Asset Liability Management: Matching the Two Sides

Asset liability management is the standing job of matching what a bank holds against what it owes, on two axes at once: when each side falls due, and when each side reprices. Asset liability management is a function rather than a measure. A committee runs it, and because a large part of a bank's deposits carry no contractual maturity at all, the matching runs on behavioural assumptions that same committee decides.

Two things have to be accepted before the rest of the subject makes sense. The first is that a bank borrows short and lends long as a matter of choice, and that the mismatch this creates earns the bank its living rather than marking an error inside it. The second is smaller and harder: where a contract does not say when money leaves, somebody has to assume it, and an assumption is a decision made by named people at a named meeting. Everything difficult in asset liability management comes out of that second sentence.

Why does matching stand inside a bank?

A corporate treasury lives inside an operating business, and Nirjhar Industries Limited, an invented steel and alloys group, is one. Matching the two sides of a balance sheet is a bank's problem in a way it is simply not a steel maker's. A steel maker funds itself and then goes and makes steel; its balance sheet is the scaffolding around an operating business. A bank has no operating business behind the balance sheet. The balance sheet is the business. A bank holds somebody else's promise to pay, owes its own promise to pay, and takes the whole of its earnings from the difference between the two. A balance sheet that is itself the business is why asset liability managementThe standing function of matching what an institution holds against what it owes, by when each side falls due and when each side reprices. is a named standing function inside a bank and a paragraph in a policy inside most companies.

The bank throughout is Vindhya Commercial Bank Limited, an invented commercial bank with a balance sheet of Rs 96,000 crore. Every ratio, factor and limit that follows is Vindhya's own working number, set by Vindhya's own committees, rather than a rule Indian banking imposes on anybody.

What are the two sides, and what is each one made of?

The phrase "the two sides" gets used so loosely that people stop looking at what is actually in them, so start with the object itself. Vindhya's assets total Rs 96,000 crore and its liabilities and equity total the same Rs 96,000 crore. Balancing to the rupee is arithmetic rather than achievement. The shape inside each side is what matters. On the asset side, Rs 57,600 crore of net advances is money that will come back over years and on dates the bank agreed with borrowers; on the liability side, Rs 36,000 crore of current and savings balances is money that can walk out this afternoon. A book of agreed dates funded by a book with no dates at all is the whole subject in one sentence.

Look at that bright green block again. The block is the reason this subject is hard. Rs 9,600 crore of current account balances and Rs 26,400 crore of savings balances add to Rs 36,000 crore of non-maturity depositsA balance the customer may take at any time, so no contract anywhere says when it leaves., being 46.9 per cent of the bank's Rs 76,800 crore deposit book. Every one of those rupees is contractually repayable on demand. Nothing anywhere says when they leave, so the bank cannot look it up. And yet the bank has to build tables that place them somewhere, and every table it builds has a different reason for placing them.

What are the two axes, and why is answering one not answering the other?

Matching sounds like one activity and it is two. An ordinary rent agreement shows why. The landlord fixes the rent for eleven months, so the date on which the price can change is known exactly. The tenant can also give a month's notice and leave. The reset date and the exit date are two different dates about one contract, and knowing one of them tells nothing at all about the other. A bank's balance sheet is thousands of contracts with exactly that property, and so it gets matched on two axes.

Axis one asks when the money leaves. The question is about cash: on any given day, is enough coming in to cover what is going out? The cash question is what the bank's maturity ladderThe table that asks when cash leaves, bucket by bucket, from the shortest period to the longest. is for, and Vindhya sets it out in eight buckets it numbers LB1 to LB8, from the first fortnight out to beyond five years. Get this axis wrong and the bank runs out of cash. Running out of cash is the failure that kills a bank in days rather than years.

Axis two asks when the rate changes. The question is about pricing: when the level of interest rates moves, how much of each side follows it, and how quickly? The pricing question is what the bank's repricing ladderThe table that asks when the rate on a balance changes, bucket by bucket, whether or not the balance itself goes anywhere. is for, in eight buckets numbered RB1 to RB8. Get this axis wrong and the bank does not run out of cash at all. The bank simply earns less than it planned, or the value of what it holds falls further than the value of what it owes, and both of those show up in a report rather than at a counter.

Why can a bank not simply match every asset to a liability of the same length?

Here is the question every reader asks at this point, and it deserves a straight answer rather than a shrug. If the mismatch causes all this trouble, why not fund a ten year loan with a ten year deposit and be done with it? Somebody could build that bank. Such a bank would be extremely safe and it would be pointless.

Think about the street corner. A moneylender who borrows from one person on Monday and lends the identical amount to another person on Monday, for the identical period, at the identical date, is not really doing anything except passing an envelope along and taking a small fee for the trouble. People need something different from a bank on each side. Depositors want their money available; borrowers want years. The bank stands between two populations who want opposite things, and the whole of its economic reason to exist is that it is willing to hold the difference between them. The willingness to hold that difference is what people mean by maturity transformation, and it is managed rather than removed.

Who actually does the matching, and what does that committee decide?

Asset liability management is not a spreadsheet and it is not a person. Asset liability management is a standing meeting with a mandate, and at Vindhya that meeting is the asset liability management committee, numbered G4 in the bank's own governance list. G4 has 9 members, it meets every month, and it is chaired by the chief executive. A chief executive in the chair shows how the bank rates the meeting: this is not a technical review passed down to a specialist. Devendra Achar, head of treasury, sits on it and runs the day to day position between meetings.

Now the part people skip. G4 does not compute the measures. The measures are computed by people with models, and G4 reads them. G4 does one thing nobody else in the bank does: it sets the behavioural assumptions those measures are computed on. Two of them matter enormously and run through the rest of this guide: the average life the bank assumes for its Rs 36,000 crore of non-maturity deposits when it works out what a rate move does to value, and the share of that same Rs 36,000 crore the bank places in the shortest maturity bucket LB1. Both are decisions. Neither is a measurement. Both are taken by the 9 people in that room, once a month.

Compare that with the board risk management committee, numbered G2. G2 has 5 members, 3 of them independent directors, meets six times a year, is chaired by an independent director, and sets every limit the bank runs against, L1 through L12. So one committee sets the assumptions and a different committee sets the tests. Hold that thought.

What a matching paper has to contain

A committee is only as good as the paper in front of it, and a matching paper has four fixed parts. Three of them get written every time because their absence is obvious. The fourth goes missing constantly. Its absence looks like tidiness rather than thinness.

What is a behavioural assumption, and why does an answer depend on one?

An ordinary savings account makes the point. The holder can empty it this afternoon and nobody can stop them. The question worth asking is how much of it has actually moved in the last three years. For most households the honest answer is very little, and the balance that has sat there through two job changes and a wedding is behaving nothing like money that is about to leave. Contractually available and behaviourally likely to leave are two entirely different claims, and a ladder needs the second one.

The gap between the two claims is where a behavioural assumptionA decision about how money will actually move when the contract does not say, made by named people at a named meeting. lives. The bank cannot read the answer off a contract, so somebody decides it, and the deciding is called slottingPlacing a balance into a time bucket in a table, which is a judgement wherever the contract is silent about the date. when it puts a balance into a bucket of a table. A behavioural assumption is not a forecast anybody can check tomorrow. Vindhya's own model inventory numbers the behavioural deposit life model V1 and records it as one of three models never put through validation. The honest reason is uncomfortable: how long a deposit stays is only observable over years, so V1 cannot be tested next Tuesday.

So the bank is placing Rs 36,000 crore, being 46.9 per cent of its deposits, using a judgement it cannot check quickly, taken monthly by 9 people. Say it plainly and it sounds alarming. The practice is not alarming, only unavoidable. Every bank on earth does this, and the only question that separates a well run one from a badly run one is whether the judgement is written down, attributed to a named committee, and shown beside every number it decides.

Why does the same Rs 36,000 crore sit in four tables four different ways?

The part worth carrying away needs its warning first. The four tables ask four different questions, so one deposit book is treated four completely different ways inside one bank in one month, and all four treatments are correct. A reader who meets the second treatment without that warning concludes somebody has made a mistake, and nobody has.

The fourth row is the one that surprises people, so here is where its number comes from. Vindhya's own thirty day assumptions treat the two halves of the non-maturity book differently: 7.5 per cent of the Rs 26,400 crore of savings balances, being Rs 1,980 crore, and 40.0 per cent of the Rs 9,600 crore of current balances, being Rs 3,840 crore. The two add to Rs 5,820 crore of assumed outflow. Divided by the Rs 36,000 crore book, the blended assumption is 16.2 per cent. Every one of those factors is this invented bank's own working number. None of them is a requirement, a market convention or a figure anybody must use, and what an Indian bank actually applies is a matter for the source named below.

The first row and the fourth row are the pair that does the teaching, so put them beside each other. The cash table assumes Rs 1,800 crore of that deposit book leaves in the first fortnight; the thirty day computation assumes Rs 5,820 crore of it leaves inside a month. The second assumption is 3.23 times the first as a rate, and the two differ by Rs 4,020 crore. Both sit in papers going to the same committee. Both are right. The first is the bank's view of an ordinary fortnight and the second is the bank's view of a severe month, and a reader who does not know that has every reason to think one of the two is a typing error.

There is one more thing worth noticing about the third card. The economic value computation gives that deposit book an average life of half a year. The repricing ladder puts the identical money in a bucket that runs from one year to three. The bank's own two tables therefore disagree about the same deposits by a factor of somewhere between two and six, and no document anywhere in the institution reconciles them. The measure that sits on top of the 0.5 year assumption, the change in the economic value of what the bank holds against what it owes, is computed separately. A number of that importance rests on an assumption 9 people set at a monthly meeting.

How far would one assumption have to move before a limit is breached?

Everything so far has been description. Now watch the assumption do actual work, on the first of the four treatments and on the smallest of the four numbers.

Vindhya's shortest maturity bucket LB1, covering the first fortnight, shows inflows of Rs 7,200 crore against outflows of Rs 9,600 crore. The gap is minus Rs 2,400 crore, being 25.0 per cent of that bucket's outflows. The bank's own limit L9, set by committee G2, says the negative gap in LB1 does not exceed 28.0 per cent of that bucket's outflows. So the bucket sits at 89.3 per cent of its limit and the paper reports it as comfortable. Of the Rs 9,600 crore of outflows in that bucket, Rs 1,800 crore is there for no reason other than that committee G4 decided to put 5.0 per cent of the non-maturity deposit book into it. The other Rs 7,800 crore is contractual: term deposits maturing, borrowings falling due, payments the bank has committed to.

Now change nothing on the balance sheet. Not one rupee moves, not one customer does anything, not one contract is signed. Move only the slotting share, and call the amount the committee places in LB1 the slotted amount.

The second column takes three lines of algebra. With a slotted amount of Rs A crore, outflows are 7,800 plus A and the gap is 600 plus A, so the ratio the limit tests is (600 plus A) divided by (7,800 plus A). Setting that equal to 0.28 gives 600 plus A equals 2,184 plus 0.28A, so 0.72A equals 1,584 and A is Rs 2,200 crore. The forward check: outflows Rs 10,000 crore, gap minus Rs 2,800 crore, and 2,800 over 10,000 is 28.0 per cent to the rupee.

The whole distance between comfortable and breached is 1.1111 percentage points of a behavioural assumption, being Rs 400 crore of a Rs 36,000 crore deposit book. Be precise about what happens at the crossing itself. At exactly 6.1111 per cent the utilisation is exactly 100.0 per cent, the cap reached rather than exceeded, and anything beyond it is a breach. One committee, meeting monthly, moving one judgement by a hair, takes another committee's limit from a number nobody looks at twice to a number that has to be reported, explained and remediated. Nothing on the balance sheet has changed.

How does a limit set by one committee end up decided by another committee's assumption?

Both halves of the problem are now in place, and they belong together. The assumption that decides where the LB1 gap sits is set by committee G4, monthly, chaired by the chief executive. The limit that tests it, L9, is set by committee G2, six times a year, chaired by an independent director. The paper G2 reads carries the output of G4's decision and never the decision itself. G2 sees a bucket at 89.3 per cent utilisation. G2 does not see, anywhere in its pack, the sentence that says this number rests on a 5.0 per cent slotting judgement and would reach its cap at 6.1111 per cent.

The fix is unglamorous and cheap. The fix is a single line added to part four of the paper: this figure is computed on a 5.0 per cent slotting assumption set by committee G4 on such and such a date, and it reaches its cap at 6.1111 per cent. No new committee, no new system, no new model. One sentence, placed beside the number it explains, in the pack of the body that reads it.

Who is accountable when the funding side concentrates?

Everything above has been about the asset side being long and the liability side being short. There is a second thing that goes wrong on the liability side and it has nothing to do with timing: the funding gets concentrated. At month 12 Vindhya has two live breaches of exactly that shape, and both of them have the same name attached.

Limit L10, the bank's own, caps wholesale funding at 20.0 per cent of total liabilities. Total liabilities are the Rs 96,000 crore balance sheet less Rs 7,680 crore of equity, being Rs 88,320 crore, so the cap in rupees is Rs 17,664 crore. Running against it are Rs 11,520 crore of wholesale term deposits plus Rs 8,400 crore of borrowings, being Rs 19,920 crore, or 22.6 per cent of liabilities. Divide the rupees rather than the percentages and utilisation is 112.8 per cent. The breach is B3, first crossed in month 11 at 21.4 per cent, and its cause is plain: retail term deposits ran off and the bank replaced them with certificates of deposit.

Limit L12 caps the top twenty depositors at 12.0 per cent of total deposits, being Rs 9,216 crore of the Rs 76,800 crore deposit book. The top twenty actually hold Rs 11,136 crore, being 14.5 per cent, so utilisation is 120.8 per cent. The breach is B4, crossed in month 10 when one state undertaking placed Rs 1,440 crore in a single deposit. Both breaches are concentrations in how the bank funds itself, one by funding type and one by depositor, and both were still open at month 12. Two breaches left open is not slackness. A breach caused by the structure of a balance sheet does not run off by itself the way a trading position does.

And both carry a name. Devendra Achar, head of treasury, is the risk ownerThe single named person accountable for a breach, so that the breach has an address rather than sitting with a department. of B3 and B4. Think about what a named owner changes. A table saying wholesale funding is at 22.6 per cent is information. A table saying wholesale funding is at 22.6 per cent, against a 20.0 per cent limit, and Devendra Achar is accountable for bringing it back, is an obligation with a person at the end of it.

What can the committee actually do about a position it does not like?

A reader who has followed this far has a fair objection waiting. Committees, assumptions, limits and papers are all very well, but what does anybody actually do on the Monday after the meeting? The honest answer is that the list of levers is short, every one of them costs something, and knowing which lever moves which axis is most of the craft.

Read the second column and one thing jumps out: the derivative is the only lever that moves the pricing axis without disturbing the cash axis, and that alone explains why banks use them at all. Everything else on the list drags both axes at once. An asset liability management committee that wants to fix a repricing problem by lengthening its funding usually finds it has also changed a maturity bucket it was quite happy with. How a swap actually pays, how it is priced and what it settles into is covered separately; it earns its place in the table as the only entry with a single tick in one column.

Now apply that to the shortest bucket. Vindhya does not like a minus Rs 2,400 crore gap in LB1, so what closes it? Bringing inflows forward, by holding more of the buffer in instruments that mature inside the fortnight. Pushing outflows back, by lengthening the wholesale money that falls due in it, though that is exactly the funding whose concentration is already in breach as B3. Or holding more cash, the least earning thing the bank can do with a rupee. Changing the slotting assumption does not close it, and the difference between those two kinds of action is the single most useful distinction in the subject. Moving the assumption changes the measured number and changes nothing about the bank. Moving the funding changes the bank and, as a consequence, changes the number. A committee that reaches for the first when it means the second has not fixed anything; it has redecorated the report.

The monthly rhythm earns its keep here too. G4 meets twelve times a year and G2 six, so every position described here gets looked at by somebody roughly every four weeks and every limit gets reset roughly twice a year. The cadence is not decoration either. A funding shape takes months to change, so a committee meeting quarterly would be reacting to a position two quarters old, and a committee meeting weekly would be making structural decisions at the speed of noise.

How does anybody outside the bank read this?

An outsider sees none of this unless somebody asks for it, so how the asking is done is worth knowing. Four readers, four different questions.

A wholesale lender deciding whether to place money with Vindhya reads the ladder and then asks what the ladder rests on. The competent version of that conversation lasts one question: what average behavioural life is the bank using on its non-maturity deposits, who set it, and when did they last change it? A treasury that answers instantly, with a committee name and a date, is showing that the judgement is governed. A treasury that has to go and find out is showing something else.

An equity analyst asks which direction the bank is managed in. A rate rise helps a bank whose assets reprice faster than its liabilities and hurts the value of a bank with long assets against short liabilities, and both of those can be true of the same institution at the same moment. So the question is not which number is right. The question is which of the two the bank steers by, and whether anybody has been shown both.

A large depositor should notice that concentration is a two sided fact. The state undertaking that placed Rs 1,440 crore with Vindhya in month 10 caused breach B4 without doing anything wrong. A corporate treasurer placing a deposit large enough to matter to the bank taking it is somebody else's concentration limit, and that is worth knowing before the bank starts pricing that deposit differently to reduce it.

And a household reader is inside these tables already. The balance sitting in a household savings account is part of somebody's Rs 36,000 crore, and a room of 9 people has decided how long it is likely to stay. A room of 9 people deciding is not sinister; it is the only way the account can pay anything at all. Money that had to be available in cash the instant it was asked for, with no assumption made about it, could not be lent to anybody, and an account that lends to nobody pays nothing.

What can asset liability management not fix?

A function with a clear mandate has a clear edge, and saying where the edge is protects the function from being blamed for things outside it. Four things sit outside.

Matching cannot make a badly chosen loan book good. Matching is about when money moves and when rates reset, not about whether the borrower pays. Vindhya's Rs 1,764 crore of non-performing advances is not an asset liability management problem and no amount of reslotting touches it.

Matching cannot conjure funding that is not there. Matching can change the shape of what is there, lengthen some of it, price some of it differently and refuse some of it. When the funding simply is not available, that is a different conversation with a different set of actions, and it belongs to the plan a bank keeps for exactly that.

Matching cannot make an assumption true. Placing Rs 34,200 crore of the deposit book in the long buckets LB5 to LB8 does not persuade a single depositor to stay. The table records a belief about behaviour; it does not cause the behaviour. The distance between recording a belief and causing it is the honest reason model V1 matters so much and can be tested so little.

And matching cannot decide how much exposure the bank wants. The board approves the appetite, committee G2 turns it into limits L1 to L12, and committee G4 works inside them. A committee that sets its own tests is not being governed, and the separation that creates the structural gap described here is also the separation that stops that happening. The fix is a line in a paper, not a merger of the two mandates.