Liquidity Gap vs Liquidity Buffer: A Timing Table and a Stock

A liquidity gap is a timing statement over a period: inflows less outflows inside a bucket, and the running total across buckets. A liquidity buffer is an availability statement on a day: a stock of assets that can be turned into cash this morning. One is a flow and the other is a stock. Comparing them requires choosing a denominator, and the choice of denominator is where most of the confusion starts.

Two tables sit next to each other in almost every liquidity paper written anywhere. One is a ladder of time buckets with a gap on each row. The other is a short list of assets with a total at the bottom. The two tables are printed side by side, carry the same currency and carry the same reporting date, and a great many treasury committees read them as two views of one position. The two tables are not two views of one position. A ladder and an asset list are two different kinds of measurement, and one difference between the two kinds is what everything else follows from.

Are a liquidity gap and a liquidity buffer two views of the same thing?

No, and the reason is older and simpler than banking. A flowAn amount moving over a period, which is what a gap measures and why a gap always has a time bucket attached to it. is an amount that moves over a period. A stockAn amount held at a moment, which is what a buffer measures and why a buffer figure is always as at a date. is an amount that is held at a moment. A salary is a flow and a bank balance is a stock. Nobody confuses those two at home. The question asked about each is obviously different: how much arrives each month, and how much is there right now. A liquidity gap is the salary question and a liquidity buffer is the balance question, and no arithmetic turns one into the other on its own.

The household version does all the work. Suppose a house runs on one salary of Rs 60,000/- a month, and the household keeps Rs 90,000/- in a savings account. The Rs 90,000/- is the stock. The month in which Rs 72,000/- goes out against Rs 60,000/- coming in is the flow, and the shortfall of Rs 12,000/- is that month's gap. Then comes the question everybody asks: is the household safe? The stock covers that one month's shortfall seven and a half times over. Seven and a half times over sounds enormous. But if the same shortfall repeats for eight months the stock is gone, and nothing in the Rs 90,000/- ever said how many months there would be. The count of months is the whole argument.

Vindhya Commercial Bank Limited, invented, is that household at Rs 96,000 crore. Everything below belongs to it, at month 12, and every assumption in it is the bank's own working number rather than a requirement. The bank keeps a maturity ladder of eight buckets, LB1 to LB8, and the ladder is the flow table. The bank also keeps a liquid asset buffer of four components, H1 to H4, and the buffer is the stock table. The two tables were built for different questions and were never designed to reconcile with one another, and one of the more useful things to understand about liquidity work is why that is a design choice rather than an oversight.

What exactly is a liquidity gap a statement about?

A liquidity gapInflows less outflows inside one time bucket of a maturity ladder, before any action is taken. is inflows less outflows inside one defined bucket of time, on a stated slotting basis, before anybody does anything about it. Three parts of that sentence carry weight. The bucket is defined, so the figure is meaningless without the period attached to it. The slotting basis is stated. The same rupee lands in different buckets depending on the question the table is asking. And a gap is measured before any action, so a gap describes the balance sheet as it stands rather than forecasting what will happen.

At this bank the ladder is slotted by contractual maturityThe date a contract says money is due, which is the starting point of a ladder and almost never the ending point. with one behavioural adjustment: of the Rs 36,000 crore of current and savings balances, 5.0 per cent being Rs 1,800 crore is placed in LB1 and the rest is spread across LB5 to LB8. The behavioural adjustment is the bank's own assumption, and the single most load-bearing one in the whole table. LB1 shows inflows of Rs 7,200 crore against outflows of Rs 9,600 crore, so the gap is minus Rs 2,400 crore. LB2 is minus Rs 1,200 crore, LB3 minus Rs 2,400 crore, LB4 minus Rs 1,200 crore and LB5 minus Rs 2,400 crore. Then the sign flips: LB6 is plus Rs 2,400 crore, LB7 plus Rs 2,400 crore and LB8 plus Rs 4,800 crore.

Reading down the bucket gaps teaches very little. A single bucket's shortfall can be met out of the surplus of the bucket before it. The figure that matters is the cumulative gapThe running total of bucket gaps to a point in time, which is what shows whether a shortfall accumulates or reverses., the running total. The running total goes minus 2,400, minus 3,600, minus 6,000, minus 7,200, then peaks at minus Rs 9,600 crore at one year, then recovers through minus 7,200 and minus 4,800 to exactly zero at LB8. Total inflows and total outflows are both Rs 96,000 crore, so the cumulative total ends at zero by construction. A ladder that ties to nothing at the end has said nothing about whether the bank is safe. The shape in the middle is the entire content of the table.

What exactly is a liquidity buffer a statement about?

A liquidity bufferA stock of assets held to be turned into cash quickly, counted after a discount for the cost of monetising them. is a stock of assets held so that they can be turned into cash quickly, counted after a discount for the cost of doing so. A buffer has no period. A buffer has a date. Vindhya Commercial Bank Limited's buffer at month 12 is Rs 14,400 crore in four components: H1, cash and balances with the central bank above the reserve requirement, Rs 1,920 crore. H2, central government securities, Rs 9,600 crore. H3, state government securities, Rs 1,920 crore. And H4, other assets after the bank's own haircut, Rs 960 crore. The four sum exactly: 1,920 plus 9,600 plus 1,920 plus 960 is Rs 14,400 crore.

The word that does the work in the buffer's definition is monetisableCapable of being turned into cash on the day, by sale or by borrowing against it, which is a stronger test than being marketable.. Monetisable is a stronger test than marketable and much stronger than owned. An asset qualifies if somebody would buy it, or lend against it, on the morning it is needed, at a price the bank has already written down its discount for. A buffer is therefore a claim about somebody else's behaviour on a particular morning. A gap is a claim about contract dates and needs nobody's agreement at all, and that is the deepest difference between the two. As shares of the total the buffer is 13.3 per cent H1, 66.7 per cent H2, 13.3 per cent H3 and 6.7 per cent H4, and the four shares sum to 100.0. Government securities H2 plus H3 are Rs 11,520 crore, being 80.0 per cent of it. To put the stock in scale, Rs 14,400 crore is 15.0 per cent of total assets of Rs 96,000 crore, 18.8 per cent of deposits of Rs 76,800 crore and 1.88 times the bank's equity of Rs 7,680 crore.

Which four questions separate the two objects?

When two figures are being argued about in a meeting and nobody can say why they disagree, the fastest way through is to ask each figure four questions in order. Over what period is this measured? On what day is it measured? In what conditions? And who else has to agree for it to be true? A gap answers the first and the third and needs nobody's agreement. A buffer answers the second and needs a counterparty. Two measures that answer different questions from that list are not rival estimates of one thing, and the argument about which is right is an argument that cannot be settled because it was never a disagreement.

How is a buffer compared against a gap without getting it wrong?

The two figures can be compared, and treasury teams do compare them every month. The comparison that means something is the buffer against the cumulative gap, bucket by bucket, expressed as coverage in timesA stock divided by a shortfall, expressed as how many times over the stock would meet it.. Coverage in times answers a defensible question: if the running shortfall to this point in time had to be met out of the stock held today, how many times over would the stock meet it? The question is honest only while one assumption holds. The assumption is that the stock is still whole when the shortfall arrives.

Run it on this bank and every figure is exact. Against LB1's cumulative gap of Rs 2,400 crore the buffer covers 6.00 times. Against LB2's Rs 3,600 crore, 4.00 times. Against LB3's Rs 6,000 crore, 2.40 times. Against LB4's Rs 7,200 crore, 2.00 times. Against LB5's Rs 9,600 crore, 1.50 times. Then it recovers: LB6's Rs 7,200 crore gives 2.00 times and LB7's Rs 4,800 crore gives 3.00 times. At LB8 the cumulative gap is zero, so the measure is undefined rather than infinite. Saying that in words beats printing a symbol nobody can act on.

The tightest point in the whole table is LB5, at one year, at 1.50 times, and it is the opposite of where a reader looks first. Almost everybody's eye goes to LB1. LB1 is the nearest bucket and the one that feels urgent, and LB1 is also the most comfortable row in the table at 6.00 times. The tight point in a cumulative table is wherever the running total peaks, not wherever the clock starts. Taken as a share instead of a multiple, the same curve repeats the shape: the cumulative shortfall is 25.0 per cent of cumulative outflows through LB1, 25.0 through LB2, 22.7 through LB3, 20.0 through LB4 and 19.0 through LB5, then falls away to 10.0, 5.9 and zero. The fall at the end is not the bank getting safer. The fall is the denominator growing and the long assets finally maturing.

Why do the same balances run off at two different rates?

One buffer reads three different percentages below. One thing has to be said out loud first, or the three percentages look like a contradiction. The bank keeps more than one view of the same money, and the views were built for different worlds.

Take the Rs 36,000 crore of non-maturity deposits, meaning current accounts of Rs 9,600 crore and savings accounts of Rs 26,400 crore. In the maturity ladder, 5.0 per cent of that balance, being Rs 1,800 crore, sits in LB1 and nothing more of it leaves until LB5. So over thirty days the ladder assumes Rs 1,800 crore of it walks out. In the bank's thirty day coverage computation the same Rs 36,000 crore is split across two categories and run off at two different rates: savings of Rs 26,400 crore at an assumed 7.5 per cent is Rs 1,980 crore, and current accounts of Rs 9,600 crore at an assumed 40.0 per cent is Rs 3,840 crore. Together that is Rs 5,820 crore, a blended 16.2 per cent, and 5,820 over 1,800 is 3.23 times as much. The difference is Rs 4,020 crore of the same money on the same balance sheet in the same month. Neither table is wrong. The ladder describes ordinary conditions and the coverage computation describes a stress, and the assumptions in each were chosen to answer that table's own question.

Why does one buffer read three different percentages?

Here is where readers lose their footing, and it is worth slowing down for. The bank's buffer is Rs 14,400 crore. The buffer figure is reported as a percentage in at least three places in the bank's own papers, and the three percentages are different. Nothing about the bank changes between them. Only the denominatorThe figure a ratio is divided by, and in liquidity work the choice of denominator changes the answer more than the numerator does. changes, and the denominator is the question being asked of the same stock.

Against the one year cumulative gap of Rs 9,600 crore, the buffer reads 150.0 per cent. The standardised thirty day net cash outflow is Rs 11,520 crore, being gross outflows of Rs 15,924 crore less counted inflows of Rs 4,404 crore, and against that denominator the buffer reads 125.0 per cent. The bank's own severe scenario consumes Rs 13,200 crore over the same thirty days, and against that denominator the buffer reads 109.1 per cent. Add a fourth for scale: against LB1's cumulative gap of Rs 2,400 crore it reads 600.0 per cent. One stock, four denominators, four answers, and every one of them is arithmetically correct. A buffer percentage quoted without naming its denominator has therefore said almost nothing at all.

The two thirty day figures deserve one more sentence. The two figures look like a contradiction and are not. The standardised computation averages Rs 384 crore of net outflow a day and the bank's own severe scenario averages Rs 440 crore a day. Dividing one by the other, 13,200 over 11,520 is 1.1458, so the bank's own scenario is 14.6 per cent harsher over thirty days than its standardised computation. Put the same buffer over the harsher denominator and 125.0 per cent becomes 109.1 per cent. The ratio of 1.1458 is the whole reconciliation, and it explains why the same institution can report a headline that sounds like a quarter more than needed while its own harsher scenario eats 91.7 per cent of the buffer inside a month.

Why do the same securities say different things in the two tables?

Government securities appear in both tables and say two different things, and that is design rather than error. In the buffer table, H2 and H3 together are Rs 11,520 crore of government securities, and the claim being made about them is that they could be turned into cash this morning. In the maturity ladder, securities are slotted by when they contractually mature, and the claim being made is about when the cash arrives on its own. A ten year government bond is monetisable today and matures in ten years. Both statements are true about the same security.

And here the honest limit of this bank's record has to be stated: it does not say which of the eight ladder buckets the buffer's own securities sit in, so nobody may assign them. The record locks the three investment books at Rs 3,600 crore held for trading, Rs 8,400 crore available for sale and Rs 14,400 crore held to maturity, and it locks the buffer components at Rs 1,920, Rs 9,600, Rs 1,920 and Rs 960 crore. The record gives no mapping between the two lists. A document that assigns a bucket to make the tables tidy has invented a fact, and the reconciliation it produces will not be reproducible by anybody who tries.

Which of the two moves when management actually does something?

This is where the distinction stops being academic. The two tables respond to completely different actions, at completely different speeds, and a treasurer who knows which lever moves which figure can do in an afternoon what a confused one cannot do in a quarter.

Take the first action. The bank raises Rs 1,200 crore of one year term deposits and uses them to replace Rs 1,200 crore of fourteen day money. LB1's outflows fall from Rs 9,600 crore to Rs 8,400 crore, so LB1's gap improves from minus Rs 2,400 crore to minus Rs 1,200 crore, and the cumulative shortfall through LB4 falls by Rs 1,200 crore at every point. The buffer does not move by one rupee: not a single asset has been added to the stock of things that could be sold this morning. And a second lesson hides in the same trade. The money has only moved inside the year: LB5's outflows rise by the same Rs 1,200 crore, so the one year cumulative gap is still exactly minus Rs 9,600 crore, and the tightest point on the cover curve has not improved at all.

Now the second action. The bank sells Rs 1,200 crore of securities held outside the buffer. The stock available this morning rises by the cash raised, and raising cash on the morning it is needed is exactly what a buffer is for. Not one liability date has moved, so no outflow row in the ladder changes. An inflow is pulled forward from wherever that security would have matured, and this bank's record does not decompose its ladder rows, so nobody can say which bucket that inflow left. Say the direction and stop there, rather than putting a number on a movement the record cannot support.

Can a bank have a large negative gap and still be comfortable?

Yes, and it is the ordinary state of a bank rather than a warning sign. The bank's first five buckets are all negative and the cumulative shortfall through one year is Rs 9,600 crore, being 1.25 times the whole equity of Rs 7,680 crore. Read cold, that sounds alarming: the timing mismatch is larger than the capital sitting behind the institution. But the shortfall is a timing statement and not a loss statement. Nothing has gone wrong. Short liabilities funding long assets is what a bank does for a living, and the equity was never the thing meant to meet a timing shortfall in the first place.

The reverse case is the one that catches people out. A bank can hold a very large buffer and still run out of cash. A stock says nothing about how many days the outflow lasts. Take this bank's own severe scenario: Rs 14,400 crore of buffer against consumption of Rs 720 crore a day for five days, Rs 600 a day for five more, Rs 420 a day for ten and Rs 240 a day for ten. The path consumes Rs 13,200 crore in thirty days and leaves Rs 1,200 crore, being 8.3 per cent of the stock, on the morning of day 31. A headline that reads a quarter more than needed and a stock that is 91.7 per cent gone in a month are the same buffer described two ways. A buffer figure alone can never answer the question people most want to ask, and the question people most want to ask is how long the bank lasts.

Who reads these two tables, and what do they do with them?

Three different people pick up the same monthly paper and take three different things out of it, and the difference between them is entirely a difference in which table they need.

Devendra Achar, head of treasury at this invented bank, reads the ladder first. He is looking for where the running shortfall peaks. His funding plan has to reach that peak, and at this bank the peak is one year out rather than the fortnight everybody worries about. He reads the buffer second and for a different purpose: not as a comfort figure but as an inventory of what he could actually monetise on a bad Monday, and he cares much more about the composition than the total. Rs 11,520 crore of it is government securities and Rs 960 crore is component H4 counted after a discount, and those two behave very differently on a morning when everybody wants to sell.

An analyst at another institution, looking at this bank as a counterparty rather than an employer, reads the two in the opposite order and for a third purpose. The buffer total tells them what could be raised if the name got difficult. The ladder tells them how long the name would need to keep raising it. Neither figure means much alone, and an analyst who quotes a coverage percentage without asking what it was divided by has repeated a number rather than understood a position.

The mechanism is identical at every scale, so the household version says the same thing. A person with a stable salary and no savings has a comfortable gap table and no buffer, and one unexpected hospital bill ends them. A person with Rs 6,00,000/- in the bank and no income at all has a magnificent buffer and a gap table that gets worse every single month, and the stock only tells them how long they have rather than whether they are safe. Most people are somewhere in between and are managing both tables at once without ever having written either one down. A bank is the same person with a committee and a spreadsheet.

Where does confusing the two actually cost something?

Four decisions, and each one lands on a different desk.

Pricing a term deposit campaign is a gap decision. The campaign moves outflows from a near bucket to a far one and changes the shape of the ladder, and it adds nothing whatever to the stock of assets that could be sold this morning. A treasurer who runs a campaign expecting the buffer figure to improve has spent money to move a number that was never going to move.

Sizing a securities sale is a buffer decision. The sale raises the stock and moves no liability date, so a ladder printed the next morning looks almost identical. Somebody who sells to fix a gap has sold for nothing.

Setting a cap on a bucket's negative gap is a gap decision that no amount of buffer satisfies. The bank has exactly one such cap in its whole limit set, and the cap applies to LB1: the negative gap in that bucket may not exceed 28.0 per cent of the bucket's outflows. The current gap is minus Rs 2,400 crore on outflows of Rs 9,600 crore, being 25.0 per cent, so utilisation is 25.0 over 28.0, being 89.3 per cent, and the cap in rupees is 28.0 per cent of 9,600, being Rs 2,688 crore, so the headroom is Rs 288 crore. The limit is measured on the flow table and the buffer is not in it, so no quantity of buffer changes that utilisation by a single decimal.

And whether the institution survives a month is neither a pure gap question nor a pure buffer question. The answer needs the stock measured against a stressed daily path, and a stressed daily path is a third measure again with its own treatment. The ladder carries no stressed path and so cannot answer that question at all.

What does this comparison not settle?

Quite a lot, and it is worth being precise about the edges rather than letting the comparison sprawl. Three denominators appear above and not one of them is derived. How a stressed scenario is designed, what makes one severe rather than merely adverse, and who signs it off, is a separate subject. The coverage computation with its run-off and inflow rules, the survival horizon as a measure, and the matching of funding to asset life over a year are each measures with their own treatment, and every one of them appears here only as a denominator to put under one stock.

Why the ladder has the shape it has, meaning the structural mismatch between short liabilities and long assets, is covered separately. So is depending on too few funding sources, and so is what an institution writes down about the day funding stops arriving. The repricing ladder belongs to the interest rate work. The repricing ladder slots the same balances by when the rate changes and the maturity ladder slots them by when the cash leaves, so the same Rs 36,000 crore of deposits sits in different rows in the two. Central bank support, on what terms and against what security, is covered separately. A repo, a certificate of deposit and a government security are named here and taught separately.