Why not P/E, and definitely not DCF

For most companies this blog has anchored valuation on P/E and, when the cash flows could be forecast, a DCF. Neither works cleanly for a bank.

A DCF needs free cash flow, and for a bank that number is close to meaningless: taking in deposits looks like a cash inflow, lending them out looks like an outflow, and the “capex” line is nearly empty. P/E works, but it sits on top of a profit figure that the provisions post showed to be partly discretionary and strongly cyclical — the same bank can post half the profit in a bad credit year with no change in franchise.

What a bank does have is a balance sheet where most assets are financial — loans, bonds, cash — carried at, or close to, what they are worth. A factory’s book value is an accounting artefact of old purchase prices and depreciation schedules. A bank’s book value is a reasonable first estimate of what the shareholders actually own. That is why bank valuation starts from price-to-book.

The formula, and the identity behind it

P/B  =  Price per share  /  Book value per share

and, because P/E = Price / EPS and ROE = EPS / Book value per share,

P/B  =  P/E  ×  ROE

That identity is the key to the whole subject. A bank’s P/B is high when the market pays a lot for each rupee of earnings and the bank earns a lot on each rupee of book. Two banks with the same P/E can have wildly different P/B purely because one earns a 16% return on its equity and the other 8%.

Worked example: HDFC Bank at 30 June 2025

Financials from HDFC Bank’s results for the year ended 31 March 2025 (standalone). Price: NSE (National Stock Exchange) close on 30 June 2025, from Yahoo Finance. Historical, for illustration only.

One wrinkle first. HDFC Bank issued 1:1 bonus shares (record date 27 August 2025), so price databases now show pre-bonus dates at half the price actually traded: Yahoo shows ₹1,000.75 for 30 June 2025, when the share changed hands at about ₹2,001.5. The FY25 EPS and book value are on the pre-bonus share count, so the pre-bonus price is the consistent one to pair with them. Mixing an adjusted price with unadjusted per-share figures would halve every multiple below.

 
Shareholders’ equity (capital + reserves), 31 Mar 2025 ₹4,97,619.43 crore
Shares outstanding (face value ₹1 → capital ₹765.22 crore) 765.22 crore
Book value per share ₹650.3
Price, 30 June 2025 (pre-bonus basis) ₹2,001.5
P/B 3.08x
EPS (basic), FY25 ₹88.29
P/E 22.7x
ROE on closing equity (PAT / equity) 13.5%
P/E × ROE (identity check) 3.06x
Market capitalisation ≈ ₹15.32 lakh crore
Dividend yield (₹22 FY25 dividend) 1.1%

The identity holds to rounding: 22.7 × 13.5% ≈ 3.06x. (Using ROE on average equity — 14.4% — would shift both terms slightly; the identity only closes exactly when P/E, ROE and book value share one basis.)

What justifies a P/B above 1

A bank trading at book value is being valued at what its shareholders nominally own. Above 1, the market is paying for the return the bank earns on that book being higher than what shareholders could demand elsewhere. Below 1, it is saying the opposite — or that the book value isn’t real.

The textbook expression, from the dividend-discount model:

Justified P/B  =  (ROE − g)  /  (Ke − g)

where  Ke = cost of equity (see the WACC post)
       g  = sustainable long-run growth

If a bank earns exactly its cost of equity (ROE = Ke), the fraction is 1 — it deserves to trade at book. Every point of ROE above Ke pushes the justified multiple up; every point of Ke above ROE pushes it below 1.

Here is the formula applied to a generic bank earning a 15% ROE — a round number in the neighbourhood of large Indian private banks, not HDFC Bank’s — with illustrative pairs of the two inputs nobody can observe. Growth is kept at or below the roughly 10% long-run nominal growth ceiling from the terminal value post, and the last row is the aggressive case, with g creeping up to within two points of Ke:

Scenario Cost of equity Growth Justified P/B
A 14.0% 6.0% 1.12x
B 13.0% 7.0% 1.33x
C 12.0% 8.0% 1.75x
D (aggressive: g within 2 points of Ke) 12.0% 10.0% 2.5x

Read that table for what it actually demonstrates. One percentage point on each of two unobservable inputs (A to B) moves the “justified” multiple from about 1.12x to 1.33x. Two points on each (A to C) takes it to 1.75x, and pushing g close to Ke (D) gets you 2.5x — the same g-too-close-to-r blow-up the terminal value post warned about. Same bank, same ROE, more than double the answer from inputs nobody can observe. The formula is a way to think about why multiples differ. It is not a way to decide whether one is right, and this post doesn’t try. (The margin of safety post made the same point about DCF inputs; banks just make it faster.)

ROA × leverage: where the ROE comes from

The DuPont post split ROE into margin, turnover and leverage. For a bank the first two collapse into one number — return on assets — and the decomposition becomes:

ROE  =  ROA  ×  (Average assets / Average equity)
FY25
ROA (PAT / average total assets) 1.79%
× Leverage (average assets / average equity) 8.0x
= ROE 14.3% (direct: 14.4%)

This is the most useful two-line summary of any bank. ROA says how well it lends: HDFC Bank earned about ₹1.79 on every ₹100 of assets. Leverage says how many times that thin return is multiplied for shareholders: about 8 times. The capital adequacy post is what caps the second number; the NIM, CASA and provisions posts are what drive the first.

Two banks with the same ROE can be very different animals: one earning 2% on assets at 7x leverage, another earning 1% at 14x. Same ROE, half the cushion. Always split it.

🧒 Explain it like I'm 10 (optional — skip if this is already clear)

Imagine two piggy banks. Each has ₹100 inside — that’s the book value.

Piggy A somehow turns its ₹100 into ₹115 every year. Piggy B only manages ₹107. If someone offered to sell you Piggy A for ₹150 and Piggy B for ₹150, you’d obviously pick A — it earns more on the same ₹100.

So people will pay more than ₹100 for Piggy A, and how much more depends on how much better than “normal” its ₹15 a year is. That’s P/B: price over the ₹100 inside, justified by how hard the ₹100 works.

Where P/E still helps

P/E isn’t useless for banks; it’s a cross-check. When P/B and P/E disagree — a high P/B with a modest P/E — it means ROE is unusually high, and the question becomes whether that ROE is sustainable or a good year in the credit cycle. When P/B is low and P/E is high, profits are depressed and the market is valuing the book, not the year. Read the pair together.

Book value itself also deserves scrutiny. It is “close to fair” only if the loans are worth what they’re carried at — which loops straight back to provision coverage. A bank with thin coverage and a rising NPA book has a book value that is partly fiction, and a low P/B on a fictional book is not cheap.

What the filings can’t show you

Everything in this module comes from what a bank publishes, and some of the biggest risks barely show up there. The mismatch between long-dated loans and short-term deposits (asset-liability management) is disclosed only in coarse maturity buckets, and stress the bank hasn’t yet recognised — loans still “standard” but quietly deteriorating — is invisible until it’s classified, which is why bank crises so often start with a clean-looking balance sheet.

Common mistakes

  • Comparing P/B across banks with different ROE. P/B = P/E × ROE. A 3x bank at 15% ROE and a 1x bank at 5% ROE can be on the same P/E. The multiple alone says nothing.
  • Reading P/B below 1 as a bargain. It may mean the market doubts the book — under-provisioned loans, or an ROE below cost of equity that destroys value every year.
  • Mixing adjusted and unadjusted figures. Bonus issues and splits halve the price in databases but not in last year’s filing. Check the basis before dividing.
  • Trusting a justified-P/B formula. As the table shows, two guessed inputs can produce almost any answer. Use it to understand why multiples differ, never to conclude that one is wrong.
  • Ignoring leverage inside ROE. The same ROE from twice the leverage is not the same quality of return. Split it into ROA and leverage every time.

Takeaway: Banks are valued on book because their assets are mostly money carried near what it’s worth, and P/B = P/E × ROE ties the multiple to how hard that book works. Split ROE into ROA and leverage every time, and treat any “justified P/B” formula as a way to see why multiples differ, not to decide whether one is right.