Provision Coverage: How Much of a Bad Loan Is Absorbed

Provision coverage is the share of a bank's bad loans, measured gross, that its provisions have already absorbed. The base is gross non-performing advances and nothing else. Suvarna Commercial Bank Limited, an invented lender, carries Rs 4,536 crore of provisions behind Rs 6,480 crore of gross non-performing advances, so its coverage is 70.0 per cent, on gross non-performing advances, for the stated year.

Five things decide how this measure lands in a real set of accounts, and each of them belongs to an authority. The minimum provision attaching to each class of advance, the rules that decide which class an advance falls into, what happens to an advance whose terms have been rewritten, what happens to one the lender has taken off its own books, and whether any floor sits under coverage itself are all settled by the Reserve Bank of India, and every one of them gets revised. Rows for the five appear further down, drawn, named and handed over with nothing typed inside them. The division itself does not move, and no revision touches it. Every rupee below belongs to a single bank, reported once for one stated year, and each figure was worked again at the spot it is used rather than carried across from anywhere. Adequacy is a separate question, and answering it needs the very minimums the Reserve Bank of India sets.

Start with what a provision already did. The measure is meaningless until that part is settled. A provision is money a bank has set aside against an advanceMoney a bank has lent and is still owed. A loan is something owed to the bank rather than by it, so the lending book sits on the asset side of a bank's balance sheet. it expects will not come back in full. Setting it aside is not a plan or an intention. The setting aside is an accounting event that has already happened: profit for the period was reduced by that amount, and the advance is now carried on the books at less than its face value. By the time a provision figure is being read, that part of the loss has already been taken.

Here is the household version, and it is exact rather than merely friendly. A scooter has started making a noise and its owner puts Rs 8,000/- aside for the repair. The Rs 8,000/- has gone out of this month's spending whatever the mechanic eventually charges. If the bill comes to Rs 8,000/- the household is square; if it comes to Rs 11,000/- it is Rs 3,000/- short and will feel it then, not now. The money set aside is a past event and the bill is a future one, and the two are not the same size by accident or by design. Money set aside for the scooter repair is a provision, exactly.

So coverage asks a narrow question: of the bad loans this bank has already identified, how much has it already paid for? How much has been paid for is a completely different question from how many bad loans the bank has, and the whole of the difficulty with this measure comes from readers answering the second question with the first one's number.

What is a provision doing on the books before anything is divided by anything?

A provision does two things at once, in two different statements, and both are worth naming before any ratio arrives. In the profit and loss account a provision is a charge. Suvarna Commercial Bank Limited's operating profitWhat a lender earns from its ordinary business over a period, counted before the charge for expected loan losses and before tax. of Rs 4,800 crore for the stated year becomes profit before tax of Rs 3,000 crore once Rs 1,800 crore of provisions is taken off it. On the balance sheet a provision is a deduction. The advances of Rs 1,44,000 crore the bank reports gross are carried at less than that once the provisions held are taken off.

Notice that those are two different numbers. Rs 1,800 crore was charged during the year. Rs 4,536 crore is standing there at the end of it, accumulated across however many years the bank has been setting money aside. A reader who treats those two figures as versions of each other has confused an amount that arose during a period with an amount that stands at a date, and every serious error on this measure starts there.

The advances at the centre of all this are the ones the bank has classified as non-performingAn advance the bank has classified as one where repayment has stopped or slipped far enough behind that the bank no longer treats it as normal. Which advances fall into that class, and at what point, is fixed by rules the regulator sets.. Suvarna Commercial Bank Limited reports Rs 6,480 crore of gross non-performing advances, and set beside a gross lending book of Rs 1,44,000 crore that comes to 4.50 per cent, on gross advances, for the stated year. The Rs 6,480 crore is the pile the provisions are being held against.

What exactly is provision coverage a percentage of?

Provision coverage is provisions held divided by gross non-performing advances. One line is the whole definition, and the important half of it is the second half. Provision coverage is defined by its denominator and by nothing else, so a coverage figure quoted without its base is not yet a figure at all.

The division is worth working once, slowly, in the order the numbers arrive. The bank has identified Rs 6,480 crore of gross non-performing advances. Against those it has set aside, over time, Rs 4,536 crore. The second divided by the first is 0.70, or 70.0 per cent. The figure is worth stating in full every time: coverage of 70.0 per cent, on gross non-performing advances, for the stated year. The full sentence takes four extra words and removes every ambiguity there is.

Which three numbers could sit under the line, and why is only one of them right?

There are three plausible denominators for a provisions figure and readers reach for all three. The arithmetic does the arguing, so laying the three out beside each other is quicker than warning about them.

The first is gross non-performing advances, and that one is coverage. Rs 4,536 crore over Rs 6,480 crore is 70.0 per cent, and it answers the question how much of the identified bad loans has already been paid for.

The second is gross advances, the whole lending book. Rs 4,536 crore over Rs 1,44,000 crore is 3.15 per cent of gross advances. The 3.15 per cent is a perfectly real number and it is not coverage. The ratio answers a different question: what share of everything the bank has lent is now sitting behind a provision. Its denominator is more than twenty times larger, so the figure comes out much smaller. Same numerator, different base, different measure, and calling the second one coverage is the commonest way this figure gets misquoted.

The third is net non-performing advances, and this one is not merely a different measure but a broken one. Net non-performing advances are what is left after the provisions have been taken off. Dividing the provisions by that leftover gives Rs 4,536 crore over Rs 1,944 crore, or 233.33 per cent. A share cannot exceed the whole, and the reason the arithmetic produces a number above one hundred is that the denominator has already had the numerator removed from it. The division measures nothing, and the fact that it still returns a tidy-looking percentage is exactly why somebody prints it once a year.

What is the difference between provisions held and provisions charged?

Most readers lose the thread at exactly this point, and the fix is one distinction held firmly. Provisions held is a stock: everything set aside up to a date and not yet used up, standing on the balance sheet at that date. Provisions charged is a flow: the amount set aside during one stated period and taken through that period's profit and loss account. Stock and flow are different sizes, they carry different time words, and they can move in opposite directions in the same year.

For Suvarna Commercial Bank Limited the stock is Rs 4,536 crore, standing at the balance sheet date. The flow is Rs 1,800 crore, arising during the stated year, and that flow is 1.25 per cent of gross advances of Rs 1,44,000 crore for the year. The stock is 2.52 times the flow. All that tells anybody is that the bank has been setting money aside for longer than one year, as every lender has.

Consider a water tank on a roof. The water in the tank at six in the evening is a stock, and it is one number. The water that went in during the day is a flow, and it is a different number. Yesterday's water did not leave, so less can go in today than went in yesterday and the day can still end with more water in the tank than ever before. A bank charging less this year than last can still be holding more provisions than ever, and a reader who cannot hold that sentence comfortably will misread half the asset quality tables they ever open.

The flow figure has a name of its own once it is expressed against the lending book, and that name is credit costThe charge a lender takes over a period for expected loan losses, usually expressed against its lending book for the same period. Credit cost is a subject in its own right and is covered separately.. Credit cost belongs to the flow column and never to the stock column.

What happens to the net bad loan ratio when coverage is raised?

Two things happen, and seeing only the first is the commonest error made with this measure. When coverage is raised, the provisions held go up. Net non-performing advances are gross non-performing advances less the provisions held, so the numerator of the net ratio falls. Most readers get that much.

Now the half that gets missed. Net advances are gross advances less the provisions held, using the same provisions. So when the provisions rise, net advances fall as well, and net advances are the denominator of the net ratio. The same rupees are being deducted from both, so both sides of the net ratio move, in the same direction, at the same instant.

Take Suvarna Commercial Bank Limited at the 70.0 per cent it reports. The provisions standing behind the bad loans come to Rs 4,536 crore. Subtract that from Rs 6,480 crore and the bad loans measured net are Rs 1,944 crore. Subtract the very same Rs 4,536 crore from Rs 1,44,000 crore and the lending measured net is Rs 1,39,464 crore. Divide the first result by the second and the reading is 1.39 per cent, on net advances.

Now push the setting either way. At 80.0 per cent the provisions come to Rs 5,184 crore, leaving bad loans of Rs 1,296 crore sitting on lending of Rs 1,38,816 crore, and the reading drops to 0.93 per cent, on net advances. Pull it back to 60.0 per cent and the provisions come to Rs 3,888 crore, leaving bad loans of Rs 2,592 crore on lending of Rs 1,40,112 crore, for a reading of 1.85 per cent, on net advances. Every one of those three runs moved what is on top and what is underneath.

The three runs, side by side

Read the bottom row last and read it slowly. Across every one of those settings the gross figures never moved, so the gross ratio sat at 4.50 per cent on gross advances the whole way through. Coverage changed how much of the loss had been paid for and changed nothing whatever about how much had gone wrong.

Why does a moving denominator cause so much trouble?

Because the error it produces does not look like an error. Suppose coverage has gone from 60.0 to 70.0 per cent and the bad loans are worked down to Rs 1,944 crore quite correctly. Rs 1,44,000 crore is the lending figure immediately to hand, so that is what gets put underneath. Out comes 1.35 per cent where the answer is 1.39 per cent. The answer is wrong by 0.04 percentage points, small enough to pass for rounding, and a wrong number that looks like rounding is far more dangerous than a wrong number that looks wrong.

One rule, said out loud once, carries a reader through every asset quality table there is. There are two ratios here and they stand on two different bases. The gross one puts Rs 6,480 crore over Rs 1,44,000 crore and reads 4.50 per cent, on gross advances. The net one puts Rs 1,944 crore over Rs 1,39,464 crore and reads 1.39 per cent, on net advances, that second base being the lending after Rs 4,536 crore of provisions has come out of it. The provisions leave both sides of the fraction, not one. So attach the word gross or the word net to what is on top and to what is underneath, together, in a single breath.

What does provision coverage refuse to tell?

Three things, and they are worth separating because they fail in different directions.

Coverage does not say whether the provision is enough. Enough is a judgement about what will eventually be recovered from a borrower or from whatever stood behind the loan, and a ratio cannot contain a judgement about the future. Coverage of 70.0 per cent against a set of advances where recoveryWhat a lender eventually gets back on an advance that went bad, whether from the borrower or by enforcing against whatever stood behind the loan. How that is pursued is covered separately. turns out to be poor is not enough; the same 70.0 per cent against a set where recovery turns out well is more than enough. The number is identical in both cases.

Coverage does not say how much bad lending there is. Coverage is struck entirely on the bad loans that have already been identified, so it takes its own denominator as given. Asked how big the problem is, coverage answers a different question altogether.

And it says nothing whatever about advances that have not been classified as non-performing, however stressed they may be. The third silence is the largest of the three. The classified pile is Rs 6,480 crore, the rest of the book is Rs 1,37,520 crore, and coverage speaks about the first figure only. Whatever condition the rest of that book is in, good or otherwise, this measure has no view about any of it.

Who decides the minimum provision on an advance?

Not the bank. The honest answer is that a lender does not simply choose how much to set aside against an advance. There is a minimum attaching to each class, there are rules deciding which class an advance sits in, and there are separate treatments for an advance whose terms have been rewritten and for one the lender has taken off its own books. Every one of those belongs to the Reserve Bank of India, and every one of them gets revised.

The body that settles each of those five rows is printed inside the row itself. Each row can be completed from the source on the day it matters. A bank can hold more than a minimum, of course, and whether that is prudence or presentation is not something a ratio will ever reveal.

What does this record not carry, and why say so out loud?

One bank and one stated year, and that is the whole of it. The lending is one figure, split into no segments and no sectors. Nothing sorts it by when it falls due. There is no quarter inside the year and no year on either side of it. Advances whose terms were rewritten are not pulled out separately, and neither are the ones taken off the books. Not one borrower is described. Branches and staff go uncounted. Rs 36,000 crore of the asset side is neither lending nor investments and stays undivided, and the interest bill for the year is one figure rather than one figure split between depositors and everybody else the bank owes.

Naming a gap costs a sentence and prevents the one failure that matters. An invented line would read exactly like a reported one, sitting in the same table, in the same font, reconciling to the rupee alongside figures that genuinely do. Where the record does not hold a figure, the gap is named.

How does anybody use this away from a classroom?

Somebody reading a lender's disclosure does three small things with these two figures, and the order is what makes it work.

First they read the gross ratio and the coverage as two separate facts and refuse to merge them. Gross at 4.50 per cent on gross advances says how much of the lending has gone wrong. Coverage at 70.0 per cent on gross non-performing advances says how much of that has been paid for. One is a size and the other is a completion, and neither substitutes for the other in either direction.

Second they put the flow beside the stock. Rs 1,800 crore charged during the stated year against Rs 4,536 crore held at the end of it tells them the pace of recognition alongside the position reached. A reader with a series in front of them would watch those two together across periods, and one stated year offers no series to watch.

Third they treat every level as somebody else's to publish. The minimum, any floor, and the handling of a rewritten or written off advance all come from the Reserve Bank of India. A person who reads the ratio correctly and looks up the level rather than remembering it will be right for longer than a person who memorises a number once.

The household version of the same discipline is short. A neighbour who says they have put aside most of what the roof repair will cost has revealed something real about their preparation and absolutely nothing about the roof. Coverage is that sentence, said about a lending book.