Reading a bank's balance sheet, where deposits are the raw material

For a manufacturer, the balance sheet is a support act. The profit-and-loss account is where the business happens - goods sold, margins earned - and the balance sheet just shows the plant and working capital that made it possible. For a bank, that relationship is reversed. The balance sheet is the business. A bank takes in deposits and turns them into loans; its entire operation is the assembling and management of that balance sheet, and the profit is a thin margin skimmed off the top of a very large pile of other people's money. illustrative

That matters because a bank almost never fails on its profit-and-loss account. It fails on its balance sheet - on loans that stopped being repaid and losses it did not set aside for - and by the time that reaches the P&L, it is usually too late. To judge a bank early, you read the balance sheet, and you read it with instincts a manufacturer's would get backwards. Deposits sit on the funding side and might look like borrowing, but they are the raw material - the cheaper and stickier they are, the better the business. Loans, called advances, are the earning asset, and their quality is almost the entire question. Net worth is a thin sliver, perhaps a tenth of the balance sheet, because a bank is deliberately levered around ten times - that is the design, not a warning.

Four ideas do the work, and none has a real manufacturer equivalent. The CASA ratio - the share of deposits in low- or no-interest current and savings accounts - measures how cheap the raw material is. The non-performing asset figure measures how much of the loan book has stopped performing. The provision coverage ratio measures how much of those bad loans the bank has already owned up to. And capital adequacy measures how much capital stands behind the whole thing to absorb losses - the one number the regulator watches most closely. Get those four and a bank's balance sheet stops being a foreign document.

Look at the shape first, then read it.

Two columns, equal height, like any balance sheet - but every block reads its own way.

How cheap is the funding? The largest block by far is deposits. A manufacturer's instinct screams "that is a huge liability", but it is the opposite of a problem: deposits are the cheap money the bank lends out at a higher rate, and the whole business is the spread between the two. What matters is not their size but their cost and stickiness. Current and savings accounts (CASA) pay little or no interest and tend to stay, so a high CASA share means cheap, stable funding. Bulk term deposits are expensive and flighty. A bank growing deposits by piling on costly term money is growing its size, not its quality.

How good are the loans, and how honest is the book? The largest asset is advances. Their quality is measured by how many have stopped being repaid - the non-performing assets. Gross NPA is the percentage of the book that has gone bad, but it is only half the story, because the bank chooses how much of that loss to set aside for. That choice is the provision coverage ratio. A bank with 3% gross NPA and 75% coverage has owned most of its problem; a bank with 3% gross NPA and 40% coverage is still hiding half of it in future profits. The honest number is the net NPA - the unprovided hole left after coverage, which is simply gross NPA × (1 − coverage).

How big is the buffer? Net worth is deliberately thin, because ten-times leverage is how a bank works. That thinness is why the regulator does not leave the buffer to the bank's discretion: capital adequacy is a regulated minimum, the ratio of capital to risk-weighted assets - loans and investments scaled by how risky each is. A bank barely above the floor has almost no room to absorb a bad year; one well above it can take losses and keep lending. The distance above the minimum, not merely clearing it, is the margin of safety - a bank that breaches the floor does not get a second chance to have held more capital.

Read those blocks - cheap-and-sticky funding, well-provided loans, a fat capital buffer - and you have read the bank. Debt-to-equity, the manufacturer's headline, is meaningless here. The profit will follow from the four; it rarely leads.

The bank balance sheet does not just add new lines - it inverts the meaning of ones you already knew. Set it beside the businesses whose lines it borrows and reverses.

The inversion worth holding onto is that a bank reverses two of the most basic balance-sheet instincts at the same time. On the funding side, a very large liability (deposits) is good, not bad. On the asset side, something called loans is an asset the bank owns, not money it owes. A reader who keeps a manufacturer's reflexes - big liabilities are dangerous, loans are debt - will misread both sides in the same glance. The manufacturer cell marks the baseline: for a cement maker, borrowings really are a risk and high leverage really is a warning. The bank is not a harder version of that. It is the reverse of it.

Read the composite private bank's balance sheet in its fifth year. Total assets of ₹338,000 crore sit against net worth of ₹33,200 crore - about ten times levered, which is normal and healthy. Now run the four questions. illustrative

Funding. Deposits of ₹282,000 crore fund most of the book, and 46% of them are CASA - cheap money that pays little interest and tends to stay. A high, rising CASA share is why this bank's cost of funds is low, the quiet foundation of its margin. If CASA were falling while deposits grew, you would know the growth was bought with expensive term money and the margin was under pressure regardless of the headline.

Loans and honesty. Advances of ₹230,000 crore are the earning asset. Gross NPA is 2.0% - two rupees in every hundred lent have gone bad - and coverage is 78%, so more than three-quarters of that loss is already provided for. The net NPA, the unprovided hole, is just 0.5%: a well-owned book. Contrast the bank's own third year, when a credit-cycle stress pushed gross NPA to 3.2% and coverage down to 66% - the same balance sheet, a worse and less honest asset picture, and the year profit dipped as provisions rose.

Buffer. Capital adequacy is 17.1%, comfortably above the roughly 11.5% regulatory minimum. That gap is the margin of safety - room to absorb a bad year and keep lending rather than raise capital at the worst moment. A bank at 11.6% would report the same "meets requirement" and be in a completely different position.

The habit to build: for any bank, ignore debt-to-equity entirely and read four things - the CASA share, the gross NPA, the provision coverage, and the capital adequacy against the floor. Those four, tracked over several years, tell you whether you are looking at a sound lender or a fragile one, long before the profit-and-loss account admits it.

The balance-sheet ratios tell you the state of the book today; they do not tell you what is about to walk into it. Gross NPA and coverage describe the loans that have already gone bad and been recognised. The loans that will go bad next year are still sitting in the "standard" book, indistinguishable on the face of the balance sheet from the good ones. A bank that has lent aggressively into a sector about to turn can show pristine NPA numbers right up until the slippage begins. The balance sheet is a snapshot of recognised trouble, not a forecast of coming trouble - and that gap is where credit cycles do their damage.

Nor can these ratios, on their own, catch a bank that is delaying recognition. NPA and provisioning depend on the bank actually classifying a loan as bad, and there are ways to postpone that: evergreening a struggling borrower with a fresh loan to keep the old one current, restructuring rather than recognising, lending to a related entity that repays just enough to stay standard. A book can look clean because the losses are real but unrecognised. So reading the disclosed numbers is necessary; noticing when they are suspiciously stable through a stressed cycle - and cross-checking against the restructured-book and slippage disclosures in the notes - is what catches the delay. When a management insists "asset quality remains robust" but will not give a figure for the restructured book or its expected credit cost, the silence is the tell: that is the exact claim a bank postponing recognition would also make, and it is unfalsifiable without the numbers the answer withholds.

Finally, the headline ratios do not reveal concentration. Two banks with identical 2% gross NPA carry very different risks if one has spread its book across thousands of small borrowers and the other has lent heavily to a handful of large groups or one cyclical sector. Concentration lives in the disclosures, not the headline - and a diversified 2% book and a concentrated 2% book are not the same bank.

The first trap is reading a bank with a manufacturer's instincts. Deposits look like a mountain of liabilities, loans sit on the asset side where they feel wrong, net worth is thin. Every one of those reflexes is inverted for a bank: deposits are the raw material, loans are the product, and thin net worth against a ten-times balance sheet is the design. A reader who does not consciously switch templates concludes that every bank is a leveraged disaster, and is wrong every time.

The second trap is trusting a low gross NPA without reading the coverage behind it. A 2% gross NPA looks better than a 3% one, and often it is - but not if the 2% book has set aside only 40% against its bad loans while the 3% book has set aside 78%. The under-provided bank has a smaller recognised problem and a larger hidden one, because the unprovided losses are still sitting in future profits waiting to land.

The third trap is treating "meets the capital requirement" as safety. Capital adequacy is a regulated floor, and clearing it is a minimum, not an achievement. A bank at 11.6% against an 11.5% requirement has essentially no buffer: one bad year can push it below the floor and force a capital raise exactly when capital is most expensive and dilutive. A bank at 17% can absorb the same losses and keep lending. Both "pass", and reading the pass rather than the distance above the floor misses everything.

So before you conclude that a bank with a low gross NPA is the safer one, state what would change your mind: coverage far below its peers, a capital position a whisker above the floor, or restructured loans parked just outside the NPA definition - any one of those flips "clean book" into "problem deferred". If you cannot name, in advance, the number that would move you from reassured to worried, you were trusting the headline, not reading the bank.