A bank is the one business where the balance sheet is not a support structure for the operations. It is the operations.
Everything else a company owns exists so that it can sell something. A bank's assets are the product: it buys money from depositors at one price, sells it to borrowers at another, and keeps the difference less whatever it loses on the loans that do not come back. That single structural fact is why almost none of the valuation habits that work for a manufacturer transfer, and why the industry settled on price to book instead of price to earnings a long time ago.
Why not price to earnings
Two reasons, and the second is the one that matters.
The first is that a bank's earnings in any single year are a function of provisioning, and provisioning is a judgement. A bank that under-provides reports higher profit this year and lower profit for the next three, and nothing in the income statement distinguishes it from a bank that genuinely had a better year.
The second is that a bank's ability to grow is limited by its capital. Every rupee of new lending consumes equity under the capital adequacy rules, so growth is not free in the way it is for an asset-light business. A bank earning ₹720 crore on ₹4,600 crore of equity can only expand its loan book as fast as its retained profit and any capital it raises allow. Book value is therefore not an accounting curiosity. It is the constraint on the business, and valuing the constraint is a more direct question than valuing one year's output.
The relationship that does the work
Price to book and return on equity are not two separate metrics. They are linked by arithmetic.
For a business in a steady state, the justified multiple of book value is the return on equity minus the growth rate, divided by the cost of equity minus the growth rate. Read plainly, it says that a bank earning exactly its cost of equity is worth its book value and no more, and everything above one times book has to be earned by a return that exceeds what shareholders require.
Work an example. A bank earning 15.7% on equity, with a cost of equity of 13% and sustainable growth of 8%, justifies a multiple of 1.54 times book. On ₹4,600 crore of equity, that is ₹7,084 crore of value.
Change the return to 9% and the same formula gives 0.2 times book. Change it to 20% and it gives 2.4 times. The multiple is extremely sensitive to the return, which is exactly right: a bank that earns more on the same book is worth more, and the relationship is not linear.
This is why the standard first exercise in financial-sector analysis is a scatter chart. Put return on equity on one axis and price to book on the other for every listed bank, and the points fall close to a line. The banks off the line are the work: either the market disagrees with the reported return, or it expects it to change.
Where the return on equity comes from
Return on equity for a bank decomposes cleanly into return on assets multiplied by leverage.
A bank earning ₹720 crore on ₹60,000 crore of assets returns 1.2% on assets. With ₹4,600 crore of equity, it is levered 13.0 times, and 1.2% times 13.0 gives 15.7% on equity.
Both terms deserve separate examination and only one of them is skill.
For an Indian bank, return on assets above 1% is respectable and above 1.5% is strong. Treat those as local reference points rather than universal ones. In Japan and much of Europe the whole distribution sits lower, and a bank there can be well run at levels that would look poor here, so the comparison has to stay inside one market. Within that market it is the honest measure of whether the bank is good, because it is taken before the leverage.
Leverage is a choice constrained by regulation. A bank can raise its return on equity by running thinner capital, right up until the regulator or a credit cycle objects. Comparing two banks on return on equity without looking at their capital ratios rewards the one taking more risk.
The four numbers that decide whether the return is durable
A return on equity is a fact about last year. What it is worth depends on whether it repeats, and four disclosures carry most of that answer.
Net interest margin. Net interest income as a percentage of average earning assets. A bank with ₹1,750 crore of net interest income on ₹48,000 crore of earning assets runs 3.65%, which sits in the ordinary Indian range of 3% to 4.5%. What matters is why. Margin earned from a low-cost deposit base is durable. Margin earned by lending to riskier borrowers is a loan against the future, and it is repaid in credit cost two or three years later.
CASA ratio. The share of deposits sitting in current and savings accounts, which pay little or nothing. This single ratio explains more of the difference in profitability between Indian banks than any other. A bank with 45% of its funding at near-zero cost can lend at the same rate as one with 25% and earn far more, and the franchise takes decades to build. It is the closest thing to a moat that exists in lending.
Asset quality, read as a flow. Gross bad loans are a stock and they lag. Slippages, meaning loans that turned bad during the period, are the flow and they lead. A bank can report a falling gross figure while slippages rise, simply by writing off faster. Read the special-mention buckets, which show accounts overdue by 30 to 90 days, because those become bad loans next quarter.
Provision coverage. A bank with ₹1,260 crore of gross bad loans that has provided ₹882 crore has 70% coverage, which is adequate. The signal to watch is coverage falling while bad loans rise, because that combination means the reported profit is being supported by under-provisioning, and the charge has been deferred rather than avoided.
The capital constraint, which sets the growth rate
Regulatory capital as a percentage of risk-weighted assets is what allows the bank to keep lending. A bank at ₹6,900 crore of eligible capital against ₹60,000 crore of risk-weighted assets reports 11.5%, and the Indian minimum including the buffer sits near that level, with a lower floor for the core equity tier.
A bank operating close to its minimum cannot grow its loan book without raising capital, whatever its growth story says. Raising capital at below book value destroys value for existing shareholders, which is why weak banks find themselves in a trap: they need capital to grow out of their problems and can only get it at prices that make the problems worse.
So the sustainable growth rate is not an independent assumption. It is return on equity multiplied by the retention ratio, and a model that assumes 18% loan growth from a bank retaining 70% of a 15.7% return is assuming an equity raise that the model should show.
The residual income route, which is the same idea done properly
Discounted cash flow does not work naturally for a bank, because free cash flow is hard to define when capital is the raw material and debt is a product rather than a financing choice.
The alternative that fits is residual income. Value the equity as its current book value plus the present value of all future earnings above the charge for equity capital.
A bank with ₹4,600 crore of book value earning ₹720 crore against a 13% cost of equity generates ₹122 crore of residual income in the year, and capitalising that stream gives the premium above book. Most of the answer sits in the book value, which is observed, rather than in a terminal value, which is assumed. That is the method's whole appeal.
It also makes the value-creation logic explicit in a way the multiple hides: a bank earning exactly its cost of equity is worth its book, and everything above that has to be earned.
What breaks the analysis
Restructured loans. Loans whose terms were changed because the borrower could not meet the originals are disclosed separately from bad loans. A bank with 3% gross bad loans and 4% restructured has 7% of its book in difficulty. Add the two before judging asset quality.
The investment book. A bank holding a large portfolio of government securities carries interest rate risk that does not appear in any lending ratio. When yields rise, the mark-to-market loss lands in other comprehensive income or in profit depending on the classification, and for a bank with a large book it can exceed the year's earnings. Read the movement in reserves, not only the profit.
Unseasoned growth. A loan book that grew 40% in a year has not yet had time to default. Credit cost on a young book is structurally understated, and the return on equity computed from it is a number about a portfolio that has not been tested.
Book value that is not book value. If provisioning is inadequate, the equity figure in the denominator is overstated, and every ratio built on it is wrong in the same direction. This is why asset quality analysis comes before valuation rather than after it.
A working sequence
Start with return on assets across five years, because it is the cleanest measure of quality and it strips out leverage.
Then decompose it: net interest margin, fee income, operating cost as a share of income, and credit cost. Each of those has a peer benchmark and each moves for different reasons.
Then check the funding: CASA ratio, cost of funds, and the credit-deposit ratio, which above 90% means growth is being funded by borrowings rather than deposits.
Then asset quality as a flow, not a stock, with the special-mention buckets and the restructured book beside the headline.
Then capital, which sets how fast any of it can grow.
Only then compute price to book and set it against return on equity, on a chart with the peer set. A bank sitting well below the line either deserves to be there because the market doubts its reported return, or it is the opportunity, and the five steps above are how you tell which.
The one sentence to keep
A bank is a levered bet on its own underwriting, funded by other people's money and constrained by regulatory capital.
Every valuation question reduces to whether the underwriting is as good as reported, whether the funding is as cheap and as sticky as reported, and whether there is enough capital to keep doing it. The multiple follows from those three answers rather than leading them.
What a bank's income statement is made of
The waterfall is short and each line has a different quality.
Net interest income is the core: interest earned on loans and investments less interest paid on deposits and borrowings. A bank earning ₹4,200 crore and paying ₹2,450 crore reports ₹1,750 crore. Growth here comes either from more loans at the same spread, which is repeatable and consumes capital, or from a wider spread, which is usually a function of the rate cycle and reverses.
Fee income arrives without consuming capital, which makes it the highest-quality revenue a bank earns. Processing fees, distribution commissions on insurance and mutual funds, card interchange and transaction charges all sit here. A bank whose fee income grows faster than its loan book is improving the shape of its earnings, because it is earning more without needing more equity behind it. The split matters: fees tied to lending volumes disappear when lending slows, while payment and transaction fees do not.
Operating expenses, measured against operating income as the cost-to-income ratio. Below 45% is efficient for an Indian bank and above 55% is a problem, though a bank investing heavily in branches or technology will run a worse ratio for several years before the investment matures.
Provisions, which are a judgement and which decide the reported profit more than any other line.
Four lines. The last one is the one that moves.
The trap that catches new analysts
A bank's return on equity looks best exactly when its underwriting is worst.
Losses fall, so standards relax, so credit grows, so asset prices rise, which appears to validate the lending. Credit costs in that phase are at their lowest, return on equity at its highest, and the loans being written are the ones that will default in three years. Then losses appear from the loosest vintage, standards tighten, and the same bank reports a return that would have been unimaginable at the peak.
Which means judging a lender on a three-year average that sits entirely inside an expansion is judging it on the easy half of the cycle.
Look at what it wrote in the boom, and what it lost in the last downturn. If there has not been one within the data you have, treat every ratio as provisional.
A five-minute version, for triage
Five numbers, from the results release rather than the annual report.
Return on assets, because it strips out leverage. CASA ratio, because it is the funding advantage. Gross bad loans and the direction of slippages, because that is the flow. Provision coverage excluding technical write-offs, because that is the cushion. And the core equity capital ratio, because that is the constraint on growth.
A bank scoring well on all five deserves the full work. A bank failing two of them does not, whatever its price to book.
The sequence, and why the order is not negotiable
Asset quality comes before valuation, and the reason is that every error in asset quality propagates in the same direction. Under-provisioning overstates profit, which overstates return on equity, which overstates the book value that the provisioning was supposed to protect, which makes the price-to-book multiple look cheaper than it is. Four errors, all pointing the same way, all compounding. So a valuation built on a book that has not been examined is a precise answer to the wrong question, and the examination is not optional or advanced work. It is the first hour. Establish that the loans are what the bank says they are, that the deposits are as cheap and as sticky as reported, and that there is enough capital to keep lending. Then, and only then, compute a multiple.
Return on assets first. It strips out the leverage.
Then funding. Then asset quality as a flow.
Then capital, then price.