Reserve Requirements: CRR, SLR and What They Constrain

A reserve requirement obliges a bank to hold part of what it owes in a form that is ready at once. One shape is a balance kept in the bank's own account at the central bank. The other is a holding of specified securities that stay on the bank's books. Both are struck on a measure of its liabilities, and the rates and that measure are set by the Reserve Bank of India.

Six things decide how these obligations bite: the two rates, the measure they are struck on, which liabilities enter that measure, what is paid on the balance and the stretch of time a bank's position is read over. Every one of them is set by the Reserve Bank of India, and every one of them moves. A figure printed for any of the six would not be slightly out of date the morning it changed. Such a figure would simply be wrong, and nothing about it would signal which of the six had moved.

The mechanism is worth more than the figures. Once the obligation's work is understood, along with where the locked cash goes, what it costs the bank carrying it and why the measure it is struck on matters as much as the rate, the six numbers can be filled in from the source in one sitting and read correctly for the rest of a working life. Somebody who leaves with six numbers and no mechanism will be confidently wrong the first time one of them moves, and will not notice.

Why does a bank need to be told to keep anything ready?

Every obligation of this kind is an answer to one mismatch and nothing else, so start with the mismatch. A bank owes money that can be asked for at any moment. Somebody walks in, or taps a phone, and the money has to be there. Against that promise the bank holds assets that cannot be produced at any moment: an advance of Rs 1,44,000 crore is a set of agreements to be repaid over years, not a stack of notes in a drawer, and a building is a building.

Left entirely to itself, each bank decides for itself how much to keep ready against that promise. And whatever is kept ready earns little or nothing while it waits, so each bank has a perfectly sound commercial reason to keep less rather than more. A rupee sitting still is a rupee not lent. Every bank faces the same pull in the same direction, and each one is making an entirely reasonable judgement about its own position.

The judgement is made one bank at a time, and the consequences of getting it wrong land on people who were never party to it. A private judgement becomes a public question at exactly that point. The depositor did not sit in the meeting. Nor did the small business whose payments run through that bank, nor the other banks that expected to be paid this afternoon. A reserve requirement takes a defined slice of that judgement out of the bank's hands and fixes it the same way for everybody, in advance, whatever any individual bank would have preferred.

Here is the everyday version. Ten shops share a mall, and each shopkeeper decides privately how much of the corridor outside to keep clear. Every one of them has a good reason to let the display creep out a little: the corridor earns nothing, the shelf earns something. Each decision is sensible on its own and the corridor is impassable by Saturday. So the mall writes one rule saying how much of the corridor stays clear, whatever any single shopkeeper would prefer, and the rule is about the corridor rather than about anybody's character.

What are the two shapes an obligation like this takes?

The two obligations are described so often as a pair that people assume they are two settings of one dial. In truth they are genuinely different things, and the difference is where the asset sits and what it is doing while it sits there.

The first is met by cash. The bank keeps a balance in its own account at the central bank, and that account is what the sequence before this one called a settlement balanceThe account every bank keeps at the central bank, and the money in it. Payments between banks are finally settled by moving amounts between these accounts. What sits in one is worked out separately.. In India this obligation is known as the cash reserve ratio (CRR). While the balance sits there, the bank cannot lend it. Whatever is paid on it, if anything at all, is decided by the authority.

The second is met by composition. The bank has bought specified securities, they stay on its own balance sheet, they earn whatever those securities earn, and it can pledgeTo promise a lender that it may seize a particular security if a borrowing goes unpaid, while ownership of the security itself stays with the borrower throughout. What such a promise raises in cash is settled elsewhere. when it needs cash against them. In India this obligation is known as the statutory liquidity ratio (SLR). The bank has not parted with anything; it has been told what a slice of its assets has to be made of.

The two names hide the distinction that actually matters: the first takes cash out of the bank's use, and the second says what a defined part of the bank's assets has to be made of. One is a claim on cash. The other is a claim on composition. Holding that distinction rules out describing them as two dials on one instrument.

What does each of the two actually constrain?

The two are routinely described as though they did the same job in different clothes, so be exact here. The cash reserve constrains how much of every rupee taken as a deposit can go anywhere else at all. A slice comes off before any other decision is made, and everything the bank might have done with that slice is off the table.

The securities obligation does something narrower and easier to misread. The securities obligation does not constrain how much may be lent, only what the part that is not lent is invested in. The lending decision is taken first and on its own terms; the obligation then says that a defined slice of what remains has to be made of specified securities rather than of anything else the bank might have preferred.

For anybody reading a bank's balance sheet that difference lands in one sentence: a bank carrying a large securities obligation is not a bank that lends less, it is a bank whose investment book has a shape somebody else chose. Such securities are not idle. Securities held this way earn, they can be pledged, and while they are pledged they are encumberedSet aside or promised to somebody else, and therefore not free to be used for anything further. An encumbered asset still belongs to its holder; it simply already has a job. rather than gone. The obligation removes the choice rather than the earning.

The wrong assumption about these obligations is common and expensive, so what neither of them does is worth saying plainly. Neither obligation says anything at all about how much may be lent to any single borrower. Neither is a rule about how risky the loan book may be. Neither is a rule about how much the bank must hold against losses. Lending limits, risk rules and loss cover are separate obligations answering separate questions, and reading any of them into this one leaves a reader believing a bank is constrained in a direction it is not.

Why does the base matter as much as the rate?

The measure is the part almost nobody is told about. Both obligations are struck on a measure of the bank's liabilities, and that measure is defined by the authority rather than by arithmetic. Which liabilities enter it. Which are left out. On what date it is read. Over what stretch it is averaged. Each of those is a choice somebody made, and each of them moves the answer.

Work it on this bank and the point stops being abstract. Suvarna Commercial Bank Limited, an invented bank, reports deposits of Rs 1,92,000 crore. Its total assets are Rs 2,40,000 crore and its net worth is Rs 24,000 crore, so what it owes altogether is the difference, Rs 2,16,000 crore. Of that, deposits account for Rs 1,92,000 crore and the remaining Rs 24,000 crore is a funding line nothing available here puts a name to. There are two entirely defensible measures of what this bank owes sitting in one set of reported figures, and they are Rs 24,000 crore apart.

Now take one unchanged setting: Rs 12.00/- locked out of every Rs 100.00/- of the measure. The setting is a declared illustration and not an obligation anywhere. On deposits alone it locks Rs 23,040 crore. On everything the bank owes it locks Rs 25,920 crore. The setting never moved. Nothing was announced. One eighth more cash is locked away.

And here is the finding: reaching Rs 25,920 crore by raising the setting to Rs 13.50/- per Rs 100.00/- and reaching it by widening the measure produce exactly the same number, and only one of the two routes is visible from outside. A change in the rate is announced, reported and discussed. A change in what enters the measure is a definitional amendment that most readers never see. Somebody watching only the rate is watching one of two levers.

The reading rule carries beyond this subject. When a requirement is stated, ask what it is struck on before asking how large it is. The same figure on two different measures is two different obligations, and the number on its own says almost nothing. Both the rate and the measure are set by the Reserve Bank of India, and both of them move.

When cash is locked away, where does it actually go?

Nowhere. Almost every loose description of this mechanism implies otherwise, so the one-word answer is worth sitting with. A rupee moved into a balance at the central bank is still a rupee in the system. The rupee was in one account and is now in another account at the same institution, and the sum of every balance is exactly what it was before the move.

Raising the obligation does not take cash out of existence, it takes it out of circulation, and lowering the obligation does not create cash, it releases cash that was already there. The total did not change. The usable part did. Out of existence and out of circulation are different sentences, and mixing them up is how people end up believing that a central bank destroys money by requiring banks to hold more of it.

The whole subject of system liquidity runs on one structure. The cash does not vanish, it moves, and the question is always where it went. Following the rupee out of one account and into another answers almost anything about system liquidity. In this case it went from a place where the bank could lend it to a place where the bank cannot, and both places are accounts kept by the same institution.

The household version is a jar on a shelf. A household decides that Rs 2,000/- of this month's salary goes into the jar and stays there. The bank balance falls, the jar rises, the household is not one rupee poorer, and the amount available to spend this week has definitely gone down. Somebody looking only at the spending would report that money had disappeared. The money is on the shelf.

What does the obligation cost the bank that carries it?

Per rupee of deposits is the scale small enough to hold in the head. On Rs 100.00/- of deposits, the bank pays the depositor what it agreed to pay on the whole Rs 100.00/-, whatever share of that rupee is sitting locked away. The depositor did not agree to a smaller rate because part of the money went into a settlement balance. The depositor was not asked and has no reason to be.

So if part of that Rs 100.00/- has to sit still, then whatever the bank earns has to be earned by the part that is left. At a declared setting of Rs 12.00/- locked, Rs 88.00/- is doing the earning while Rs 100.00/- is doing the paying. Nothing about the arrangement with the depositor changed. Everything about the arithmetic on the bank's side did.

The mechanism to keep is that the obligation raises the cost of the rupees that remain usable without changing the rate paid on any single deposit by even a paisa. The cost shows up instead in a bank's net interest marginWhat a bank earns in interest less what it pays in interest, read against the assets that actually earn. Which assets go into that base is a choice, and it is worked out separately. and never in its deposit rates. The cost of a reserve obligation does not appear in what banks advertise on savings accounts, and looking for it there leads to the wrong conclusion that there is no cost.

The second half is the one that gets left out. The rate paid on the balance held at the central bank, if any rate is paid at all, changes how large that cost is. A balance that earns something costs the bank less than a balance that earns nothing, and the difference can be the whole of the argument. The Reserve Bank of India sets that rate, and it moves.

Why is it a floor rather than a target?

A bank may hold more than it is obliged to hold, and banks routinely do. The reason is practical rather than dutiful: a balance at the central bank is the one asset that settles a payment instantly and without argument. When a large payment has to go out this afternoon, nothing else on the balance sheet will do the job in the time available.

So the picture to keep is three layers. The obligation is the ground. The amount banks actually choose to hold sits above it. The difference between the two layers can move, and that makes it the interesting part.

The gap between the two layers is what actually gets lent in the overnight market between banksThe market in which banks lend to one another for a single day, usually to square off a shortfall at one bank against a surplus at another. How it is priced and settled is worked out separately., so the obligation sets the level the market clears around rather than deciding the market. A bank sitting on more than it needs today can lend the surplus to a bank sitting on less than it wants. Sending the obliged part away is what breaching the obligation means, so the obliged part cannot go anywhere. Only the layer above it is merchandise.

Raising an obligation is not the same instrument as draining cash, even though both reduce what is available. Raising the obligation lifts the ground under the market. Draining cash lowers the level sitting on top of it. Either can squeeze the gap to nothing. The two arrive at that result from opposite directions, and the direction matters a great deal to anybody working out what happens next.

Does it matter whether the position is read every instant or on average?

The reading rule matters more than the size of the balance, and it is the design point that almost never gets explained. Suppose the position is read at every moment. Then the balance is dead weight. Using the balance puts the bank below the obligation the instant it is used, and being below at any instant is a breach, so the balance can never be used for anything. The bank holds a large asset that does no work at all.

Now suppose the position is read as an average across a period instead. The bank can be below on Tuesday and above on Thursday, and as long as the average across the period holds, nothing has been breached. The same balance has become a genuine cushion. A bad afternoon can be absorbed and made up afterwards. A bank needs precisely that when a large payment arrives without warning.

Whether the balance is dead weight or a working cushion is decided entirely by how the position is read, not by how large it is. Two banks holding identical balances under the two arrangements are in completely different situations, and nothing on either balance sheet would show which. The arrangement, the length of the period and everything about how the position is read are set by the Reserve Bank of India, and they move.

The everyday version is a school attendance rule. A rule requiring presence every single day makes one fever a breach. A rule requiring a certain average attendance across the term allows a pupil to be ill in October and make it up in November. The requirement has not become softer; it has become usable, and the difference is in the reading rule rather than in the number.

What does this bank's own rupee actually look like?

Everything below is worked from figures this bank reports rather than from anything supplied here. Deposits are Rs 1,92,000 crore, advances are Rs 1,44,000 crore and investments are Rs 60,000 crore. Divide rather than quote, every time.

Rs 1,44,000 crore divided by Rs 1,92,000 crore is three quarters, so Rs 75.00/- has been lent for every Rs 100.00/- of deposits. Rs 60,000 crore divided by Rs 1,92,000 crore is five sixteenths, so another Rs 31.25/- sits in securities. The two add to Rs 106.25/-, exceeding the deposit rupee by Rs 6.25/-.

The Rs 6.25/- excess points at something the figures do not settle: deposits are not the only thing funding this bank's earning assetsThe assets that actually produce interest, chiefly what has been lent and what has been invested. Which assets belong in that base is a choice, and reading a margin against the wrong one is a common error worked out separately., and this record does not name the rest. A funding line of Rs 24,000 crore arrives with no label attached. The identity of that line is the precise question to put to whoever keeps the real figures, and the question beats a plausible guess in every way that matters.

The second absence is sharper, and it sits on the very quantity at issue. Set the Rs 2,40,000 crore of total assets beside the Rs 2,04,000 crore that lending and investing account for between them, and Rs 36,000 crore of assets is neither. The unexplained remainder is Rs 15.00/- of every Rs 100.00/- of total assets. The record does not split that Rs 36,000 crore, so how much of it is a balance kept at the central bank cannot be established.

What is a reserve requirement not?

The wrong picture is common enough to be worth correcting head on. A reserve requirement is not a fund set aside for depositors, and it is not insurance. The requirement sits on the asset side and says nothing whatever about the difference between what the bank has and what it owes, so it is not capitalThe difference between what a bank has and what it owes, being the part of the balance sheet that absorbs losses. How much of it a bank has to carry is a separate obligation answering a separate question, and it is worked out separately. either.

The balance at the central bank belongs to the bank, and it protects no depositor in particular. The balance carries no account number and is divided among nobody. If the bank failed tomorrow, not one rupee of it would be paid to anybody because of this obligation, and what a depositor is actually covered for runs through an entirely separate arrangement.

The positive work of the obligation matters as much as its limits. The obligation makes a defined part of every bank's promise payable immediately, at every bank, whatever any individual bank would have chosen for itself. The claim is about the system rather than about any single account, and it is precise rather than vague. The corridor is clear at every shop in the mall, and none of the clear corridor belongs to any one shopkeeper.

How does anybody actually use this on a working day?

Who sets the rates, the measure and the averaging?

Six things belong to somebody else. The sheet below carries a label on every row with the authority printed inside it, and that keeps the sheet usable away from here.

All six are set at the same address, and all six move, so a figure printed for any one of them would be an error rather than an old fact. There is a second reason as well. Somebody who leaves with six rows and a site can fill the sheet in one sitting and will still be right next year. Somebody who leaves with six numbers will be confidently wrong the first time one of them moves, and will have no way of knowing which.