Valuing financials — why DCF breaks

Everything on this shelf so far has quietly assumed a certain kind of company: one that borrows money to finance operations that produce cash, so that you can value the operations, then subtract the debt, and arrive at what the shares are worth. For most businesses that separation — operations here, financing there — is clean and useful. For a bank, an NBFC (a non-bank lender that does much of what a bank does without a banking licence), or an insurer, it collapses entirely.

The reason is a single, disorienting fact: for a financial company, debt is not how it pays for the business — debt is the business. A bank takes in deposits (which are debt it owes you) and lends them out; the gap between what it pays depositors and what it charges borrowers is its product. Borrowing cheap and lending dear is not the financing behind the operation — it is the operation. So the moment you try to "value the operations and subtract the debt," you find there is nothing left, because the debt was the raw material you were trying to set aside.

This is why a DCF and every enterprise-value multiple quietly break for financials. There is no clean free cash flow to discount, no sensible enterprise value to compute, and EV/EBITDA becomes gibberish. This module shows exactly why the familiar tools fail, and what to reach for instead: valuing the equity directly, through the relationship between price-to-book and returns, and — for insurers — through embedded value.

Take an ordinary manufacturer. It buys steel, runs machines, sells goods, and generates operating cash. If it borrows to build a factory, that debt is a choice about how to finance an operation that would exist and produce cash either way. You can strip the financing out, value the cash-producing operation on its own — its free cash flow, the cash left after running and reinvesting in the business — and then subtract the debt to get to equity. The separation is real because the operation and the financing are genuinely different things.

Now take a bank. Its "raw material" is money, and it acquires that raw material by borrowing — from depositors, from bond markets, from other banks. Its "product" is also money, lent out at a higher rate. Interest paid is its cost of goods; interest earned is its revenue; the difference, the net interest margin (the spread between the rate a lender earns on loans and the rate it pays for funds), is its gross profit. There is no version of a bank with the "financing" removed, because removing the borrowing removes the raw material and there is no business left to value.

Three consequences follow, and each kills a familiar tool. First, EBITDA is nonsense for a bank, because it adds back interest — deleting the bank's single largest and most essential cost. Second, enterprise value is nonsense, because it nets debt against cash as if debt were incidental, when for a bank debt and cash are the inventory. Third, free cash flow has no clean meaning, because a growing bank must retain capital to support a bigger loan book and to satisfy regulators — the faster it grows, the more capital it swallows, so "free" cash is not free at all. Value is still the discounted worth of cash returned to owners — but for a bank the only cash you can honestly discount is what is left after it has held back the capital its growth and its regulator demand.

Since you cannot value a bank's enterprise, you value its equity. The workhorse relationship links two numbers: price-to-book (the share price divided by book value — the accounting net worth, assets minus liabilities, per share) and return on equity (ROE — annual profit as a percentage of that book value).

The logic is intuitive once you see it. Book value is the equity capital the bank stands on. ROE is how much profit each rupee of that capital earns. If a bank earns more on its capital than shareholders demand for the risk — its cost of equity, the return owners require to hold the shares — then each rupee of book is worth more than a rupee, and the bank deserves to trade above book. If it earns less than that hurdle, each rupee of book is worth less than a rupee, and it deserves to trade below book. A single tidy formula captures it:

Justified P/B = (ROE − growth) ÷ (cost of equity − growth).

Work two composite banks, identical except for ROE. Both have a cost of equity of 14% and grow at 8%. illustrative

• Sound Bank, ROE 16%: justified P/B = (16 − 8) ÷ (14 − 8) = 8 ÷ 6 = 1.33×. It earns above its 14% hurdle, so it should trade at a premium to book. On a book value of ₹500 per share, fair value ≈ ₹665.
• Thin Bank, ROE 11%: justified P/B = (11 − 8) ÷ (14 − 8) = 3 ÷ 6 = 0.50×. It earns below its 14% hurdle, so each rupee of book destroys value, and it should trade at a discount. On the same ₹500 book, fair value ≈ ₹250.

Same book value, same growth, same required return — and fair values that differ by more than two-and-a-half times, entirely because of ROE. This is the whole secret of reading a bank's valuation: P/B is simply ROE wearing a price tag. A high P/B is not "expensive" and a low one is not "cheap" until you know the ROE that earns it.

Watch the method flex across the three kinds of financial. illustrative

A bank. Sound Bank trades at 1.6× book while earning a 16% ROE against a 14% cost of equity. Your justified P/B said 1.33×, so at 1.6× the market is paying up — it may be assuming the ROE rises, or the growth lasts longer, than your inputs allow. The useful question is not "is 1.6× expensive?" in the abstract but "what ROE and durability does 1.6× require, and are they believable?" You have turned a vague price into a testable expectation. But before trusting even the ROE, you check its quality: a bank can flatter ROE for years by under-providing for bad loans, so a high ROE built on thin provisioning is a borrowed number that the next credit cycle reclaims.

An NBFC. The same P/B-against-ROE logic applies, with two extra worries magnified. An NBFC borrows wholesale rather than from sticky retail depositors, so its funding can vanish in a liquidity squeeze; and its ROE leans harder on leverage. Two NBFCs with identical 15% ROEs are not equally valuable if one reaches it on modest leverage and stable funding and the other on aggressive leverage and flighty borrowing. The multiple looks the same; the risk beneath it does not.

An insurer. Here even book value misleads, because an insurer's real worth is largely the future profit locked inside policies it has already sold — profit that accounting recognises only slowly, over decades. So insurers are valued on embedded value: the company's net worth plus the present value of the profits expected from its in-force policies. A life insurer might trade at a multiple of embedded value, just as a bank trades at a multiple of book — and just as with a bank, that multiple is only sane if the returns and growth behind it are real.

The equity-based tools are the right tools — and they still hide the very things that sink financial companies.

They cannot see the quality of the loan book. Book value assumes the assets are worth what the accounts say. For a lender, the assets are loans, and their real worth depends on how many will be repaid — which is unknown until a downturn tests them. A bank at "0.9× book" is only cheap if the book is honest; if a fifth of the loans are quietly bad, the true book is far smaller and the "discount" is imaginary.

They cannot see the leverage risk until it bites. A bank stands on a thin sliver of equity beneath a mountain of deposits. A small percentage of loans going bad can wipe out a large slice of equity. ROE looks serene right up to the point where the leverage that magnified the good years magnifies a bad one into a crisis.

They cannot judge the accounting assumptions inside embedded value. An insurer's embedded value rests on projections — how long policyholders keep paying, how investments perform, how many claims arrive — stretched over decades. Small, optimistic tweaks to those assumptions inflate embedded value quietly and enormously. The number looks precise; its foundations are forecasts. A rough, conservative read of a bank's through-cycle earnings power beats a precise valuation built on this year's flattering ROE and an untested loan book.

Financials punish the reader who brings ordinary-company habits to them. The recurring traps:

1. Quoting EV or EV/EBITDA for a bank. The single clearest tell that someone has not understood financials. Interest is the bank's cost of goods; adding it back and netting debt against cash produces a number with no meaning.

2. Calling a low P/B "cheap" without checking ROE. A 0.5× book bank may be perfectly priced if it earns below its cost of equity. Low P/B is a discount the business has earned through weak returns as often as it is a bargain the market has missed.

3. Trusting a high ROE without testing provisioning. The easiest way to manufacture a high ROE for a few years is to under-provide for bad loans. The reported number gleams; the credit cycle sends the invoice later.

4. Ignoring how the ROE is levered. Two banks at 15% ROE differ enormously if one is modestly geared with stable retail funding and the other is aggressively geared on wholesale money. The ratio hides the fragility.

5. Taking embedded value as a hard number. It is a projection over decades, exquisitely sensitive to assumptions the company itself chooses. Treat it as an estimate with a wide range, not a fact.