have gone wrong in completely different ways. One instrument out of that first list is the strangest in the set, the only instrument on the shelf that is designed to work and known to go unused.
Where can a bank short of cash actually get it?
There are exactly two places, and everything in this guide comes out of the difference between them. A bank that will be short of settlement balancesMoney a bank holds at the central bank, in an account of its own. Payments between banks are settled by shifting it about, and it is a separate thing altogether from the notes sitting in a branch. by the close can go to other banks, holders of those balances already, or to the central bank, a party able to make more of them.
The two routes do very different things to the total. When another bank lends, cash that already exists changes hands: one account falls, another rises, and the amount in the system at the end of the afternoon is the amount that was there at the start. When the central bank lends, the total goes up. The party keeping the accounts is the one party whose own transaction can change what the accounts add to. Everything else about the window, the price and the stigma included, follows from the fact that the second route is always open and the first one is not.
The everyday version below does most of the work later, so hold on to it. A tea stall outside an office block runs out of change at four in the afternoon. The stall can ask its neighbour, a stall that may or may not have any to spare and may or may not feel like parting with it. Or it can walk to the bank on the corner. The bank is definitely open, definitely has change, and definitely charges for the trip. Both routes end with the same coins in the same tin.
What is the discount window, and who starts the transaction?
The window is a standing arrangement: a counter that is open whether or not anybody walks up to it. The bank decides that it wants cash, decides how much, arrives, hands over securities as a pledgeHanding over a security as security for a loan without selling it. The borrower keeps it on its own books and gets it back on repayment, which is what makes a pledge different from a sale., and takes settlement balances away. Nobody rang to offer.
The one feature that defines the window is that the bank starts it. Compare that with an operation the central bank runs, where the central bank decides that it will happen, decides how big it will be, decides when, and decides which institutions may take part. In an operation, a bank is a participant in somebody else's decision. At the window, a bank is the author of its own.
The bank's own initiative is doing two opposite jobs at once. A route that can be taken without waiting to be invited is the only kind of route that is any use when the trouble is today, and that is what makes the window useful at four o'clock on a bad afternoon. The same initiative is also the entire source of the stigma. A transaction somebody chose to enter says something about the person who chose it. A gift says something about the giver. A purchase says something about the buyer.
What is the one feature that separates the discount window from an operation the central bank runs?
Two places, one shortfall: what is the same and what is not?
The abstraction dissolves the moment a number is put on it. Suvarna Commercial Bank Limited, an invented bank with invented figures, finds that it will be short of Rs 9,600 crore of settlement balances by the close. The Rs 9,600 crore is a control setting, chosen so the arithmetic works out cleanly, and it reads two ways at once. Set beside deposits of Rs 1,92,000 crore, the shortfall comes to 5.0 per cent. Set beside total assets of Rs 2,40,000 crore, those very same rupees come to 4.0 per cent. Nothing moved but the base underneath the division.
Source one is the rest of the system: the banks that received the payments Suvarna made and are sitting on the proceeds. Those banks lend to each other in the interbank marketThe market in which banks lend settlement balances to one another, most of it for a single night. It is where a bank goes before it goes anywhere else, and it is covered separately., usually overnightLent today and repaid on the next working day. Most lending between banks is done this way, which is why a shortfall at four in the afternoon is a problem for this afternoon rather than for the year.. Source two is the central bank at the window.
The amount is identical, the effect on Suvarna's own balance sheet is identical, and the three things that differ are who is on the other side, what it costs and whether anybody finds out. The split used throughout is 65.0 per cent of the need from the rest of the system and 35.0 per cent at the window. Those shares are Rs 6,240 crore and Rs 3,360 crore, adding back to Rs 9,600 crore exactly. Changing the split changes the name on the other side of the transaction, and nothing else.
Why is the window deliberately priced above the market?
Cash at the window costs more than cash from another bank. The gap is not an accident of the market and it is not a fine. The gap is a design decision, and the reasoning behind it is worth following all the way through. The same reasoning makes the stigma inevitable.
Ask what would happen if the ordering were reversed. Suppose the window were the cheapest place to get settlement balances, as well as the one place that is always open. No bank would ever ring another bank again. Why would it? The interbank market would empty out, not through anybody's fault but through a straightforward comparison that every treasury desk would make in the same direction on the same morning. And the central bank would find itself the counterpartyThe other side of a transaction, the party a bank is exposed to if things go wrong. Which party a bank ends up facing, and how that exposure is managed, is covered separately. to every short position in the system, having replaced a market with a queue.
The price is set so that the window is the second call and not the first, and being the second call is exactly what the word last in last resort is doing. Read the phrase as an instruction about ordering rather than as a description of desperation and it stops being dramatic. Last means after the others, and the price is the mechanism that produces after.
The price itself, the period it runs over, and its relation to any rate the central bank announces all sit with the Reserve Bank of India. All of them shift, so the label and the address stand below in place of a figure.
Why is cash at the window priced above what banks charge each other?
What exactly is the stigma, and is it a reasonable inference?
Here is where most explanations go wrong. Such an explanation describes the stigma as a superstition, a hangover, an irrational fear that sensible people ought to get over. The superstition framing feels sophisticated and explains nothing. Why is the fear so stubborn among people who are otherwise extremely good at arithmetic? Superstition has no answer.
So build it up instead, one careful decision at a time. The window costs more than the market and is therefore used rarely, and the arrangement intends both. The market is cheaper and any treasury desk works down the list in price order, so a bank that turns up at the window has almost certainly tried the market first. If it is at the window, the market did not fill it. Anybody who learns that a particular bank borrowed at the window can therefore infer, with decent odds, that other banks would not lend to it, or would not lend it enough.
Every single one of those inferences is sensible, and the aggregate of all of them is a facility that nobody will touch. The lender who declines is being careful with its own position. The observer who infers is reasoning correctly from the evidence available. And the bank that stays away from the window rather than be seen there is being careful too. Being seen there costs it more than the cash is worth. Nobody in the story is behaving badly and the outcome is still bad.
The point is worth saying plainly. A bank that declines to lend has not done anything wrong. Neither has a bank that would rather not be seen asking. The problem is the shape of the chain rather than a failing in anybody standing in it.
An observer learns that a bank borrowed at the window and concludes that other banks would not lend to it. Is that inference reasonable?
Every bank holding a surplus of settlement balances decides to sit on it rather than lend it. What has happened to the quantity of cash in the system?
If the cash still exists, why is the system short?
The whole argument reaches its sharpest form in this question. The Rs 9,600 crore that left Suvarna Commercial Bank did not evaporate. The money was paid to somebody, so it arrived in an account at another bank and is sitting there right now. Somebody's shortfall is somebody's surplus, and the system's total did not shift by a single rupee when the payment settled.
So the cash that would fix the shortfall exists. The only question is whether it will move. And here is the half that readers miss: a bank that decides to hold on to its surplus rather than lend it has not created a single rupee of extra safety for the system. The bank has removed a rupee from circulation.
Hoarding does not make cash, it stops cash moving, and a system where every holder of a surplus decides to sit on it is a system that is short in every practical sense while holding exactly as much as it held yesterday. Nothing was destroyed. Nothing needs to be replaced. The quantity is identical and the usable amount is not. The distinction is worth carrying around. One problem is fixed by making more, and the other by getting what exists to move.
The household version is a street on a festival morning. Every household on it has change in a drawer. Word goes round that change is scarce, so every household decides to keep what it has rather than break a note for a neighbour. There is exactly as much change on that street as there was at dawn, and not one shop can give any. Nobody hid anything, nobody stole anything, and the street is out of change.
Rs 9,600 crore leaves one bank and that bank cannot replace it. Does the cash that would fix its shortfall exist somewhere?
What levers work against a stigma?
If the stigma is made of correct inferences rather than of shame, then the only thing that can move it is the quantity those inferences are drawn from. Three levers are available to whoever designs such a facility. Which of them should be pulled is not a question arithmetic answers.
The first lever is visibility: what is published about who borrowed and when, and how long afterwards it appears. The second is routine. A facility used by many participants on ordinary days carries less information when one of them uses it on a bad day. Ordinary use is noise, and noise dilutes a signal. The third is breadth: the wider the set of institutions that qualify and the wider the set of securities that may be pledged, the less any single appearance narrows down who is in difficulty.
The stigma is made of information rather than of shame, so every one of the three works by reducing how much information a single act of borrowing carries. Nothing else can work. Notice what is missing from that list. Nothing about the price. Nothing about the quantity available. A designer who responds to an unused facility by making it cheaper has changed a number that was never the problem. A bank at a cheaper window is a bank that could not raise cash even at a bargain, so every appearance becomes more informative, not less.
Which of these reduces stigma: widening the set of institutions that may use the facility, publishing borrower names sooner, or raising the price?
A bank fills the same Rs 9,600 crore shortfall at the window instead of from other banks. What looks different on its balance sheet?
What does borrowing at the window do to the balance sheet?
Nothing distinctive whatever. Cash borrowed at the window raises the bank's settlement balance at the central bank and raises what the bank owes to the central bank, by the same amount, and leaves net worth precisely as it stood. Cash borrowed from another bank raises the settlement balance and raises what the bank owes to that other bank, by the same amount, and leaves the difference between the two sides untouched.
Worked through on Suvarna Commercial Bank at the split used throughout, Rs 9,600 crore of settlement balances arrived, so total assets now read Rs 2,49,600 crore where this morning they read Rs 2,40,000 crore. The liability side rises by the same Rs 9,600 crore. Both sides moved together, so net worth stays at Rs 24,000 crore. And the bank's leverageTotal assets read as a multiple of net worth. It is the second limb of a return decomposition and is settled separately; here it is just a way of saying that both sides of the sheet grew while the difference between them did not. goes from 10.0 times to 10.4 times, and the move is arithmetic rather than an event. Rs 2,49,600 crore divided by Rs 24,000 crore is 10.4. A pledge is not a sale, so the securities pledged come out of the Rs 60,000 crore of investments and stay on the books.
Every amount above belongs to Suvarna Commercial Bank Limited, and the balance sheet was built to be worked by hand, with a shortfall picked so a pencil can check each division. Six operations produce the lot. Deposits of Rs 1,92,000 crore times 0.05, then Rs 2,40,000 crore of total assets times 0.04, both landing on Rs 9,600 crore. The same Rs 9,600 crore times 0.65 and times 0.35, splitting into Rs 6,240 crore and Rs 3,360 crore. Adding Rs 9,600 crore to Rs 2,40,000 crore for Rs 2,49,600 crore. Then Rs 2,49,600 crore over Rs 24,000 crore for 10.4 times. And the cover of 12.35 times shown below is Rs 24,000 crore of net worth over Rs 1,944 crore of net bad advances.
The balance sheet cannot show which route the cash came from, so everything the market believes about window borrowing comes from disclosure rather than from accounts. The claim is a strong one, so test it the hard way with the control below: slide the split from every rupee at the window to every rupee from other banks, and watch the four figures underneath refuse to move.
Slide the split all the way across and watch the accounts stay put
One variable moves: the share of the same Rs 9,600 crore need that comes from the rest of the system rather than from the window. The shortfall stays pegged at Rs 9,600 crore wherever the control is set, so what moves is the source of the cash and never the size of the want. The price difference between the two sources is real and is the reason a bank would prefer one over the other. The Reserve Bank of India sets it and it shifts, so no setting of the control shows it.
65.0 per cent from the rest of the system, and the remainder at the window
With 65.0 per cent of the need met by the rest of the system, Rs 6,240 crore comes from other banks and Rs 3,360 crore comes from the window, which is Rs 9,600 crore in total. The sheet underneath reads Rs 2,49,600 crore of total assets against Rs 24,000 crore of net worth, the first being 10.4 times the second. Slide the control and read that again.
Educational illustration. Invented bank, invented shortfall and an invented split. The price the two routes charge belongs to the Reserve Bank of India and moves, so no setting shows one. Nothing else on the balance sheet moves either: no deposit leaves, no advance is made and no security is sold. The frozen row is the finding rather than an oversight, recomputed from the same whole rupee arithmetic on every move of the control.
The failure: reading an appearance at the window as a statement about solvency
Suvarna Commercial Bank takes Rs 3,360 crore at the window on an afternoon when the market between banks is thin. A reader who has absorbed the stigma concludes that the bank was refused elsewhere, and from there concludes that something is wrong with its book. The second step is where the reasoning breaks. An inference about who would lend has quietly become an inference about what the assets are worth.
The figures already set out carry the point. Net worth Rs 24,000 crore. Net non-performing advancesThe bad advances left after subtracting what the bank has already set aside against them. Gross and net readings sit on different bases and are settled separately. Rs 1,944 crore, so net worth covers them 12.35 times over. Rs 24,000 crore divided by Rs 1,944 crore is 12.35. One asset and one liability rose by Rs 3,360 crore together, so not one figure in that list moved when the cash was borrowed.
Who makes this reading: nearly everybody. The uncomfortable part is that the inference is not stupid. A rarely used facility does carry real information about how the afternoon went. The cost is the sequence that follows. Depositors and counterparties act on the inference, the bank loses more cash, and the next bank in the same position watches what happened and decides to sell assets at speed instead of borrowing. Selling at speed is how a shortage of cash turns into a loss that lands on net worth. The borrowing never touched net worth at all.
One substitution fixes it, and it fits in a line. On learning that a bank borrowed, the question to ask is what moved on its balance sheet. The answer is one asset up, one liability up, and nothing else.
How does somebody on the other side actually use this?
The five minutes a treasury desk, a lender and an analyst each spend on the same fact
The whole chain hangs on the desk with a surplus to place, so start there. The desk has spare settlement balances at four in the afternoon and a request from a bank it does not know well. Its question is not whether that bank is sound in some cosmic sense. Its question is whether it will get the cash back tomorrow morning and what it will earn for the night. The desk may say no for reasons that have nothing to do with the borrower: a limit it is already close to, a large payment of its own tomorrow, a policy about names it does not usually deal with. A refusal at four in the afternoon is very often a statement about the lender rather than about the borrower, and reading it the other way round is where the whole stigma chain begins.
Now the analyst reading a disclosure. The habit worth building is to separate two questions that arrive fused together. Question one: what did the accounts do? One asset up, one liability up, net worth flat, and that much can be checked. Question two: why did the bank prefer this route to the other one? The accounts are silent on that, and the honest answer is usually that it cannot be known from outside. A bank might come to the window for a dull operational reason, or because a large payment landed late. Filling the silence with the worst available story is a choice, not an inference.
And the household version of the same shape, at a wedding scale. A relative asks to borrow money for a function. The answer is no. The household's own money is committed next week. Somebody else hears about the refusal and concludes the relative is in trouble. The refusal was about a calendar. The conclusion was about a character. The gap between a calendar and a character, repeated across a system with money in it, is the whole of the matter.
Who sets the price, the period and the collateral terms?
Four terms decide how this arrangement actually behaves. The charge, and the stretch it covers. The securities that qualify, and the amount knocked off a pledged one before cash goes out, a deduction the trade calls a haircutThe amount taken off a security's market value when working out how much cash it will support as collateral. How a pledged security is valued at all is covered separately.. How long the cash may stay out, and any ceiling on coming back for more. And the publication duty that attaches to the whole thing, together with the lag before it bites.
Every one of them belongs to the Reserve Bank of India, and every one of them shifts, so any figure written down here would be wrong rather than merely dusty on the morning it changed. The table below carries each label with the authority that sets it, and one visit to that source completes the sheet.
The entire stigma is a question of what is visible to whom and when, and the fourth row is where that is decided. Moving the disclosure arrangement moves the stigma, without touching the price, the collateral or the period. A guess at that row would teach the wrong thing about the only variable that decides whether the facility works at all.
Four terms named here and filled in nowhere
| The term | What appears here | Where the number lives |
|---|---|---|
| The charge on cash a bank takes at the counter under its own steam, and the stretch that charge covers | An empty cell | Reserve Bank of India, rbi.org.in |
| The securities eligible to be pledged, and the deduction applied to a pledged one before cash goes out | An empty cell | Reserve Bank of India, rbi.org.in |
| How long the cash may stay out, and any ceiling on repeat visits to the counter | An empty cell | Reserve Bank of India, rbi.org.in |
| The publication duty attaching to such a borrowing, and the lag before it bites | An empty cell | Reserve Bank of India, rbi.org.in |
Copy these four labels down, visit the address sitting in the third column, and write the answers in yourself. Even empty, the sheet earns its place. Knowing which questions exist is most of the battle.
One more, and this is the habit worth walking out with. A bank took cash at the window yesterday. What does that establish about its balance sheet?
Four counters to walk up to, and what each one settles
Four terms decide how this arrangement behaves, and none of them is printed anywhere above. Here is the counter to walk up to for each. Give the last one a second read: everything before it hangs on that.
| Who sets it | The term left blank here | Site | Value here |
|---|---|---|---|
| Reserve Bank of India | The charge a bank meets on cash it draws at the counter under its own steam, together with the stretch that charge covers | rbi.org.in | Left blank. Read off the source on the day it is needed. |
| Reserve Bank of India | The securities that qualify as a pledge for that cash, and the amount knocked off a pledged one before cash goes out | rbi.org.in | Left blank. Read off the source on the day it is needed. |
| Reserve Bank of India | The outer limit on how long the cash may stay out, and any ceiling on coming back for more | rbi.org.in | Left blank. Read off the source on the day it is needed. |
| Reserve Bank of India | The publication duty that attaches to a borrowing of this kind, and the lag before it bites | rbi.org.in | Left blank. Read off the source on the day it is needed. |
Suvarna Commercial Bank Limited is invented.
Educational material. Not advice on any investment, tax, budget or market position.