How to value a bank

In almost any company, debt is a capital structure decision and gets subtracted to arrive at equity value. In a bank, debt is the raw material of the business: deposits and wholesale funding are the input with which it lends. Subtracting them corrects nothing — it destroys the calculation. That is why a bank is valued on its equity and not on its enterprise value.

The mistake of carrying enterprise value over

Applying an enterprise value multiple to a bank produces a meaningless number: it subtracts from the valuation the very thing that generates the revenue. The same holds for an insurer, where float — premium money collected before claims are paid — is the company's own funding rather than excess cash.

The two metrics that do apply are a multiple on normalised earnings and a multiple on book value, particularly tangible book value. The second exists because a bank's equity is its regulatory cushion: how much is paid per unit of own capital is an economically meaningful question in a way it would not be for an industrial company.

Return on tangible common equity is the bar

The quality of a bank is not measured with return on invested capital but with return on tangible common equity: profit attributable to ordinary shareholders over equity, net of goodwill and intangibles. Netting those matters because in a bank that grew by acquiring other banks, book equity carries the price paid in those deals, and measuring against it flatters the return.

The bar it is compared against is absolute, not relative to the volatility of the share: above 20% is exceptional, 15% to 20% excellent, 10% to 15% good, and below 10% falls short of the opportunity cost of simply owning the market.

Normalising expected-loss provisioning

Current accounting requires provisioning, at the moment a loan is originated, for the loss expected over its entire life. In a fast-growing lender this pulls into the present losses that will occur over years, and depresses accounting profit precisely when the business is doing well. Reported earnings from an expanding book understate its steady-state profitability.

To value it, that provision has to be normalised to a steady-state loss, approximated with normalised net charge-offs rather than the period's provision. And the signals that say whether the normalisation is defensible have to be read: coverage of the book, delinquency and its trend, and net interest margin after losses.

What exit multiple is reasonable

A bank is leveraged by design and exposed to the credit cycle, so it does not sustain a platform business's multiple. The reference band on normalised earnings runs from ten to fifteen times, and a high-quality niche franchise can approach sixteen; the multiple on book value typically sits between one and a half and two and a half times depending on sustainable return.

The sanity check is the implied yield: the inverse of the multiple is the earnings yield. Twenty times equals 5%, which is only justified if the business will keep growing; for a lender it is almost always too much.

One note on buybacks: when valuing on a per-share multiple, earnings per share already incorporate the falling share count. The target price is shown per share, without dividing again, or the same effect would be counted twice.

Where this shows up in the analyses

Three different shapes of the same method, each with its derivation:

Keep reading