Lender of Last Resort: The Backstop and Its Conditions

A lender of last resort is a central bank lending to a sound institution that nobody else will lend to, against security it takes, at a price above the market. The conditions are not trimmings on the loan, they are the instrument. Drop the security and it is a gift; drop the soundness test and it is a loss postponed; drop the price and it becomes the first call rather than the last.

Almost everything written about this subject spends its time on the first half of that answer and hurries past the second. The emphasis is the wrong way round. A central bank keeps the accounts, so of course it can lend to a bank. The interesting question is the one the four conditions answer. Something has to stop that power from becoming something else entirely the first time somebody is frightened enough to use it.

So the argument for a backstop comes before the definition of one, and then the conditions come one at a time. Each gets the same two questions: what is the condition for, and what does it cost the person who has to apply it on the afternoon it matters? The doctrine is clean, the application never is, and four conditions offered as a tidy checklist teach only the easy half.

Why does a banking system need a backstop at all?

Two things already settled on this platform meet here, and the meeting is the whole argument. The first is the arrangement a bank runs on: it owes more cash on demand than it can produce on demand. The mismatch is not a flaw somebody failed to fix. A bank that could produce every rupee it owes on demand would be storing money rather than lending it, and would be a very expensive locker rather than a bank.

The second is where the cash goes when it moves. The cash goes somewhere. When a depositor at Suvarna Commercial Bank Limited, an invented bank, moves money out, that money lands in an account at another bank, and the settlement balanceThe money a bank keeps in its own account at the central bank. It is what one bank actually pays another with at the end of the day, and what it is used for is settled separately. that carried it lands there too. The rest of the system now has it. Nothing was destroyed. But a bank holding a surplus tonight can look at an institution it is unsure about and decide, quite reasonably and with no ill will at all, not to lend it out.

The two facts together produce the situation the whole instrument is built for: an institution can be entirely sound and still be unable to get cash, not because the cash is missing but because it will not move. The situation is a very strange kind of problem. The trouble is not a shortage. Every rupee that would answer the demand exists, is sitting in an account, and is doing nothing in particular. Willingness is what is missing, and willingness is exactly what everybody withdraws at the same moment for the same reason.

The household version is closer than it looks. A shopkeeper on the street is good for the money, has stock in the back, and has never missed a payment. One difficult week, a rumour goes round. Nobody who has cash wants to be the one who lent to them just before the rumour turned out to be true, so nobody lends, and the shopkeeper cannot pay a supplier on Friday over a gap that any single neighbour could have covered without noticing. The cash on the street did not fall. The cash stopped circulating around one address. A backstop is the one lender whose willingness does not rest on the same judgement everybody else is making.

Whose doctrine is this, and what does it actually say?

The name is part of the term, so it gets said properly. Walter Bagehot set the four conditions below out in Lombard Street, published in 1873. He was writing about a market that no longer exists in the form he saw it. The argument has outlived that market by a hundred and fifty years, and outliving a market by that much is unusual enough to be worth a sentence.

Lend freely, and let everybody see the cash is there. Lend against good security: with security the central bank is lending, and without it the central bank is giving. Lend at a price above the market rate, and the backstop stays the last place a bank comes rather than the first. And lend only to institutions that are sound. A loan cannot put back what is missing from one that is not.

The four conditions are not preferences to be traded off against one another, they are one instrument, and removing any single one of them does not weaken the operation, it turns the operation into something else. Hold on to that sentence. Three of the conditions come apart below, one at a time, and each shows exactly what the operation becomes when it is dropped. Lend freely is the least contested of the four: everybody agrees a backstop that lends in a trickle has not lent.

Why must the cash go out against security?

Start by clearing away two reasons that sound plausible and are both wrong. Security is not there so the central bank makes something on the deal, and it is not there because the central bank might otherwise run short. A body that keeps the accounts everybody settles in cannot run short of the thing it keeps the accounts in. Neither of those explains a single line of the requirement.

The real reason is much simpler and much more important. Security is there to keep the operation a loan. Cash handed over against nothing is a transfer, the loss on it settles onto the central bank's balance sheet, and a central bank's losses are in the end somebody else's. Cash handed over against nothing is a capital injection wearing a loan's clothes. The point is not that such a thing is unthinkable; the point is that it is a completely different decision, taken by different people, answerable to different questions, and calling it emergency lending would hide all of that behind a word.

The mechanism itself, said plainly, is an exchange. A depositor at a counter will not accept a security certificate, so the bank has assets it cannot pay a depositor with. The bank pledges some of those assets and receives something it can pay with. Suvarna Commercial Bank Limited carries investments of Rs 60,000 crore. A pledge is not a sale, so securities pledged out of that block stay on the bank's own books throughout. The securities come back when the cash is repaid. Nothing leaves the bank permanently, and an operation that reverses itself is one the central bank can run again.

There is a real cost buried in that exchange. Not every security is accepted, and what is accepted is not accepted at its full market value: an amount is taken off first. The deduction has a name, a haircutThe amount taken off a pledged security's market value before working out how much cash it can raise. The size of it is set by the authority named in the routing block below., and a size set by an authority, and that size moves. Which securities count, and how much is taken off each, are two of the empty rows in the sheet at the end.

Why must the price sit above what the market is charging?

The price is the condition readers argue with, and the reason is that it looks cruel. An institution is in difficulty and the response is to charge it more than anybody else would. Sit with the alternative for a moment and the design becomes obvious.

Suppose the backstop were the cheapest cash available anywhere. Then no institution would ever look anywhere else. Looking anywhere else would mean paying more for the identical thing. The market between banksThe place where a bank with more cash than it needs tonight lends it to a bank with less, closing at the end of the day. What it is and how it is priced is settled separately. would stop doing the one job it exists to do: moving surpluses to shortfalls without anybody being asked. And the central bank would end up as the counterpartyThe party on the other side of a transaction, and the one being relied on to do what was agreed. Which party stands where is covered separately. to every short borrowing anybody in the system has, not on the bad day but on every ordinary Tuesday.

The price is what makes the word last mean anything: a backstop priced attractively is not a backstop at all, it is a funding line with a dramatic name. Everything else in the doctrine describes an emergency. The price is the part that keeps the emergency an emergency.

The everyday version is the pawnbroker two streets over, and the comparison is meant plainly rather than unkindly. Nobody goes there because it is cheap. People go because it is open, it is certain, and the alternative that afternoon is nothing at all. If it were the cheapest money in the neighbourhood, everybody would borrow there first and it would stop being the place of last resort when the others have said no. The rate actually charged, how it stands against any other rate, and how long the borrowing may run for, being its tenorThe length of time a borrowing runs before it must be repaid. The one allowed here is set by the authority named in the routing block below., are all set by the Reserve Bank of India.

Why can a loan never repair a hole in net worth?

The soundness test is the hardest condition to hold on to under pressure and the easiest to argue against in the moment, so the case for it is arithmetic rather than assertion. One loan puts the same number on both sides of the sheet at the same instant: something the bank has goes up, and something the bank owes goes up with it. Net worthWhat is left of a business when everything it owes is taken off everything it has. It is settled separately and is used here as a reading rather than rebuilt. is what stands between those two sides. Push both of them up by an identical number and what stands between them has not shifted by a single rupee.

So lending to an institution whose net worth has already been wiped out saves nothing at all; it changes who is holding the claim when the loss is finally recognised, and it puts the recognition off. The central bank ends up as a creditor of a borrower that cannot repay. Read that sentence twice. The shape is the same as everything else on this subject: the problem did not vanish, it moved, and the only question worth asking about any of these operations is where it went.

Two consequences follow and both are uncomfortable. The first is that the security taken does not rescue the position either. Security protects the lender, and protecting the lender is its job. Security does nothing whatever for the borrower's net worth, and a fully secured loan to an institution with a hole in it simply means the loss surfaces somewhere further down the line rather than at the central bank. The second is that the operation is now doing something nobody voted for. Deciding to put capital into an institution is a decision about whose money, on what terms, answerable to whom. Lending to an unsound institution reaches part of that outcome without ever asking the question.

Set the numbers of the invented bank against this so the test is not abstract. On total assets of Rs 2,40,000 crore, Suvarna Commercial Bank Limited reports net worth of Rs 24,000 crore against net non-performing advancesAdvances that have stopped performing, counted after the amount already set aside against them has been taken off. How the counting is done is settled separately. of Rs 1,944 crore, so its net worth stands at 12.35 times that block. Whatever a difficult afternoon does to its cash, it does not move either of those two figures, and that is exactly why the two questions can be answered separately. The condition is not a moral test of the borrower, it is a question about which of two entirely different problems is actually on the table.

What does each condition cost the person applying it at four o'clock?

Everything above is the doctrine as it reads in print, and most treatments stop there. The application is where the subject actually gets hard.

Take the soundness test first. The test has to be answered in hours. The figures available are weeks old. The book being judged is one that nobody can revalue overnight, on an afternoon when the very thing making the judgement urgent is also making it unreliable. And whoever answers will be criticised whichever way they go: lend to an institution that turns out to be unsound and the criticism writes itself; refuse one that was sound and the refusal becomes the reason it was not.

Security is not much easier. Security has to be valued on an afternoon when prices are moving, and the value that matters is not the one on anybody's books, it is what that security would actually fetch that afternoon. The two values are different numbers on exactly the days they most need to be the same. And the price condition reads as the simplest of the four. Applying it requires somebody to name the market rate at a moment when the market may barely be trading.

The doctrine is clean and the application never is, and holding both of those in mind at once is what separates understanding this subject from reciting it. None of that is an argument against the conditions. The argument is against reading them as a checklist somebody ticks. The conditions are three genuinely difficult judgements that have to be made in order, at speed, by people who will own the answer afterwards.

What does a backstop change on the ordinary days?

Everything so far has been about the bad day. The more interesting effect happens on all the other ones, and this part states it as mechanism rather than as a verdict about anybody's character.

If cash is known to be available when things go wrong, then holding less of it when things are fine costs less than it otherwise would. The trade is not cynicism, it is arithmetic. An institution deciding how much to keep ready is weighing what that readiness costs against what it buys, and a backstop changes one side of that weighing for everybody at once, whether or not a single rupee is ever drawn.

Which produces the difficulty at the centre of this whole subject, and it does not have a solution, only a management. The certainty that makes a backstop work on the bad day is the same certainty that changes behaviour on the ordinary one, and the two properties cannot be pulled apart because they are one property seen at two moments. A backstop nobody believes in does not calm anything. A backstop everybody believes in changes what everybody does beforehand.

So look again at the three conditions and notice what they are actually doing. A price above the market, a requirement to hand over security, and a test that can be failed all put a cost on relying on the backstop. The cost of relying on the backstop is the answer to the problem the backstop creates. The cost does not remove the problem. Every condition in the doctrine is an attempt to keep the first property while giving up as little as possible of the second, and that attempt is a balance rather than a fix.

Lender of Last Resort vs Deposit Insurance: which side of the counter does each act on?

The backstop and deposit cover get muddled more often than anything else on this subject, and the muddle is worth taking apart properly rather than settling with two definitions. Four questions do it.

Who is protected. The backstop protects the institution's ability to pay. Deposit cover protects a covered depositor's claim. The institution and the depositor are two different parties standing on two different sides of the same counter, and no amount of careful reading leads from one to the other.

When it acts. The backstop acts once the demand has already started; it is a response. Deposit cover acts on the depositor's reason to join a queue before anything has happened at all. Deposit cover carries on working on every day when nothing happens, and almost every day is one of those.

Which thing changes. The backstop changes what the bank can produce today. Deposit cover changes whether a covered depositor has any reason to be at the counter in the first place. And who administers it: the central bank for the first, and in India the Deposit Insurance and Credit Guarantee Corporation for the second, whose site is dicgc.org.in.

Read those four rows together. One is a decision taken in the moment and the other is a promise fixed in advance that does its work by never being needed, and neither can stand in for the other. A backstop cannot stop a queue forming; it can only answer one. Cover cannot produce a rupee of settlement balance at four o'clock; it can only make the afternoon less likely to arrive. How much of a deposit is covered, the deposits that are covered, what the depositor pays for the cover, and how quickly a covered depositor is paid are all settled at dicgc.org.in, and all of them move.

What does one drawing do to this bank's balance sheet, line by line?

Enough principle. Put the invented bank through the doctrine one condition at a time and watch what each one does, with every figure on the table. The situation below is a declared setting rather than a report of anything: this bank, with total assets of Rs 2,40,000 crore, needs Rs 9,600 crore of settlement balances this afternoon and no other bank will lend it to them.

Condition one, security. The bank pledges securities out of its investments of Rs 60,000 crore, a block 6.25 times the amount it needs and more than enough to pledge from. The pledged securities stay on the bank's own books and come back on repayment. Condition two, the price. Above what the market between banks is charging, and the rate itself belongs to the Reserve Bank of India. Condition three, soundness. Net worth of Rs 24,000 crore stands against net non-performing advances of Rs 1,944 crore, a cover of 12.35 times, and nothing in that comparison moved when the cash was needed.

Now the entries, set out as rows because they are rows. Watch the last two lines in particular, and notice that one of them is doing nothing at all while everything above it moves. The quickest way to see both facts in one figure is to read how many times total assets of Rs 2,40,000 crore stand above net worth of Rs 24,000 crore, and that multiple is what leverageHow many times a business's assets stand above its own net worth. It is established separately and is used here as a reading rather than rebuilt. means here.

Read that last line slowly. The line is the whole reason the third condition exists: the operation gave this bank Rs 9,600 crore of cash and no capital whatever. Every rupee that arrived arrived as a debt. If there had been a hole between the two sides of the sheet, the hole would be sitting there untouched at the end of the afternoon, and the only thing that would have changed is who is owed money by an institution that cannot pay.

The record behind this bank gives no figure at all for what it keeps at the central bank, so the balance rises by Rs 9,600 crore from a starting level that is not stated. The absence is worth more than a made up starting figure would be. A blank identifies exactly which number to go and ask for.

Which single operation in all of this changes the total?

Time to close on the one thing this operation does that nothing else covered here can do. Every other movement had another bank on the far side. A payment took settlement balances out of one account and put them into another. One bank borrowing overnight from another did the same in the opposite direction. A pledge between two banks moved cash one way and a claim the other. In every one of those the two entries cancel, and the system's total of settlement balances is exactly where it was however busy the afternoon has been.

Central bank lending is different in kind, not in degree: the other side of that entry is the central bank's own books, so nothing cancels and the system's total of settlement balances rises. The difference is not a detail about accounting. The difference is the reason this arrangement has to sit with the body that keeps the accounts and cannot be handed to anybody else, however well capitalised or well intentioned they might be. A very rich institution lending to a bank in difficulty is doing something genuinely useful. Its own cash came out of the same total, so it is not doing this.

So here is the sentence to leave with, and it is the one that ties this whole subject together. The cash does not vanish and it does not appear, anywhere, except at one desk. Everything else is a question of where it sits, and that is exactly why the conditions on what happens at that one desk are written down in advance rather than settled on the afternoon.

How does somebody outside a central bank actually use any of this?

Who sets the conditions, and who administers deposit cover?

Six things a reader arrives wanting are set by somebody else. Each is set by a body, each of them moves, and a printed value would be wrong rather than merely stale on the morning it changed. The sheet below carries six labelled rows with the body and its site printed inside each one and nothing in the value column at all.

The sheet has two different bodies on it, and that split teaches the last thing here: the two instruments separated throughout are run by two different organisations, and two organisations is about as clear a confirmation as could be asked for that they are not two names for one thing. Four rows go to the Reserve Bank of India and two go somewhere else entirely. A reader who leaves knowing which body to ask has something that will still be true long after any figure printed here would have stopped being true.