How to Value a Bank: Book Value, P/B, ROE, NIM

The first time I tried to value a bank, I did the obvious thing: I looked up its P/E ratio, compared it to a soap company, and concluded the bank was "cheaper." It took me an embarrassingly long time to see why that comparison was meaningless. A soap company's raw material is palm oil. A bank's raw material is money itself — and the moment you understand that one sentence, every odd thing about how banks are valued falls into place.

So let's build the whole idea around it. Picture a bank as a shop. But it's a strange shop, because the thing on its shelves — its inventory — is rupees.

A bank's inventory is money

A normal shop buys stock cheap and sells it dearer. A bank does exactly that, except its stock is cash. It buys money cheap — your savings deposit pays you maybe 3%, a fixed deposit a few percent more — and it sells that same money dearer, as a home loan or a business loan at 8–9%. The gap between the two is its gross margin.

That gap has a name every bank investor must know: Net Interest Margin, or NIM — interest the bank earns on its loans minus the interest it pays on deposits, expressed as a percentage of its lending assets. (If the mechanics of how a bank earns are still fuzzy, start with how a bank makes money.) NIM is the bank's shelf margin. A wider NIM means it's buying its inventory cheaper and selling it dearer; a shrinking NIM means competition is squeezing the spread.

Here's why this matters for valuation. For an FMCG company you might glance at EBITDA — earnings before interest. But for a bank, interest isn't a sideshow; interest is the entire business. Subtracting it out would be like valuing a kirana store after ignoring the cost of the goods on its shelves. That single fact is why the usual tools — P/E on its own, EV/EBITDA — quietly stop working for banks, and why a different toolkit takes over.

Why P/E and EBITDA don't travel well

Two reasons, both flowing from the money-as-inventory idea.

First, EBITDA is nonsense for a bank. EV/EBITDA strips out interest to compare core operations. Strip interest out of a bank and you've deleted the business. Nobody serious uses it here.

Second, a bank runs on borrowed money to a degree no normal company would survive. A bank funds itself almost entirely with other people's money — deposits are borrowings — and lends out a multiple of its own capital. Warren Buffett put the danger plainly in his 1990 letter: "The banking business is no favorite of ours. When assets are twenty times equity — a common ratio in this industry — mistakes that involve only a small portion of assets can destroy a major portion of equity." (Berkshire Hathaway, 1990 Chairman's Letter)

Twenty-to-one. A 5% loss on the loan book can wipe out the entire owner's capital. That extreme leverage is why a bank's balance sheet — what it owns and owes — matters more than for almost any other business, and why bank valuation starts not with earnings but with book value.

Start with book value and P/B

Book value (also called net worth or shareholders' equity) is simply what the owners would be left with if the bank collected every loan, paid off every depositor, and shut the doors: total assets minus total liabilities. On the balance sheet it's share capital plus accumulated reserves.

For a bank, book value is roughly the real, marketable thing the business is — a pile of financial assets net of financial liabilities — which is why the headline valuation ratio for banks is Price-to-Book (P/B): market value divided by book value. It answers a clean question: how many rupees are you paying for one rupee of the bank's net worth?

Take ICICI Bank, one of India's largest private lenders (this is an illustration of the method, not a buy or sell call). Its market value is about ₹9,20,611 crore. Its book value — share capital of ₹1,432 crore plus reserves of ₹3,58,946 crore — is about ₹3,60,378 crore (Inve data, 2026). So:

• P/B ≈ 9,20,611 ÷ 3,60,378 ≈ 2.6

You're paying about two and a half rupees for each rupee of net worth. On its own that number is meaningless — a great bank deserves more than one times book, a troubled one trades below it. The number only comes alive when you pair it with the next one.

The number that justifies the price: ROE

A P/B of 2 is only sensible if the bank earns a high return on that book. The metric is Return on Equity (ROE): net profit divided by book value — how much profit the bank squeezes out of each rupee of owner capital every year.

ICICI Bank earned about ₹57,936 crore of net profit over the trailing twelve months on that ₹3,60,378 crore of net worth (Inve data, 2026):

• ROE ≈ 57,936 ÷ 3,60,378 ≈ 16.1%

Now P/B and ROE click together. A bank earning ~16% on equity, year after year, is compounding the owners' money at ~16% — so paying 2.6x book to own that engine can be perfectly rational. A bank earning 6% on equity at 2.6x book is being over-paid for. The pairing is the whole game: P/B is only as justified as the ROE behind it. (For the deeper logic of why return on capital is the quality number, see Return on Capital: The Quality Number.)

One raw figure ties it to your screen: that book value, divided by ICICI Bank's roughly 716 crore shares (face value ₹2), is a book value per share of about ₹503 (Inve data, 2026). The share price quoted against that is just P/B in per-share form.

Asset quality: the hidden trapdoor

Now the part that separates bank valuation from everything else — and the reason a "cheap" bank can be the most expensive mistake you ever make.

A bank's book value is only as honest as its loans. If borrowers stop repaying, those loans are worth less than the balance sheet claims, and book value is a fiction. So before trusting any P/B, you check asset quality — chiefly the Gross NPA ratio (GNPA): the share of the loan book that has stopped paying (NPA = non-performing asset, a loan overdue past 90 days).

This is the bank's version of rotting inventory. A shopkeeper whose shelves are full of spoiled stock isn't really sitting on the inventory the books show — and a bank stuffed with bad loans isn't worth the net worth it reports. A low GNPA means the inventory is fresh; a high one means quiet rot.

The Indian system, for now, looks clean. The RBI's Financial Stability Report put scheduled commercial banks' gross NPA ratio at 2.3% as of March 2025, a multi-decadal low (RBI Financial Stability Report, via News On Air, 2025). ICICI Bank ran below even that, reporting that "the gross NPA ratio was 1.67% at March 31, 2025 compared to 1.96% at December 31, 2024" (ICICI Bank, Performance review: quarter ended March 31, 2025) — a relatively fresh shelf. (These asset-quality figures are as of March 2025, one cycle behind the FY26 book value and ROE above; always line up the dates yourself, because — as the next section shows — a clean-looking GNPA can lag the moment the loans actually start to sour.)

The trapdoor is the timing. NPAs always look lowest right before the cycle turns, because loans go bad slowly. A bank can post a gorgeous ROE and a clean GNPA for years by lending recklessly — the rot only shows up later. So you never read GNPA from a single quarter. You watch its direction across the cycle. (This is a read on asset quality now, not a lifetime verdict on the bank.)

Watch the margin guidance, not just the margin

The four numbers — P/B, ROE, NIM, GNPA — are a snapshot. What turns a snapshot into a judgement is whether management's commentary about where they're heading actually comes true.

NIM is the one to watch, because it drives ROE, which justifies P/B. When deposit costs rise faster than loan rates, NIM compresses and the whole valuation chain weakens. On a recent earnings call (concall), ICICI Bank's management guided that NIM would be "range-bound from here on" (Inve data, 2026). Whether a bank's NIM actually troughs and recovers as guided — or quietly keeps slipping while management changes the subject — is exactly the kind of thing worth tracking quarter after quarter. Doing that by hand, across a 10–15 stock portfolio, is the real grind; it's the job Inve's Promise Tracker was built to take off your plate.

What a five-year owner must believe

So put the strange shop back together. You're buying a business whose inventory is money, funded many times over with other people's savings — ICICI Bank itself carries roughly ₹8 of assets for every ₹1 of equity (Inve data, 2026), and weaker banks run far higher — where one bad lending decision can swallow a large slice of the owners' capital. You don't value it on P/E and EBITDA. You value it on what it owns net of what it owes (book value), what you pay for that (P/B), what it earns on it (ROE), the spread that drives those earnings (NIM), and whether the shelves are full of fresh loans or quiet rot (GNPA).

A five-year owner of any bank must believe three plain things: the loans on the shelf will get repaid, the spread won't collapse, and the people running a heavily leveraged machine are honest and unhurried. Get those right and the ratios take care of themselves. Get the first one wrong, and no P/B was ever low enough to save you.

Frequently asked questions